Apple ML Research explores LLM agents, translation, speech, and safety
ByPulseAugur Editorial·[301 sources]·
Apple's Machine Learning Research division has published several papers detailing advancements in AI. One paper introduces SelfCompact, a system that allows language model agents to autonomously manage their context by deciding when and how to compact information, improving efficiency and reducing token costs on benchmarks. Another study evaluates Large Language Models (LLMs) for machine translation quality annotation, finding that while LLMs show promise, their agreement with human annotators varies significantly across tasks and languages, highlighting the need for collaborative annotation pipelines. Additionally, research demonstrates that enhancing language discrimination in multilingual speech models can reduce the performance gap compared to monolingual models, and a new framework called TRACE is proposed for safer language generation by assessing and editing risks during the generation process. Other papers explore methods for improving LLM decision-making, managing temporal validity in parametric memory, and addressing numerical data failures in language models.
AI
IMPACT
These papers advance LLM capabilities in context management, translation evaluation, multilingual speech, decision-making, safety, and memory, potentially influencing future AI development.
RANK_REASON
Cluster consists of multiple research papers published by Apple's ML Research and on arXiv, detailing new methods and findings in LLM capabilities and safety.
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Research
Cluster consists of multiple research papers published by Apple's ML Research and on arXiv, detailing new methods and findings in LLM capabilities and safety.
Source corroboration
301 independent sources
Strong cross-source corroboration — multiple independent publishers covered this within the clustering window.
Topics
paper, model release, safety
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
69 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.
Coverage growth since scoring
+80 source(s) since last score
New sources have picked up this story since our last re-score. Score will update on the next scoring pass.
Long agent traces composed of chains of thought and tool calls accumulate stale content that anchor subsequent generations, and eventually outgrow the context window. Existing scaffolds mitigate it with fixed-interval compaction triggered at a token threshold. Such triggers pay n…
Apple Machine Learning Research
TIER_1English(EN)·
Large Language Models (LLMs) are considered to be a more efficient and cost-effective alternative to human judgment for Machine Translation (MT) evaluation. With MT evaluation spanning a large number of language pairs, domains and levels of annotation granularity, LLMs must be th…
Apple Machine Learning Research
TIER_1English(EN)·
Multilingual self-supervised speech models can benefit from sharing information across languages, but under a matched total pretraining data budget they still fall short of monolingual models. We show that strengthening the model’s ability to discriminate languages during pretrai…
arXiv:2610.22636v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used in critical domains, where their susceptibility to hallucinations makes reliable and interpretable uncertainty quantification essential. We introduce the concept of mechanistic un…
arXiv:2610.23722v1 Announce Type: new Abstract: Understanding and simulating how subjective cognition and mental states emerge, evolve, and shape behavior is essential for advancing artificial intelligence toward deeper social cognition and general intelligence. Existing large la…
arXiv:2610.22425v1 Announce Type: cross Abstract: Safety-aligned language models are expected to refuse harmful requests while remaining useful for benign ones. However, because many evaluations score only the final response, they often conflate intent understanding with response…
arXiv:2610.23641v1 Announce Type: new Abstract: We introduce no-regret generation, which requires a generator to make a sublinear number of mistakes uniformly over all demonstration sequences and compatible target languages. We also formulate an agnostic variant that allows arbit…
arXiv:2610.23251v1 Announce Type: new Abstract: Restoring service after distribution-grid faults requires decisions from incomplete and conflicting information while respecting electrical constraints. Addressing this need, we introduce FLARE (Feeder-Level Agent Restoration Evalua…
arXiv:2610.22754v1 Announce Type: new Abstract: LLM safety evaluations via jailbreaking typically index on attack success rate (ASR): whether a generated response violates safeguards. However, ASR only evaluates sampled outputs and provides little information on the underlying ge…
arXiv cs.CL
TIER_1English(EN)·Andrey Goncharov, Nikolai Kondusov, Alexey Zaytsev·
arXiv:2510.13849v4 Announce Type: replace Abstract: Multilingual Large Language Models (LLMs) often exhibit hallucinations such as unintended code-switching, reducing reliability in downstream tasks. We propose latent-space language steering, a lightweight inference-time method t…
arXiv:2610.23225v1 Announce Type: new Abstract: Drifting Models generate a sample with a single function evaluation, making them attractive for fast parallel language generation, but their one-step formulation does not support iterative refinement. We propose Recurrent Language G…
arXiv cs.CL
TIER_1English(EN)·Qiang Yan (Institute of Computing Technology, Chinese Academy of Sciences)·
arXiv:2610.23170v1 Announce Type: new Abstract: Many language-based applications require probabilities over a finite set of decisions rather than open-ended responses. We present Chinese Decision, a family of Chinese-language models for single-pass probabilistic decision-making. …
arXiv:2610.23157v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used as computational models of human behavior. Recent models such as Centaur have advanced this goal by fine-tuning LLMs on trial-by-trial decisions from psychological experiments. Howe…
arXiv:2610.22980v1 Announce Type: new Abstract: Large language model (LLM) agents are increasingly evaluated on long-horizon tasks, including autonomously running a business. Existing benchmarks for this setting are almost exclusively English-language, Western in retail context, …
arXiv:2610.22961v1 Announce Type: new Abstract: Brain alignment of large language models (LLMs) offers a way to investigate representational similarities between LLMs and the human brain during language processing. While previous studies have mainly focused on passive language co…
arXiv:2610.22889v1 Announce Type: new Abstract: Formulating optimization problems from raw inputs of problem descriptions and data is time-consuming and costly. While Large Language Models (LLMs) offer a promising way to automate this process, existing benchmarks overestimate the…
arXiv:2610.22807v1 Announce Type: new Abstract: Recent studies have shown that position bias remains a critical vulnerability in large language models (LLMs), undermining both the integrity of multiple-choice evaluations and the reliability of retrieval-augmented generation. Howe…
arXiv:2610.22721v1 Announce Type: new Abstract: Large language model (LLM) unlearning aims to forget certain information while preserving others, without retraining from scratch. SoTA methods such as NPO, SimNPO, and SatImp balance forgetting and retain through whole model fine-t…
arXiv:2610.22468v1 Announce Type: new Abstract: Action nominal constructions (ANCs) express events in nominal form, with their argument marking following patterns found in clauses or ordinary noun phrases. In this paper, we investigate how different grammatical configurations aff…
arXiv:2610.22428v1 Announce Type: new Abstract: In a small language model the vocabulary is the single largest part. In Gemma 3 270M the embedding table holds 168M of 268M parameters, and because quantisation methods leave the embedding in 16-bit, it is also most of the file once…
arXiv cs.AI
TIER_1English(EN)·Mayank Mishra, Shawn Tan, Ion Stoica, Joseph E. Gonzalez, Tri Dao·
arXiv:2603.14360v3 Announce Type: replace-cross Abstract: Transformers are highly parallel but are limited to computations in the TC$^0$ complexity class, excluding tasks such as entity tracking and code execution that provably require greater expressive power. Motivated by this …
arXiv cs.AI
TIER_1Deutsch(DE)·Benaya Trabelsi, Jonathan Shaki, Sarit Kraus·
arXiv:2601.13749v2 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly employed for decision-support across multiple domains. We ask whether these models favor artificial intelligence (AI) itself. We examine this question through three experiments…
arXiv cs.AI
TIER_1English(EN)·Adrians Skapars, Edoardo Manino, Youcheng Sun, Lucas C. Cordeiro·
arXiv:2507.01693v3 Announce Type: replace-cross Abstract: Recent work shows that distinct inputs produce distinct hidden representations in LLMs, so the input can in principle be recovered from them. Yet, no existing method can reliably do so from the model's output text or logit…
arXiv:2604.00445v2 Announce Type: replace Abstract: Uncertainty estimation (UE) aims to quantify uncertainty in large language model (LLM) outputs to flag potential hallucinations and support reliable deployment. However, UE metrics often exhibit unstable performance across model…
arXiv:2504.19599v4 Announce Type: replace Abstract: Post-training plays a crucial role in refining and aligning large language models to meet specific tasks and human preferences. While recent advancements in post-training techniques, such as Group Relative Policy Optimization (G…
arXiv cs.AI
TIER_1English(EN)·Dongyoon Hahm, Woogyeol Jin, Dennis Wei, Yi Zhou, Nathalie Baracaldo, Kimin Lee·
arXiv:2610.23847v1 Announce Type: cross Abstract: Fine-tuning large language models (LLMs) can induce behaviors beyond those explicitly targeted. For example, training a model to respond more warmly can make it more sycophantic, and training it to write insecure code can cause br…
arXiv:2610.23625v1 Announce Type: cross Abstract: Presence-aware intelligence is the capacity of an AI system's intelligence layer to adapt appropriately to the functional state of the person it serves (here readiness, flow, and stability) while preserving task quality and safety…
arXiv cs.AI
TIER_1English(EN)·Gabriel Hortea, Simon Pietro Romano, Juan Tapiador, Mario Varlese·
arXiv:2610.23534v1 Announce Type: cross Abstract: Small Language Models (SLMs) are increasingly attractive for resource-constrained environments due to their low computational footprint. These characteristics also make them candidates for future autonomous malware, where connecti…
arXiv cs.AI
TIER_1English(EN)·Ponhvoan Srey, Jason Lucas, Pittawat Taveekitworachai, Junyoung Lee, Anh Tuan Luu·
arXiv:2610.23527v1 Announce Type: cross Abstract: Large Language Models (LLMs) encode substantial factual knowledge. Yet, estimating confidence in factual claims reliably remains challenging. We study likelihood-based factual confidence from the complementary perspectives of disc…
arXiv:2610.23344v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used to automate AI research, including language-model (LM) development, offering a potential path toward recursive self-improvement. However, turning experimental feedback into a lear…
arXiv:2610.23190v1 Announce Type: cross Abstract: Neural combinatorial optimization (NCO) rapidly produces high-quality solutions for problem structures encountered during training, but operational requirements may change after deployment and be communicated in natural language. …
arXiv cs.AI
TIER_1English(EN)·Basar Demir, Emily Shen, Ksheeraj Sai Vepuri, Sourabh Kulkarni, Harihar Subramanyam, Ankit Jain, Nan Jiang, Chirag Nagpal·
arXiv:2610.22942v1 Announce Type: cross Abstract: Small variations in prompt characteristics can meaningfully alter the semantics, tone, and contextual assumptions of large language model outputs. Such characteristics can involve group variables such as demographics or other sens…
Large language model (LLM) agents are increasingly evaluated on long-horizon tasks, including autonomously running a business. Existing benchmarks for this setting are almost exclusively English-language, Western in retail context, and single-agent, with the agent observing the b…
arXiv:2610.21810v1 Announce Type: cross Abstract: Conditional memory expands language-model capacity through selective lookup, but the full embedding table must remain available for every possible input. We introduce Conditional Program (CP), a compressed representation that reco…
arXiv:2610.21230v1 Announce Type: new Abstract: Unlearning aims to remove selected knowledge from large language models to mitigate privacy and safety risks while preserving useful capabilities. For models that continue to be fine-tuned after release, this goal is insufficient: t…
arXiv:2610.16717v2 Announce Type: replace Abstract: Does a large language model have a self, and if so, of what kind? We separate two functional levels familiar from biology: a self-index, which distinguishes one's own material from external material and shapes processing, as cor…
arXiv cs.CL
TIER_1English(EN)·Xin Xu, Clive Bai, Kai Yang, Tianhao Chen, Cai Yuchen, Yangkun Chen, Weijie Liu, Hao Chen, Yang Wang, Saiyong Yang, Can Yang·
arXiv:2602.12036v3 Announce Type: replace Abstract: Large-scale verifiable prompts underpin the success of Reinforcement Learning with Verifiable Rewards (RLVR), but they contain many uninformative examples and are costly to expand further. Recent studies focus on better exploiti…
arXiv:2601.21132v2 Announce Type: replace Abstract: I demonstrate that large language models can infer ethnicity from names with accuracy exceeding that of Bayesian Improved Surname Geocoding (BISG) without additional training data, enabling inference outside the United States an…
arXiv:2610.21256v1 Announce Type: cross Abstract: Language models are often asked to make judgments: whether a proposed answer is correct, whether one statement follows logically from another, or whether a response is safe. Yet changing the instructions that specify how the judgm…
arXiv:2610.22169v1 Announce Type: new Abstract: The vocabulary of a language model is its native semantics: the embedding and the unembedding fix the meaning of a state at the two ends of the residual stream. Between them, existing methods either reuse a fixed lens or treat the s…
arXiv:2610.22131v1 Announce Type: new Abstract: Interactive agents repeatedly choose among a finite set of actions, labels, or ordinal outcomes. Generating a free-form response for each choice adds latency and complicates the interface between a model and its environment. We pres…
arXiv cs.CL
TIER_1Deutsch(DE)·Kalyan Bhetwal, Ceren Mert, Ian Whitehouse·
arXiv:2610.22053v1 Announce Type: new Abstract: Names carry demographic indicators such as age, gender, and race, which makes them a subtle but strong target of bias in large language models (LLMs). In this work, we present an evaluation of name-based bias across 6 local large la…
arXiv:2610.21900v1 Announce Type: new Abstract: An agent's memory must remain useful as information changes across conversations. A new fact may replace an earlier value without making that value irrelevant to a later question. We present Temporal State Tracking Memory (TEM), a t…
arXiv:2610.21681v1 Announce Type: new Abstract: Data strategy plays a central role in training large language models (LLMs). It governs the selection, weighting, and scheduling of training data, thereby determining the capabilities a model can acquire from a fixed data budget. We…
arXiv cs.CL
TIER_1English(EN)·Antonio Lopardo, Luis Frentzen Salim, Shikhar Srivastava, Kevin Jiang, Mridul Sharma, Catherine Arnett·
arXiv:2610.21679v1 Announce Type: new Abstract: Tokenization is a critical component of the language modeling pipeline, but it is poorly understood. Tokenizers can impact downstream performance, but there is currently mixed evidence about which tokenizer features predict a better…
arXiv:2610.21383v1 Announce Type: new Abstract: Block Influence (BI) ranks Transformer blocks for depth pruning from their input-output cosine change, but the ranking depends on the calibration distribution. We study whether English calibration transfers to Indian languages in Qw…
arXiv:2610.21304v1 Announce Type: new Abstract: Current value alignment evaluations for large language models (LLMs) heavily rely on output compliance, overlooking whether LLMs maintain value consistency when alternating between generator and evaluator roles. Such consistency is …
arXiv:2610.21165v1 Announce Type: new Abstract: Long-context capability has become essential for large language models (LLMs) to handle complex tasks, yet continuously expanding the context window does not remove the computational, storage, and information-selection pressures inc…
arXiv:2610.20808v1 Announce Type: new Abstract: Recent literature shows that unsafe tool calls can often accompany textual refusals in language models. To investigate this hidden vulnerability, we introduce PRISM, a framework for evaluating the transfer of activation intervention…
arXiv:2409.12500v2 Announce Type: replace-cross Abstract: Large language models have become increasingly popular and demonstrated remarkable performance in various natural language processing (NLP) tasks. However, these models are typically computationally expensive and difficult…
arXiv:2605.06898v2 Announce Type: replace Abstract: At the heart of a language model agent is an orchestrator program which is responsible for the state transition between consecutive turns. In existing systems, this program is implemented in the harness. This paper introduces se…
arXiv:2610.22099v1 Announce Type: cross Abstract: Masked diffusion language models expose an inference-time choice: which positions become visible after each denoising pass. Many existing decoding methods use hand-designed rules or require visibility to grow monotonically. We pre…
arXiv cs.AI
TIER_1English(EN)·Bruno Arcanjo, Zhaoxing Li, Sebastian Stein·
arXiv:2610.21938v1 Announce Type: cross Abstract: An agent that learns a user's preferences from their choices faces a cold start problem: before any choices are observed, it must rely solely on a prior. Informative priors are usually fitted to historical choice data, which new t…
arXiv:2610.21176v1 Announce Type: cross Abstract: Large language models (LLMs) reproduce a human working-memory signature on the N-back task: performance declines as memory load N increases, and errors concentrate on lures, items that repeat the current stimulus at a neighboring …
arXiv cs.AI
TIER_1English(EN)·Matteo He, William F. Shen, Alex Iacob, Andrej Jovanovic, Xinchi Qiu, Nicholas D. Lane·
arXiv:2610.21025v1 Announce Type: cross Abstract: When does a language model acquire a capability? Developmental studies usually answer this by tracking when models start producing the correct token. But that prediction passes through the unembedding matrix, or readout, which is …
arXiv:2610.20824v1 Announce Type: cross Abstract: Wolof is spoken by 23M people, mostly in Senegal, but remains weakly supported by open-weight language models, while API models raise cost, governance, and sovereignty constraints for local deployments. In this paper, we present 0…
arXiv:2610.22068v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) can act unsafely when individually benign images and texts are interpreted jointly. Beyond implicit reasoning risk, where reasoning drifts toward harm, we identify intent-level risk, where cr…
arXiv:2610.21808v1 Announce Type: new Abstract: Weight quantization below four bits shrinks memory and bandwidth at a perplexity that looks acceptable, but often at the cost of factual knowledge and task accuracy. Ternary weights have mostly been obtained by training from scratch…
arXiv:2610.21286v1 Announce Type: new Abstract: We investigate whether variation in information density across chain-of-thought (CoT) reasoning steps predicts answer correctness and can serve as a controllable inference-time target. We prompt frozen pretrained models to solve pro…
arXiv:2610.21195v1 Announce Type: new Abstract: On-policy distillation (OPD) improves a student language model by training on its own generated trajectories with supervision from a stronger teacher. However, when conditioned on student-generated prefixes, the teacher can favor in…
arXiv:2610.21045v1 Announce Type: new Abstract: Tool-using language agents must revise prepared actions when evidence changes or authority is revoked. Selective withdrawal means suspending affected actions, preserving supported work, and resuming only after sufficient repair. We …
Evaluating language models and agents requires combining benchmark accuracy, judge-based assessment, runtime performance, tool-use behavior, and retrieval quality. Differences in interfaces, metric semantics, and trace formats complicate these comparisons and can hide failures in…
Interactive agents repeatedly choose among a finite set of actions, labels, or ordinal outcomes. Generating a free-form response for each choice adds latency and complicates the interface between a model and its environment. We present Intern-Decision, a family of 0.8B, 2B, and 4…
Multimodal large language models (MLLMs) can act unsafely when individually benign images and texts are interpreted jointly. Beyond implicit reasoning risk, where reasoning drifts toward harm, we identify intent-level risk, where cross-modal ambiguity silently commits the model t…
Names carry demographic indicators such as age, gender, and race, which makes them a subtle but strong target of bias in large language models (LLMs). In this work, we present an evaluation of name-based bias across 6 local large language models spanning multiple providers and pa…
An agent that learns a user's preferences from their choices faces a cold start problem: before any choices are observed, it must rely solely on a prior. Informative priors are usually fitted to historical choice data, which new tasks and deployments lack. To address this, we pre…
Context compression in long-running applications of large language models (LLMs) must preserve goal-relevant information, often before specific downstream questions are known. We introduce GACC-Bench, a benchmark for Goal-Adaptive Context Compression that evaluates how reliably L…
Conditional memory expands language-model capacity through selective lookup, but the full embedding table must remain available for every possible input. We introduce Conditional Program (CP), a compressed representation that reconstructs only the rows an input requests. CP repre…
arXiv:2610.15365v1 Announce Type: new Abstract: Large language models (LLMs) can produce harmful content while generating a response. A useful safeguard should not only detect this risk during generation but also intervene while leaving safe responses largely unchanged and provid…
arXiv:2610.15235v1 Announce Type: new Abstract: Despite remarkable progress in language understanding and generation, decoder language models continue to exhibit striking failures on numerical data. Numerical examples are typically serialized as text by concatenating feature name…
arXiv:2610.15218v1 Announce Type: new Abstract: A recurrent-depth language model reuses one block of weights several times in a row. The number of repetitions, which we call loops, is chosen when the model runs, not when it is trained, and running fewer is cheaper and faster, so …
arXiv:2610.18566v1 Announce Type: new Abstract: Large language model (LLM) routers reduce serving cost by choosing which model to call and how many output tokens to grant it for each query. Existing routers rank model-budget actions using a single quality metric. This does not wo…
arXiv cs.AI
TIER_1English(EN)·Hanbo Xie, Yue Zhao, Robert C. Wilson·
arXiv:2610.17310v1 Announce Type: new Abstract: Language-model agents are expected to adapt to others whose behavior changes through interaction. Yet adaptation may rely on learned behavioral patterns rather than theory of mind (ToM): the ability to infer others' mental states, i…
arXiv:2610.17266v1 Announce Type: new Abstract: Diffusion language models are a promising and more efficient alternative to autoregressive models. However, most RL post-training techniques are not easily adapted to these models. One issue is that diffusion language models generat…
arXiv cs.AI
TIER_1English(EN)·Shuhul Razdan, R Balamurugan·
arXiv:2610.16325v1 Announce Type: new Abstract: When a steering intervention changes a language model's behavior, which part of the intervention deserves the credit? We test this question for a target-centered, scaled conceptor complement by removing its fitted covariance filter …
arXiv:2610.16141v1 Announce Type: new Abstract: LLM agents increasingly operate over multiple entities, where reliable response generation requires not only retrieving relevant evidence but also preserving its association with the correct entity. We identify Phantom Merge (PM), a…
arXiv cs.AI
TIER_1English(EN)·Sam Siavoshian, Omar Ramadan, Amir K. Saeed, Benjamin A. Johnson, Benjamin M. Rodriguez·
arXiv:2610.15882v1 Announce Type: new Abstract: A stored fact may stop being valid when a user changes a preference or an event replaces an earlier one. We study whether a language model can learn these changes in its parameters and answer correctly from only a question and the c…
arXiv:2610.15501v1 Announce Type: new Abstract: Large language models make decisions through parallel prefill and sequential decoding. Although effective, this approach incurs decoding costs and does not explicitly expose decision probabilities. Existing task-specific classifiers…
arXiv cs.AI
TIER_1English(EN)·Peizhi Niu, Evelyn Ma, Boya Hou, Huiting Zhou, Duo Zhou, Huan Zhang, S. Rasoul Etesami, Olgica Milenkovic·
arXiv:2506.10946v3 Announce Type: replace-cross Abstract: Unlearning large language models (LLMs) is important due to copyright and privacy concerns. A key challenge is unintended forgetting, where removing data inadvertently impairs model utility. While prior work has focused on…
arXiv cs.AI
TIER_1English(EN)·Daking Rai, Mor Geva, Ziyu Yao·
arXiv:2605.09129v2 Announce Type: replace Abstract: Circuit study aims to explain how language models (LMs) implement a task by localizing and interpreting a circuit: a computational subgraph responsible for a specific model's behavior. Existing methods for circuit localization a…
arXiv cs.AI
TIER_1English(EN)·Nicolas Martorell, Bruno Bianchi·
arXiv:2603.18893v3 Announce Type: replace Abstract: Language models are often asked to rate their internal states with a number, but it is unclear how reliable this measure is. We train probes for four affective concepts (wellbeing, interest, focus and impulsivity) in eight instr…
arXiv cs.AI
TIER_1English(EN)·Simon Henniger, Gabriel Poesia·
arXiv:2602.17831v3 Announce Type: replace Abstract: Evaluating the reasoning capabilities of Large Language Models is increasingly challenging as models improve. Human curation of hard questions is highly expensive, especially in recent benchmarks using PhD-level domain knowledge…
arXiv:2506.17585v4 Announce Type: replace Abstract: Large language models can answer questions correctly and still hallucinate their citations. Most citation systems address this by retrieving documents at inference time. Retrieval adds latency and depends on external infrastruct…
arXiv:2610.18674v1 Announce Type: cross Abstract: Large language models rely on prompt context to perform in-context learning, follow formats, and adapt to local evidence. This ability, however, also raises a robustness concern: models may reuse irrelevant tokens from earlier con…
arXiv:2610.18604v1 Announce Type: cross Abstract: 3D part assembly aims to reconstruct a complete shape from a set of unassembled parts by predicting a rigid SE(3) transformation for each part. Because the placement of a part is rarely determined by its own geometry, the task dem…
arXiv:2610.18520v1 Announce Type: cross Abstract: Autoregressive language models produce, at every prediction step, a vocabulary-sized logit vector whose structure determines the conditional next-token distribution. We empirically identify that the ordered vocabulary-head logits …
arXiv cs.AI
TIER_1English(EN)·Jophin John, Michael Hoffmann, Piroska Lendvai, Florent Dufour·
arXiv:2610.18343v1 Announce Type: cross Abstract: Post-training quantisation reduces the memory cost of large language models, but it remains unclear whether compression preserves language skills of regional varieties as reliably as their standard-language counterparts. We study …
arXiv:2610.17590v1 Announce Type: cross Abstract: Diffusion language models (DLMs) have attracted wide attention for their fast, parallel decoding, but they are believed to be far less efficient to train than autoregressive language models (ARMs): a widely cited study reports tha…
arXiv cs.AI
TIER_1English(EN)·Ayse Sila Okcu, Weisheng Wang, Tongmeng Xie, Niall Dalton, Angelos Nalmpantis, Raghu Arghal, Mario Giulianelli·
arXiv:2610.17444v1 Announce Type: cross Abstract: Language model (LM) agents can fail for different reasons, requiring different explanations and interventions. We study how state beliefs, action values, and action preferences interact as LM agents reason and commit to decisions …
arXiv:2610.16837v1 Announce Type: cross Abstract: Syntactic language models (SLMs) that generate sentences along with their syntactic tree structures have shown promise in both syntactic generalization and downstream tasks. Prior studies, however, have mostly evaluated smaller mo…
arXiv:2610.15415v1 Announce Type: cross Abstract: Before an agent acts, it must decide whether action is warranted. A single accuracy cannot distinguish failure to recognize a warranted action from reluctance to act. Signal-detection theory separates sensitivity from criterion, b…
arXiv:2610.14476v1 Announce Type: cross Abstract: Comparing large language models can reveal shared representational structure and systematic differences between them. However, aggregate similarity measures can obscure how agreement varies across geometric scales and which lingui…
arXiv:2610.14443v1 Announce Type: cross Abstract: Model cards, system cards, and safety reports increasingly document LLM safety behaviors, yet safety evaluations typically start from predefined benchmarks rather than asking which documented claims those benchmarks actually test.…
arXiv:2610.14007v1 Announce Type: cross Abstract: A familiar line of reasoning runs as follows: language models are flexible approximators trained under a strictly proper loss, so scale brings them closer to Solomonoff induction, and their predictive success is evidence of that a…
arXiv cs.AI
TIER_1English(EN)·Zhihua Liu, Kurt Butler, Damian Machlanski, Danilo Bueno, Sotirios A. Tsaftaris·
arXiv:2610.17439v1 Announce Type: new Abstract: The success of autonomous research systems relies on the quality of the scientific hypotheses they generate. Recent studies report that human reviewers often rate LLM-generated ideas as more novel than those from human experts, rais…
arXiv cs.LG
TIER_1English(EN)·Haofei Yu, Yixing Chen, Yining Zhao, Xuyan Ye, Yuxiang Liu, Zhigang Hua, Yan Xie, Shuang Yang, Jiaxuan You·
arXiv:2610.20597v1 Announce Type: cross Abstract: Training language agents is bottlenecked by the high cost of long-horizon rollouts, yet standard agentic reinforcement learning (RL) learns only from a terminal scalar reward, leaving the supervision in intermediate steps largely …
arXiv:2610.20401v1 Announce Type: new Abstract: Reinforcement learning makes language models better at solving tasks, but the same mechanism also rewards guessing. When success is the only way to earn reward, opting for a safe fallback is always worse than a long-shot attempt. Wh…
arXiv:2610.19439v1 Announce Type: new Abstract: While LLM-based agent systems have demonstrated remarkable proficiency in reasoning and solving complex real-world tasks, they also come with substantial token consumption and inference latency, limiting their practical deployment. …
arXiv:2604.12128v2 Announce Type: replace Abstract: Do language models struggle with self-reference itself, or with structures where propagating truth values along references cannot settle a value in finitely many steps? Our results support the latter. In our main test, statement…
arXiv:2601.08689v3 Announce Type: replace Abstract: Large Language Models (LLMs) have shown strong capabilities in finance-oriented question answering, yet their evaluation on professional quantitative workflows remains fragmented and often stops at factual or document-grounded Q…
arXiv cs.CL
TIER_1English(EN)·Jungsoo Park, Ethan Mendes, Gabriel Stanovsky, Alan Ritter·
arXiv:2509.20645v4 Announce Type: replace Abstract: Modern ML research increasingly faces an evaluation bottleneck: building benchmarks and running experiments is costly, slowing iteration. We ask how much can be inferred about evaluation outcomes before experiments, given only a…
arXiv:2311.04902v3 Announce Type: replace Abstract: Large Language Models (LLMs) with billions of parameters are prime targets for network pruning, which removes some model weights without hurting performance. Prior approaches such as magnitude pruning, SparseGPT, and Wanda eithe…
arXiv cs.CL
TIER_1English(EN)·Alexander Spangher, Daniel E. Ho, Diyi Yang, Noah Goodman, Sanmi Koyejo·
arXiv:2610.19721v1 Announce Type: cross Abstract: Language captures only a subset of the perceptions, intuitions and intents that form meaning between speakers. What, then, are the limits of language? We take steps towards measuring this in LLMs as follows: for a set of concepts,…
arXiv:2610.19642v1 Announce Type: cross Abstract: Reinforcement learning provides a direct mechanism for optimizing task objectives in large language model-driven agents, but task rewards are typically sparse trajectory-level signals that only evaluate task success, without direc…
arXiv cs.CL
TIER_1English(EN)·Rodney Lafuente-Mercado, Robert Sincero, Phil Sage, Yuan Xue·
arXiv:2610.19138v1 Announce Type: cross Abstract: How well can a language model turn a large, incomplete and partly contradictory body of documents into a plan when no single correct answer exists? We introduce Mission Analysis (MA)-Bench, built on mission analysis, the first ste…
arXiv cs.CL
TIER_1English(EN)·Jingqin Tang, Tim Tomov, Sebastian Schmidt, Dominik Fuchsgruber, Stephan G\"unnemann·
arXiv:2610.20052v1 Announce Type: new Abstract: Ambiguous inputs can admit multiple interpretations, each potentially yielding a different answer. Existing work largely treats this ambiguity as a property of the input, yet for a particular language model, only some alternatives m…
arXiv:2610.19848v1 Announce Type: new Abstract: We present OpenTWBench, an open evaluation suite for Taiwan-domain knowledge in large language models, built around two benchmarks. tw-exam-bench contains 255,854 four-option questions across 148 academic subjects, from fifteen year…
arXiv cs.CL
TIER_1English(EN)·Enrico Liscio, Anne Arzberger, Aditya Shankar, Pradeep K. Murukannaiah·
arXiv:2610.19817v1 Announce Type: new Abstract: Human values shape how we expect language models to behave. Dominant alignment approaches treat values as global objectives, although the appropriate value orientation is personal and contextual. Training-free alignment can steer mo…
arXiv:2610.19696v1 Announce Type: new Abstract: Multi-agent systems allow language models to solve tasks collaboratively by sharing complementary information. These systems typically communicate through text, while recent approaches also explore exchanging internal model states. …
arXiv:2610.19601v1 Announce Type: new Abstract: Knowledge distillation (KD) improves language-model pretraining by matching teacher distributions at corpus prefixes, but provides no direct supervision at contexts induced by student predictions. We introduce PA1-OPD, which augment…
arXiv:2610.19528v1 Announce Type: new Abstract: Retrieving external knowledge can improve large language model (LLM) translation by providing supplementary information such as idiom interpretations and terminology. However, existing retrieval-augmented machine translation systems…
arXiv cs.CL
TIER_1English(EN)·SungHo Kim, Eojin Jeon, SangKeun Lee·
arXiv:2610.19399v1 Announce Type: new Abstract: Korean neologisms arise through diverse mechanisms, a demanding test of language models beyond lexical memorization. To enable systematic evaluation, we introduce NeoTrace, an automated, rerunnable framework that collects and dates …
arXiv:2610.19247v1 Announce Type: new Abstract: Transformers are intuitively considered to be efficient at copying from their inputs. We actually show that the ability of LLMs to copy in context can be improved through explicit modeling. We propose Copy-Paste Language Models (CPL…
arXiv cs.CL
TIER_1English(EN)·Shenbin Qian, David Samuel, Jind\v{r}ich Helcl, Andrey Kutuzov, Stephan Oepen·
arXiv:2610.19185v1 Announce Type: new Abstract: Evaluation of large language models (LLMs) remains largely pre-scientific: it involves many design decisions, with little guidance on how to make them. These choices can substantially affect measured performance and, consequently, d…
arXiv:2610.19113v1 Announce Type: new Abstract: In-context learning (ICL) allows language models to adapt their behavior from instructions and demonstrations in a prompt, but accuracy alone does not reveal how contextual information controls their predictions. We study the acquis…
arXiv:2609.27603v2 Announce Type: replace-cross Abstract: In-Context Learning (ICL) has become a cornerstone of modern LLM deployment. However, existing ICL post-training methods have a critical blind spot: they excel at extracting patterns from demonstrations while often neglect…
arXiv:2605.16343v2 Announce Type: replace-cross Abstract: Looped language models (LoopLMs) increase effective computational depth by repeatedly applying shared Transformer blocks without increasing the parameter count. However, this reuse makes them particularly sensitive to post…
An ensemble of language models is a jury, not an electorate. The answers are testimony from witnesses of unknown reliability, and the question is not what the majority said but who is credible. Most ensemble methods assign each expert a fixed reputation, often learned from labels…
Lexical reference frames interpret hidden states through a model's own remaining layers, enabling accurate attribution, steering, and early layer exit.
Intern-Decision introduces compact multimodal models that output structured choices with probabilities via masked language modeling, achieving high accuracy and low latency for interactive agents.
Replacing a learned embedding table with a frozen binary code table reduces model size and parameters while maintaining or improving performance on several benchmarks, and enables faster CPU inference with a ternary body.
Long-horizon tasks expose a weakness of monolithic tool-use loops: one agent must decompose, execute, and verify a task within a single trajectory. We introduce LAGO, which treats specialized language agents as reusable temporally extended options. A meta-agent retrieves options …
Language captures only a subset of the perceptions, intuitions and intents that form meaning between speakers. What, then, are the limits of language? We take steps towards measuring this in LLMs as follows: for a set of concepts, we measure how well each concept can be transmitt…
Multi-agent systems allow language models to solve tasks collaboratively by sharing complementary information. These systems typically communicate through text, while recent approaches also explore exchanging internal model states. However, sharing these states across architectur…
Reliable use of large language models requires responses to satisfy formatting requirements across changing instructions. Standard compliance scores do not distinguish preserving a model's observed format from correcting it, or reveal whether that correction holds across variants…
Knowledge distillation (KD) improves language-model pretraining by matching teacher distributions at corpus prefixes, but provides no direct supervision at contexts induced by student predictions. We introduce PA1-OPD, which augments corpus-path distillation with teacher feedback…
Retrieving external knowledge can improve large language model (LLM) translation by providing supplementary information such as idiom interpretations and terminology. However, existing retrieval-augmented machine translation systems typically apply predefined retrieval strategies…
Retrieving external knowledge can improve large language model (LLM) translation by providing supplementary information such as idiom interpretations and terminology. However, existing retrieval-augmented machine translation systems typically apply predefined retrieval strategies…
Hidden states encode task information before the readout expresses it, so behavioral timelines reflect readout learning rather than model acquisition, and faster readout training closes this gap earlier.
Transformers are intuitively considered to be efficient at copying from their inputs. We actually show that the ability of LLMs to copy in context can be improved through explicit modeling. We propose Copy-Paste Language Models (CPLMs), which add a lightweight Copy-Paste head to …
Evaluation of large language models (LLMs) remains largely pre-scientific: it involves many design decisions, with little guidance on how to make them. These choices can substantially affect measured performance and, consequently, decisions about subsequent pretraining. To identi…
Copy-Paste Language Models enhance in-context copying by adding a lightweight head that improves training efficiency, downstream performance, and long-context retrieval with minimal overhead.
Benchmarks are essential for evaluating and improving large language models (LLMs). Yet many widely used benchmarks are redundant, making evaluation unnecessarily expensive. Benchmark compression lowers this cost by estimating full-benchmark scores and rankings from a partial eva…
How well can a language model turn a large, incomplete and partly contradictory body of documents into a plan when no single correct answer exists? We introduce Mission Analysis (MA)-Bench, built on mission analysis, the first step of military operational planning, in which a sta…
arXiv:2603.17917v2 Announce Type: replace-cross Abstract: Large language models (LLMs) contain billions of parameters, but many exact weight values are not essential. We apply scalar $K$-means clustering to pretrained models, replacing each weight matrix with only 16--64 shared v…
arXiv cs.AI
TIER_1(CA)·Max van Duijn, Michiel van der Meer, Aske Plaat, Niki van Stein·
arXiv:2610.13455v1 Announce Type: new Abstract: Large language models (LLMs) have taken the world by storm. Their ability to communicate with us in our own language, and to reason about our world, has made them a favorite tool for writing, answering questions, and much more. Thei…
arXiv cs.AI
TIER_1English(EN)·Dongkyu Cho, Rumi Chunara, Ali Behrouz·
arXiv:2610.12602v1 Announce Type: cross Abstract: Continual learning (CL) seeks to adapt pretrained models while preserving prior capabilities. Yet the tasks supplied for preservation cover only part of what a model knows, while forgetting can extend beyond them. Recent work show…
arXiv cs.AI
TIER_1English(EN)·Sovesh Mohapatra, Dani S Bassett·
arXiv:2610.13895v1 Announce Type: cross Abstract: Large language models reuse the same attention heads across tasks, but identifying these shared heads does not explain what they do differently in each task. A natural source of inspiration is task-dependent interactions among sha…
arXiv:2605.08037v2 Announce Type: replace-cross Abstract: Direct Preference Optimization (DPO) aligns language models using pairwise preference comparisons, offering a simple and effective alternative to Reinforcement Learning (RL) from human feedback. However, in many practical …
arXiv cs.CL
TIER_1English(EN)·Faisal Ladhak, Shantanu Acharya, Santiago Akle Serano, Simeng Sun, Boris Ginsburg, Krishna C Puvvada·
arXiv:2610.12582v1 Announce Type: new Abstract: A latent recurrent language model applies a shared block of layers repeatedly, so one checkpoint can trade inference compute for accuracy by changing the loop count. At one loop, however, such models typically fall well short of a n…
arXiv cs.CL
TIER_1English(EN)·David Yiyang Liu, Aditya Joshi, Paul Dawson, Andrew Piper·
arXiv:2610.12989v1 Announce Type: new Abstract: We investigate whether language models contain internal components that contribute to representing narrativity, the degree to which a text exhibits narrative properties. Across GPT-2-Small, Gemma-2-2B, and Llama-3.1-8B, we identify …
arXiv cs.CL
TIER_1English(EN)·Juan Terven, Diana-Margarita C\'ordova-Esparza, Ruben Gomez, Carlos Miranda Rocha, Julio-Alejandro Romero-Gonz\'alez, Edgar Chavez-Urbiola·
arXiv:2610.13821v1 Announce Type: new Abstract: A useful language model should know when to answer and when to say "I don't know." Can we improve that decision by training the model itself or by training a separate component to check its answers? We study both approaches on Qwen3…
arXiv:2610.12792v1 Announce Type: cross Abstract: Technologies like large language models make it easier to convert cultural material into analyzable variables, while LLM-driven agentic systems extend machine assistance to other parts of the research workflow. Rather than a full …
arXiv cs.CL
TIER_1English(EN)·Christopher Kissling, Elena Merdjanovska, Alan Akbik·
arXiv:2601.18901v2 Announce Type: replace Abstract: Knowledge probing quantifies how much relational knowledge a language model (LM) has acquired during pre-training. Existing knowledge probes evaluate model capabilities through metrics like prediction accuracy and precision. Suc…
arXiv:2605.11845v2 Announce Type: replace Abstract: Language models are increasingly used in settings where outputs should follow an intended distribution, yet their generation probabilities are often poorly calibrated to it. In many such requests the distribution is implicit and…
arXiv cs.LG
TIER_1English(EN)·Oscar Davis, Daan Roos, Alice Bizeul, Amitis Shidani, Victor Turrisi, Louis B\'ethune, Pierre Ablin, Marco Cuturi·
arXiv:2610.13834v1 Announce Type: new Abstract: Continuous diffusion language models (CDLMs) promise to bring the benefits of continuous flow modelling to language, including accelerated sampling and controllable generation. The field has grown rapidly, and the capabilities of CD…
arXiv:2406.07177v2 Announce Type: replace Abstract: Large language models (LLMs) have achieved remarkable performance on Natural Language Processing (NLP) tasks, but they are hindered by high computational costs and memory requirements. Ternarization, an extreme form of quantizat…
Proximal Policy Optimization (PPO) is widely used in reinforcement learning with verifiable rewards for language-model reasoning. While value evolution along trajectory characterizes relative token-level credit, yet standard critics often fail to track it under sparse terminal re…
Large language models (LLMs) have taken the world by storm. Their ability to communicate with us in our own language, and to reason about our world, has made them a favorite tool for writing, answering questions, and much more. Their answers already influence our decisions and ou…
arXiv cs.LG
TIER_1English(EN)·Frederikke Isa Marin, Panagiotis Antoniadis, Dionysia Danai Brilli, Andreas Bjerregaard, Rachael DeVries, Yan Li, Ole Winther, Wouter Boomsma·
arXiv:2610.11761v1 Announce Type: new Abstract: The genome holds the blueprint that governs the biological properties of the cell. Consequently, advancing our knowledge of genomic function is crucial both for a broader understanding of biology and for continued biomedical advance…
arXiv:2610.10625v1 Announce Type: cross Abstract: Looped Language Models (LoopLMs) provide a parameter efficient approach to scaling model capabilities through repeated use of shared parameters across recurrent steps. Since each recurrent depth can be read out independently, a si…
arXiv:2610.10854v1 Announce Type: new Abstract: Knowledge distillation is an established technique for improving the capabilities of small, efficient student models by training them with the representations of larger, more capable teacher models. Much of the recent work in the di…
arXiv:2610.12235v1 Announce Type: new Abstract: AI research agents automate the cycle of proposing, implementing, and evaluating experiments, opening a path toward recursive self-improvement. Yet their ability to propose experiments outpaces their capacity to execute them in real…
arXiv cs.CL
TIER_1English(EN)·Thomas Gebhart, Russell J. Funk·
arXiv:2610.12207v1 Announce Type: new Abstract: Pre-trained transformer models are increasingly being used to study scientific and technological progress. Encoders tuned to paper or patent text outperform general-purpose models on downstream classification, regression, and proxim…
arXiv cs.CL
TIER_1English(EN)·Clara Meister, G\"ul Sena Alt{\i}nta\c{s}, Antoine Bosselut·
arXiv:2610.12144v1 Announce Type: new Abstract: Tokenizer choice affects multilingual language modeling, but vocabulary capacity is finite and vocabulary size is often constrained: improving representation for some languages often comes at the expense of others. We therefore ask …
arXiv:2610.11915v1 Announce Type: new Abstract: Machine unlearning in large language models aims to remove unwanted knowledge while preserving the model's remaining capabilities. Although existing methods use retention objectives or restrict where edits occur, achieving the desir…
arXiv cs.CL
TIER_1English(EN)·Yuhao Chen, Shuochen Liu, Jiayao Shi, Jian Hong, Chen Cheng, Xinyun Ding, Tao Wang, Ya Li, Quan Liu, Tong Xu·
arXiv:2610.11776v1 Announce Type: new Abstract: Long-term personalization requires language models to use interaction history to track users' preferences across sessions. Parametric memory encodes this interaction history into model parameters or adapters, reducing the need to in…
arXiv:2610.11305v1 Announce Type: new Abstract: A language model may revise the same proposition after receiving genuinely relevant evidence or after receiving directional user pressure that adds no relevant fact. The observable response shift alone therefore does not reveal whic…
arXiv:2610.10946v1 Announce Type: new Abstract: Large language models (LLMs) are entering wildfire management, where overstated evaluations can cost property and lives. How do they perform on wildfire tasks, with and without grounding? Bare means a model receives the task input a…
arXiv cs.CL
TIER_1English(EN)·Hala Almaghout, Christian Federmann, Qin Gao·
arXiv:2610.10918v1 Announce Type: new Abstract: Large Language Models (LLMs) are considered to be a more efficient and cost-effective alternative to human judgment for Machine Translation (MT) evaluation. With MT evaluation spanning a large number of language pairs, domains and l…
arXiv:2610.10592v1 Announce Type: new Abstract: Historical Polish is well documented as a language but annotated in machine-readable form only to about a million words for the period this paper covers; the rest sits behind optical character recognition of variable quality. We pre…
arXiv cs.AI
TIER_1English(EN)·Gabriel U. Talasso, Meghdad Kurmanji, Allan M. de Souza, Nicholas D. Lane, Leandro A. Villas·
arXiv:2604.00050v3 Announce Type: replace-cross Abstract: Federated Learning (FL) has emerged as a promising technique for training language models on distributed and private datasets of diverse tasks. However, aggregating models trained on heterogeneous tasks often degrades the …
arXiv:2604.05716v2 Announce Type: replace Abstract: LLMs have shown strong potential to advance scientific discovery. Whether they possess the capacity for foundational innovation, however, remains an open question. In this work, we focus on a prerequisite for foundational innova…
arXiv cs.AI
TIER_1English(EN)·Noam Mazor, Andrew Morgan, Rafael Pass·
arXiv:2610.11031v1 Announce Type: cross Abstract: A long-standing hypothesis in artificial intelligence and neuroscience posits that intelligence is closely related to compression: the ability to compress information efficiently intuitively reflects capacities associated with int…
arXiv cs.AI
TIER_1English(EN)·Sungyoon Kim, Kaan Ozkara, Youngsuk Park·
arXiv:2610.10975v1 Announce Type: cross Abstract: Muon has motivated a growing family of optimizers that compose multiple matrix normalizations, but these methods remain fragmented and lack a unified perspective. We introduce chained linear minimization oracles (chained LMOs), wh…
arXiv cs.AI
TIER_1English(EN)·Marjan Celikik, Ana Peleteiro Ramallo, Javier Morales·
arXiv:2610.10827v1 Announce Type: cross Abstract: Corrective feedback is among the best-evidenced drivers of second-language acquisition, yet corrections delivered during lessons rarely accumulate into an actionable view of grammar mastery. Prompted frontier models can provide su…
arXiv cs.AI
TIER_1English(EN)·Peter Devine, Nick Ryan, Benjamin Sirb, Alex Chiocchi·
arXiv:2610.11715v1 Announce Type: new Abstract: Hypernetworks that map a context directly to a LoRA adapter let a large language model carry that context in its weights, but prior work has demonstrated them only on base models of up to 14 billion parameters. We present the Intern…
arXiv cs.CL
TIER_1English(EN)·Tobias Braun, Nils Loose, Alexander Herzog, Virginia Ceccatelli, Marcus Rohrbach, Thomas Eisenbarth, Lorenzo Cavallaro·
arXiv:2610.09087v1 Announce Type: new Abstract: Large language models are informing decisions with ever-higher stakes. As the consequences of their errors grow, a central question becomes harder to ignore: how much can we trust an individual answer? Yet recognizing when to defer …
arXiv cs.CL
TIER_1English(EN)·Mingyan Liu, Min Huang·
arXiv:2610.10304v1 Announce Type: cross Abstract: We study whether latent sequence compression of prompt prefixes preserves the capabilities that large language models rely on during inference. We introduce SemanticFold, a compression scheme that folds prefix hidden states at lea…
arXiv cs.CL
TIER_1English(EN)·James Ravi Kirkpatrick, Alexandru Radulescu, Rachel Katharine Sterken·
arXiv:2610.09064v1 Announce Type: cross Abstract: When we interact with large language models (LLMs), are we having a conversation? They are designed to invite us to treat them as intelligent interlocutors who remember, act, and make commitments. But appearances deceive. We intro…
arXiv:2610.10049v1 Announce Type: new Abstract: Reasoning models allocate extra computation at inference time and present their answers as the product of deliberate thought. If this deliberation works the way dual-process accounts of human cognition suggest, longer thinking shoul…
arXiv:2610.09346v1 Announce Type: new Abstract: Quantization-aware training (QAT) can recover much of the accuracy lost when large language models are compressed below four bits. Existing re- covery stages, however, are commonly optimized on fixed completions or teacher-generated…
arXiv cs.CL
TIER_1English(EN)·Marek \v{S}uppa, Ivan Vykopal, Andrej Ridzik, Kristi\'an Sopkovi\v{c}, Nat\'alia K\v{n}a\v{z}ekov\'a, Jaroslav Kop\v{c}an, Miroslav Bl\v{s}t\'ak, Vikt\'oria Ondrejov\'a, Daniel Hl\'adek, Michal Gregor, Martin Tamajka, Mari\'an \v{S}imko·
arXiv:2610.09152v1 Announce Type: new Abstract: Multilingual LLM benchmarks omit Slovak, a morphologically rich West Slavic language of five million speakers, or cover it only by machine translation. We present sk-bench, a native-first Slovak benchmark with 30 datasets (33 scored…
arXiv:2610.09033v1 Announce Type: new Abstract: Standard safety evaluations of large language models assess harmful requests written in canonical plain text, while models in real-world deployment routinely receive inputs containing emojis, altered spellings, encoded strings, and …
arXiv:2610.06387v2 Announce Type: replace Abstract: Large language models (LLMs) have achieved substantial performance gains through increases in model size, training data, and computational resources. However, traditional scaling approaches produce diminishing returns, rising fi…
arXiv:2610.10261v1 Announce Type: cross Abstract: Context: Once defined a taxonomy of stages structuring Machine Learning (ML) pipelines (e.g. Data Preprocessing, Modeling...), extracting these stages from source code is key for better understanding ML practices. However, the div…
arXiv:2610.05366v2 Announce Type: replace Abstract: \`{I}r\`{o}y\`{i}nSpeech is a 42-hour, 80-speaker Yor\`ub\'a read-speech corpus whose audio has been distributed by ELRA since 2024. This paper describes the release of its text component: 24,905 unique, hand-verified, tone-mark…
arXiv:2505.19912v3 Announce Type: replace Abstract: We present Adjacent Possible Exploration (APE), a selective fine-tuning method for adapting large language models that systematically explores parameter modifications while maintaining model stability. Inspired by evolutionary o…
arXiv cs.CL
TIER_1English(EN)·Maverick Morales, Tom\'a\v{s} Dominik, Vermut Gao, Katrina Shirey, Paulius Rimkevi\v{c}ius, Aaron Schurger, Uri Maoz·
arXiv:2610.10405v1 Announce Type: cross Abstract: Monitoring the chain-of-thought of reasoning artificial intelligence (AI) models remains a key approach to detecting deception and other forms of misbehavior in such models. However, semantic chain-of-thought monitoring depends on…
Continual learning (CL) seeks to adapt pretrained models while preserving prior capabilities. Yet the tasks supplied for preservation cover only part of what a model knows, while forgetting can extend beyond them. Recent work shows that language models can supplement this limited…
A latent recurrent language model applies a shared block of layers repeatedly, so one checkpoint can trade inference compute for accuracy by changing the loop count. At one loop, however, such models typically fall well short of a non-recurrent model of the same size, a one-loop …
This is a book about large language models. As indicated by the title, it primarily focuses on foundational concepts rather than comprehensive coverage of all cutting-edge technologies. The book is structured into six main chapters, each exploring a key area: pre-training, genera…
Muon has motivated a growing family of optimizers that compose multiple matrix normalizations, but these methods remain fragmented and lack a unified perspective. We introduce chained linear minimization oracles (chained LMOs), which cast these methods as compositions of LMOs. De…
Monitoring the chain-of-thought of reasoning artificial intelligence (AI) models remains a key approach to detecting deception and other forms of misbehavior in such models. However, semantic chain-of-thought monitoring depends on reasoning traces being legible and sufficiently f…
Large language models (LLMs) are increasingly used for emotional support, raising concern that sustained use may draw users away from their real-world relationships. Yet existing evaluations primarily focus on the safety, empathy, or helpfulness of responses, leaving under-examin…
A retrieval-augmented model can match a document without relying on it. Controlled knowledge conflicts make source choice observable and let us ask a second question that prediction alone cannot answer: which internal-state properties define useful intervention directions? We stu…
arXiv:2610.07853v1 Announce Type: cross Abstract: Ternary language models such as BitNet b1.58, Falcon-E and BitCPM are fine-tuned with higher-precision latent weights and deployed as ternary codes produced by an export step that, in the labs' documented pipelines, first casts th…
arXiv:2610.07587v1 Announce Type: new Abstract: LLM orchestration investigates how an orchestrator coordinates a group of autonomous agents to achieve common goals or maximize collective welfare. The agents are typically heterogeneous, each holding a private preference that it pu…
arXiv:2610.06940v1 Announce Type: new Abstract: Continual adaptation of language models can change their output distribution on prompts learned earlier, while retaining every old prompt-answer pair may be undesirable or impossible. We study condition-anchored generative distillat…
arXiv:2610.07606v1 Announce Type: new Abstract: This paper establishes a theoretical framework for vertical adaptive layer skipping, proving three foundational results: (i) an Expected FLOPs formula (theorem 2) giving a closed-form expression for the computational cost of arbitra…
arXiv cs.AI
TIER_1English(EN)·Minjia Mao, Shi Chen, Bowen Yin, Xiao Fang·
arXiv:2610.07661v1 Announce Type: new Abstract: Massive activations, a phenomenon in which a small number of hidden channels exhibit exceptionally large magnitudes, are pervasive in large language models (LLMs). However, the mechanism by which a token develops massive activations…
arXiv:2610.08258v1 Announce Type: new Abstract: Auditing the claimed outcomes of large language model (LLM) training is challenging when model weights and training data are private, while cryptographically proving the full training process is prohibitively expensive at Transforme…
arXiv cs.AI
TIER_1English(EN)·Xingru Zhou, Luis Sentis, Aarti Choudhary·
arXiv:2610.07276v1 Announce Type: cross Abstract: Deployment-time safety of language models is commonly implemented through runtime guardrails such as input moderation, routing, retrieval verification, and output filtering. Existing deployment frameworks provide increasingly capa…
arXiv:2610.07553v1 Announce Type: cross Abstract: LoRA fine-tuning adapts small language models (SLMs) to heterogeneous instruction data within a low-rank update subspace, making it vulnerable to three structural problems: conflicting gradients that cancel, static data selection …
arXiv cs.AI
TIER_1English(EN)·Byeonghu Na, Donghyeok Shin, Yeongmin Kim, Mina Kang, Il-Chul Moon·
arXiv:2610.07706v1 Announce Type: cross Abstract: Autoregressive large language models (LLMs) have rapidly advanced in capability, but their increasing scale comes with substantial computational and memory costs at inference time. Knowledge distillation (KD) offers a practical so…
arXiv:2610.07848v1 Announce Type: cross Abstract: Parameter-efficient fine-tuning (PEFT) has become a standard approach for adapting large language models to downstream tasks. However, most existing PEFT methods rely on uniform and static adaptations, without accounting for the s…
arXiv:2610.07940v1 Announce Type: cross Abstract: Looped language models apply the same stack of layers T times to each token, which deepens the model without adding parameters but multiplies its key-value (KV) cache by T. The larger cache limits how many sequences a GPU can deco…
arXiv:2511.20471v3 Announce Type: replace Abstract: Recent advances in Large Language Model (LLM) reasoning have improved conventional problem solving, but creative reasoning remains comparatively underexplored. Inspired by cognitive science, we formalize combinational, explorato…
arXiv:2503.22764v3 Announce Type: replace-cross Abstract: The large language model (LLM) is typically integrated into the mainstream optimization protocol. However, it remains underexplored whether maintaining the model integrity is \textit{indispensable} for promising performanc…
arXiv:2601.16390v2 Announce Type: replace-cross Abstract: Large language models exhibit strong multilingual capabilities, yet significant performance gaps persist between dominant and non-dominant languages. Prior work attributes this gap to imbalances between shared and language…
arXiv:2610.08341v1 Announce Type: cross Abstract: Recent training-free pruning approaches for Multimodal Large Language Models (MLLMs) effectively cut computational overhead by exploiting visual redundancy or text-vision attention. However, they frequently suffer from semantic de…
arXiv:2610.07247v1 Announce Type: new Abstract: Large language models have shown strong reasoning capabilities, but their high inference costs make knowledge distillation an important approach for transferring such capabilities to compact models in resource-constrained scenarios.…
arXiv:2510.22954v2 Announce Type: replace Abstract: Language models (LMs) often struggle to generate diverse, human-like creative content, raising concerns about the long-term homogenization of human thought through repeated exposure to similar outputs. Yet scalable methods for e…
arXiv:2610.08164v1 Announce Type: cross Abstract: Low-rank quantization error compensation (LQEC) recovers the accuracy lost under aggressive weight quantization by attaching a closed-form rank-$r$ adapter beside each frozen quantized weight, without any training. We show that ex…
arXiv:2610.08630v1 Announce Type: new Abstract: Recently Large Language Models (LLMs) and LLM-based agents increasingly need to incorporate knowledge acquired after pretraining, e.g., domain facts, user preferences, documents, and interaction experience. In-context learning (ICL)…
arXiv:2508.10020v2 Announce Type: replace-cross Abstract: Enhancing LLM reasoning in federated settings is nontrivial due to stringent computational, communication, and privacy constraints, especially in healthcare, where clinically consequential decisions require not only accura…
Large language models are informing decisions with ever-higher stakes. As the consequences of their errors grow, a central question becomes harder to ignore: how much can we trust an individual answer? Yet recognizing when to defer remains difficult because language models can pr…
Low-rank quantization error compensation (LQEC) recovers the accuracy lost under aggressive weight quantization by attaching a closed-form rank-$r$ adapter beside each frozen quantized weight, without any training. We show that existing compensators are limited by two shared simp…
This paper establishes a theoretical framework for vertical adaptive layer skipping, proving three foundational results: (i) an Expected FLOPs formula (theorem 2) giving a closed-form expression for the computational cost of arbitrary per-sample skip schedules as a function of la…
LLM orchestration investigates how an orchestrator coordinates a group of autonomous agents to achieve common goals or maximize collective welfare. The agents are typically heterogeneous, each holding a private preference that it pursues but does not reveal. Inferring such hidden…
Large language models are informing decisions with ever-higher stakes. As the consequences of their errors grow, a central question becomes harder to ignore: how much can we trust an individual answer? Yet recognizing when to defer remains difficult because language models can pr…
Recently Large Language Models (LLMs) and LLM-based agents increasingly need to incorporate knowledge acquired after pretraining, e.g., domain facts, user preferences, documents, and interaction experience. In-context learning (ICL) and ICL-based agent harness remain flexible, bu…
arXiv:2610.02986v1 Announce Type: new Abstract: Membership inference on large language models (LLMs) aims to determine whether a given text sample was included in an LLM's training data, without access to its training corpus. Despite recent progress, existing benchmarks suffer fr…
arXiv:2610.02856v1 Announce Type: new Abstract: Multi-task post-training of large language models (LLMs) aims to improve performance across tasks with unequal amounts of training data. Existing methods focus primarily on balancing task contributions during single-model training. …
arXiv:2610.02549v1 Announce Type: new Abstract: Time and event expression extraction are fundamental temporal reasoning tasks, but the problem remains difficult due to annotation ambiguity, domain sensitivity, and unstable model behavior. Existing evaluations focus on in-domain p…
arXiv cs.AI
TIER_1English(EN)·Han Wang, Ishwar B Balappanawar, Huan Zhang·
arXiv:2610.02695v1 Announce Type: cross Abstract: Reliable large language models (LLMs) must not only produce accurate answers but also express confidence that faithfully reflects their probability of being correct. Such calibration is essential for identifying uncertain predicti…
arXiv:2610.02705v1 Announce Type: cross Abstract: The Muon optimizer derives its update rule for hidden linear layers by solving a local linearization of the loss penalized by the spectral norm, motivated by an RMS-stability argument for dense linear layers. Standard Muon impleme…
arXiv cs.AI
TIER_1English(EN)·Mohsen Larni (Department of Computer Science, University of Nevada, Las Vegas), Sobhan Ebrahimi Azar (Department of Computer Science, University of Nevada, Las Vegas), Pouyan Nahed (Department of Computer Science, University of Nevada, Las Vegas), Kazem …·
arXiv:2610.03329v1 Announce Type: cross Abstract: Large language models are increasingly used where small syntactic errors matter, yet character-level reasoning is still evaluated mostly through isolated probes and aggregate accuracy. We introduce SyntaxBench, a diagnostic benchm…
arXiv cs.AI
TIER_1English(EN)·Shuyang Cao, Karthik Radhakrishnan, David Rosenberg, Steven Lu, Pengxiang Cheng, Lu Wang, Shiyue Zhang·
arXiv:2505.21870v2 Announce Type: replace-cross Abstract: Retrieval-augmented generation (RAG) generally enhances large language models' (LLMs) ability to solve knowledge-intensive tasks. But RAG could also lead to performance degradation due to imperfect retrieval and the model'…
arXiv:2610.02766v1 Announce Type: new Abstract: Retaining language-model prefix states trades recomputation against storage time. Optimizing each cached block independently can overcount savings: a resident block is usable only when the required preceding prefix is also available…
arXiv:2610.02632v1 Announce Type: new Abstract: As large language models are increasingly deployed for multi-step reasoning, verifying the correctness of their outputs has become essential for maintaining reliability at scale. Verifying the correctness of large language model out…
arXiv cs.LG
TIER_1English(EN)·Zhenghao Zhao, Gaowen Liu, Zhiling Lan, Yan Yan·
arXiv:2610.02593v1 Announce Type: new Abstract: Data selection is already a central bottleneck in large-language-model training, where web-scale corpora are noisy and token budgets are finite. In continual pre-training (CPT), it becomes a forgetting-control problem: a poorly chos…
arXiv cs.LG
TIER_1English(EN)·Nicolas Martorell, Wendy Brau, Gonzalo A. Heredia, Tom\'as Pablo Korenblit, Gaspar Labasti\'e, Tom\'as Gimenez Molina·
arXiv:2610.02303v1 Announce Type: new Abstract: Language models increasingly assist people with power-related requests, so systematic differences in whom they help could shift the distribution of power at scale, or be exploited by users who learn which identities are refused less…
arXiv:2610.03199v1 Announce Type: cross Abstract: Large language models fine-tuned from a shared base can be merged by averaging their task vectors, but some merges collapse far below the base model, and common merge operators give no warning before evaluation. We show that one s…
arXiv:2610.02911v1 Announce Type: cross Abstract: Methods for training language models on stale samples are judged by comparisons against importance-corrected baselines. We show that details of the experimental harness can reverse the observed ranking of methods, and we introduce…
arXiv:2610.02432v1 Announce Type: cross Abstract: Large language models (LLMs) in production systems face prompt injections, trojans (backdoors), and manipulation of automatic quality metrics. This thesis develops models, methods, and algorithms for evaluating and improving LLM r…
arXiv:2610.01165v1 Announce Type: new Abstract: Layer interventions are widely used to probe the internal organization of language models, yet most analyses examine a single training checkpoint even though model representations and computations evolve throughout pretraining. This…
arXiv:2610.00265v1 Announce Type: new Abstract: Large Language Bayes (LLB) answers an informal modelling question by sampling candidate probabilistic programs from a language model, running approximate inference on each, and averaging them with weights proportional to an exponent…
arXiv:2610.01560v1 Announce Type: cross Abstract: Model intelligence and fast response jointly shape the quality of interaction with speech language models, yet remain difficult to achieve together. Explicit chain-of-thought (CoT) improves reasoning and audio understanding, but g…
arXiv cs.LG
TIER_1English(EN)·Bangji Yang, Jingyuan Li, Jiajun Fan, Yi Evie Zhang, Ruihan Guo, Hongba Ma, Neil He, Chumeng Liang, Qinglong Zheng, Zhanghan Ni, Ge Liu·
arXiv:2610.01110v1 Announce Type: new Abstract: How much can test-time computation improve a language model, and at what cost? Test-time scaling is widely proposed as a substitute for larger models, but existing comparisons mostly evaluate one domain at a time and rarely charge s…
arXiv cs.LG
TIER_1English(EN)·Chayne Thrash, Kevin Chen, Soheil Kolouri·
arXiv:2609.35579v2 Announce Type: replace Abstract: Residual stream pruning methods reduce inference cost by shrinking the model's hidden dimension, but existing approaches typically choose these dimensions by minimizing activation reconstruction error. This criterion implicitly …
arXiv:2508.13408v3 Announce Type: replace Abstract: Chemical Language Models (CLMs) are increasingly used in de novo drug design, driven by recent growth in model scale, compute, and dataset size. However, the relationship between design choices, training dynamics, and downstream…
arXiv:2610.01592v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly applied to process streaming data, with practitioners relying on benchmarks to select the best model even though these signals only approximate real performance. While oracle annotatio…
Numerical Linear Algebra (NLA) has consistently played a vital role in advancing science by providing tools to solve fundamental problems encountered in scientific and engineering applications. Over the decades, it has continually evolved to meet the demands driven by successive …
A language model reads long text in one quadratic forward pass, stops at the context window, and loses accuracy with length before reaching it. We ask whether the read can be factorized when deciding over a finite set: which document is relevant, which option is supported, which …
arXiv:2610.00694v1 Announce Type: cross Abstract: Compression reports summarize how far a compressed language model moved from the dense one, usually by a KL divergence; a deployment that relies on the dense model's outputs needs to know how many of its decisions changed. We show…
arXiv:2610.00568v1 Announce Type: cross Abstract: Large language models are characterized by three key properties: capability, alignment, and faithfulness. Prior work studies the tradeoffs between capability and alignment, and between capability and faithfulness, but a third tens…
arXiv cs.AI
TIER_1English(EN)·Timoth\'ee Lesort, Alejandra L\'opez de Aberasturi G\'omez, Tristan Karch, Tom Veniat, Philippe Modard, Karl Tuyls, Ludovic Denoyer·
arXiv:2610.00416v1 Announce Type: new Abstract: Evaluating large language models becomes increasingly challenging as their capabilities advance: benchmarks can saturate, public test sets risk contamination, and assessing harder tasks can require expensive grading or execution inf…
arXiv cs.AI
TIER_1English(EN)·Oguzhan Baser, Elahe Sadeghi, Eric Wang, Nico Vergauwen, Sam Kazemian, Hong Kang, Sandeep P. Chinchali, Sriram Vishwanath·
arXiv:2602.12630v2 Announce Type: replace-cross Abstract: Most large language models (LLMs) run on external clouds: users send a prompt, pay for inference, and must trust that the remote GPU executes the LLM without any adversarial tampering. We critically ask how to achieve veri…
arXiv:2606.22676v2 Announce Type: replace Abstract: Refusal on a safety benchmark does not reveal how stable that behavior will remain after model updates. Benign downstream fine-tuning can weaken refusal, yet behavioral evaluations typically expose this fragility only after an i…
arXiv cs.AI
TIER_1English(EN)·Jiangfeng Chen, Xinyu Wang, Tianshuo Yan, Hanwei Wu, Xiao-Wen Chang, Yang Zhang, Lei Ding·
arXiv:2610.00717v1 Announce Type: cross Abstract: Post-training compression of LLM attention is often formulated as independent matrix approximation, ignoring both the shared structure among attention projections and the representation shift introduced by earlier compression. We …
arXiv:2610.00673v1 Announce Type: cross Abstract: Looped language models increase effective depth by repeatedly applying a shared block of layers, but existing large-scale recipes require multi-stage training over trillions of tokens, while the benefits of recurrence remain diffi…
Language-model systems can separate contextual computation, persistent storage, and exact execution instead of updating all capabilities through one shared parameter system. We investigate FEM-ASM, a finite-element-method-inspired organization in which independently constructed d…
A trainable input embedding table assigns each vocabulary item an independently adjustable vector. We investigate whether this token-specific parameterization is required for substantial language-modeling capability, or whether a shared Transformer can learn from fixed token iden…
A language model reads long text in one quadratic forward pass, stops at the context window, and loses accuracy with length before reaching it. We ask whether the read can be factorized when deciding over a finite set: which document is relevant, which option is supported, which …
arXiv:2603.26164v2 Announce Type: replace-cross Abstract: Data-centric training has emerged as a promising direction for improving large language models (LLMs) by optimizing not only model parameters but also the selection, composition, and weighting of training data during optim…
arXiv:2609.39869v1 Announce Type: new Abstract: Grammar-constrained generation guarantees syntactic validity, but can substantially degrade semantic quality when the model's preferred outputs are poorly aligned with the imposed grammar. This trade-off is particularly severe when …
arXiv:2609.38261v1 Announce Type: cross Abstract: In this paper we propose NinaXander, a series of composed language models obtained by connecting layers of frozen language models from different architecture families with a single trained shared-latent adapter. A composed model r…
arXiv:2609.39321v1 Announce Type: cross Abstract: Group Relative Policy Optimization (GRPO) has emerged as a memory-efficient reinforcement fine-tuning (RFT) technique for reasoning-intensive tasks. How- ever, GRPO training dynamics on small language models (SLMs) remain poorly u…
arXiv cs.AI
TIER_1English(EN)·Zhen Liang, Hai Huang, Wentao Chen·
arXiv:2609.39902v1 Announce Type: cross Abstract: Large language models have achieved remarkable capabilities across diverse domains, yet their safety alignment remains vulnerable to jailbreak attacks. In this work, we identify a previously underexplored failure mode - safety gen…
arXiv cs.AI
TIER_1English(EN)·Hoa Quynh Nhung Nguyen, Jacopo Staiano, Michael Sullivan·
arXiv:2609.40121v1 Announce Type: cross Abstract: While Abstract Meaning Representation (AMR) has historically improved performance on a range of NLP tasks, the benefit---or lack thereof---of AMR augmentation for modern LLMs is thus far unclear. In this paper, we attempt to repro…
arXiv:2511.00763v3 Announce Type: replace Abstract: We investigate the performance of large language models (LLMs) on repetitive deterministic prediction tasks and study how the sequence accuracy rate (SAR) scales with output length. Each such task involves the repetition of the …
arXiv cs.AI
TIER_1English(EN)·Mansi Sakarvadia, Aswathy Ajith, Arham Khan, Nathaniel Hudson, Caleb Geniesse, Kyle Chard, Yaoqing Yang, Ian Foster, Michael W. Mahoney·
arXiv:2410.02159v3 Announce Type: replace-cross Abstract: Language models (LMs) can "memorize" information, i.e., encode training data in their weights in such a way that inference-time queries can lead to verbatim regurgitation of that data. This ability to extract training data…
arXiv cs.AI
TIER_1English(EN)·Kenan Alkiek, David Jurgens, Vinod Vydiswaran·
arXiv:2510.13935v3 Announce Type: replace-cross Abstract: The facts a language model stores are tied to its parameter count, so small models that fit on edge devices fail on expert problems, which need specialized knowledge and follow multi-step procedures. Fine-tuning for a spec…
arXiv cs.CL
TIER_1(CA)·Juan M Zambrano Chaves, Peniel Argaw, Risa Ueno, Carlo Bifulco, Kristina Young, Rom Leidner, Tristan Naumann, Hoifung Poon·
arXiv:2609.38181v1 Announce Type: new Abstract: Survival analysis estimates time-to-event outcomes from patient covariates and is widely used for medical risk assessment. Patients seeking prognostic information after a diagnosis may turn to large language models (LLMs), now readi…
arXiv cs.CL
TIER_1English(EN)·Parisa Salmani, Peter R. Lewis·
arXiv:2609.38357v1 Announce Type: new Abstract: Conversational safety evaluations often test language models with a single harmful prompt, even though real-world systems interact with users through long, adaptive conversations. This study examines whether models continue to respo…
arXiv cs.CL
TIER_1English(EN)·Jhen-Ke Lin, Chung Chun Wang·
arXiv:2609.38612v1 Announce Type: new Abstract: As natural language drives more applications, language models increasingly run inside programs as decision components: the program sends them the current state and acts on the returned decision until a newer one arrives. When eviden…
arXiv:2609.40185v1 Announce Type: new Abstract: Constrained generation aims to sample from language models (LMs) conditioned on hard constraints. Existing constrained-generation techniques for nondeterministic finite automaton (NFA) constraints either distort the distribution or …
arXiv cs.CL
TIER_1English(EN)·Eduard Tulchinskii, Kristian Kuznetsov, Laida Kushnareva, Anastasia Voznyuk, Andrei Andriiainen, Irina Piontkovskaya, Evgeny Burnaev, Serguei Barannikov·
arXiv:2410.02343v2 Announce Type: replace Abstract: Large language models (LLMs) routinely fail to output the correct option in multiple-choice question answering (MCQA) while encoding the answer internally. We expose this latent knowledge via the Query--Key (QK) score, defined f…
arXiv:2511.10661v2 Announce Type: replace Abstract: It is increasingly important to evaluate the characteristics of systems based on large language models (LLMs). Evaluations in this context often rely on a curated benchmark set of input prompts provided to the LLM, where the out…
arXiv cs.MA (Multiagent)
TIER_1English(EN)·Roberto Di Pietro·
Large Language Model (LLM) agents are increasingly deployed as populations of interacting entities, in which consensus --agreement on a shared answer-- emerges as a collective, unengineered behaviour. Prior work on LLM consensus shows that agents can cross-verify their answers an…
arXiv cs.AI
TIER_1Deutsch(DE)·Yuansen Liu, Yixuan Tang, Anthony Kum Hoe Tung·
arXiv:2609.37700v1 Announce Type: new Abstract: Language models expose internal signals that predict whether an answer is correct, readable from a single forward pass of a frozen model without additional generations. Yet existing probes often commit to one signal family or layer …
arXiv:2609.37076v1 Announce Type: new Abstract: Large language models trained on vast corpora inherently risk memorizing harmful content that may later re-emerge in their outputs. To mitigate this issue, existing unlearning methods typically rely on training-based parameter updat…
arXiv cs.CL
TIER_1English(EN)·Elsayed Eshra, Ali Al-Lawati, Dongwon Lee, Suhang Wang·
arXiv:2609.33638v2 Announce Type: replace-cross Abstract: We introduce RareTrap, a framework for estimating the probability of severe behaviors in black box large language models (LLMs). A key challenge for probability estimation is defining a tractable distribution over the inpu…
arXiv cs.CL
TIER_1English(EN)·Prosper Arineitwe Asiimwe, Francois Meyer, Jan Buys·
arXiv:2609.37543v1 Announce Type: new Abstract: Cross-lingual zero-shot transfer and multilingual fine-tuning are promising approaches for NLP tasks such as Named Entity Recognition (NER) in low-resource languages, but in the absence of target language benchmarks, it is unclear w…
arXiv:2602.13540v2 Announce Type: replace-cross Abstract: Accurate confidence estimation is critical for reliable use of large language models (LLMs). Prior work on LLM calibration largely focuses on response-level confidence, which estimates the correctness of a single generated…
arXiv:2406.00083v3 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by retrieving relevant information from external knowledge bases to provide more accurate, contextually informed, and up-to-date responses. However…
arXiv cs.AI
TIER_1English(EN)·Rebecca Ramnauth, Brian Scassellati·
arXiv:2609.37371v1 Announce Type: cross Abstract: Model adaptation is typically governed by a fixed recipe, even though different update programs can produce substantially different behavioral outcomes. We introduce adaptation compilation, which reframes where, how, and to what e…
arXiv cs.AI
TIER_1English(EN)·Kenan Alkiek, Moontae Lee, David Jurgens, V. G. Vinod Vydiswaran·
arXiv:2609.38006v1 Announce Type: new Abstract: Running a language model locally offers advantages in privacy, latency, and cost, but local hardware fits only small models, which are less capable than frontier models. The usual remedy for a hard query, escalating it to a cloud mo…
Language models expose internal signals that predict whether an answer is correct, readable from a single forward pass of a frozen model without additional generations. Yet existing probes often commit to one signal family or layer and can be brittle under distribution shift; in …
arXiv:2501.09223v3 Announce Type: replace-cross Abstract: This is a book about large language models. As indicated by the title, it primarily focuses on foundational concepts rather than comprehensive coverage of all cutting-edge technologies. The book is structured into six main…
arXiv:2609.29504v1 Announce Type: cross Abstract: Aggregate factuality scores hide where a language model succeeds, which relations it confuses, and whether an answer survives innocuous changes to the question or decoder. We introduce PROOF, a profile-oriented benchmark for factu…
We study a practical question: can a small correction module fix errors in a frozen language model's outputs without degrading its base capabilities? We propose CRN v2, a lightweight logit-level correction module (~34M trainable parameters, 0.73% of the 4.65B text module) that si…
arXiv:2610.12718v1 Announce Type: cross Abstract: Large language models are stochastic by nature. Repeated executions under the same experimental conditions may produce different responses. We propose a framework that can be applied across the model lifecycle, from task design to…
arXiv cs.CV
TIER_1English(EN)·Pengcheng Zheng, Chaoning Zhang, Jiaxin Yan, Sihan Cao, Jianwei Zhang, Xudong Wang, Jiaquan Zhang, Jewon Lee, Tae-Ho Kim, Yang Yang, Heng Tao Shen·
arXiv:2610.08830v1 Announce Type: new Abstract: Multimodal Large Language Models (MLLMs) have demonstrated remarkable reasoning capabilities across vision and language tasks. However, their massive computational and memory demands hinder real-world deployment. While recent effort…
arXiv:2505.06592v2 Announce Type: replace Abstract: In this paper, we propose batch augmentation with unimodal fine-tuning for multimodal learning. We start with pre-trained unimodal models. We fine-tune the unimodal models with the application data. After that, we form a Multi-L…
Latent Space (podcast video)
TIER_1English(EN)·Latent Space·
From GPU kernels and KernelBench to Recursive Language Models, agent harnesses, and massive multi-agent swarms, Alex Zhang is exploring how much capability we’re leaving on the table by wrapping increasingly powerful models in primitive systems. In this episode, the MIT researche…
Medium — fine-tuning tag
TIER_1English(EN)·Photon neurons·
<p>You’re at a dinner party.</p><p>The conversation has drifted from mortgage rates to whether sourdough starters qualify as pets, when someone across the table leans forward and asks the question everyone’s been dodging for three years:</p><p><em>“So… how does ChatGPT actually w…
<!-- SC_OFF --><div class="md"><p>If you build on open-weight reasoning models, you have probably wanted a "how sure is the model about this answer?" signal that does not cost 5x sampling, a judge model, or a labeled dataset. Hence, we built the <strong><em>U-Lens.</em>…
<h1> Integrating Large Language Models into Production: Practical Considerations </h1> <p>Large language models (LLMs) have moved from research labs into real-world services. The excitement around their capabilities often masks the engineering effort required to run them reliably…
<h2> Parallelizing the Transformer, Masking the Future, and the Language Modeling Head </h2> <p>Last post built self-attention and the transformer block, but left two promises unkept: I showed the computation one token at a time (so where's the famous parallelism?), and I never s…
<p>Today we're releasing THX-01, a model for the boring decisions that<br /> agents and backends make thousands of times a day: route this ticket, is this spam, which team, what's the invoice total,<br /> which sentence supports that.</p> <p><strong>What it is</strong></p> <ul> <…
<h1> Introduction </h1> <p>Deploying large language models (LLMs) in a production environment presents a different set of challenges than running them in a notebook. Engineers need to balance latency, cost, and reliability while keeping the model up to date with the latest data. …
<p>Há um tempo estudei Engenharia de IA, usando como base o livro Engenharia de IA da Chip Huyen e fui escrevendo alguns artigos ..<br /> Agora que a série está completa, achei que valia a pena compartilhar por aqui.</p> <p>Se você está estudando IA, LLMs ou só quer entender melh…
<!-- SC_OFF --><div class="md"><p>Learning from data as we observe it is easy for humans, but most machine learning models have limited ability to learn from new data that they have not seen during training. Prior-fitted networks (the idea behind TabPFN) showed that a model train…
<p>German AI company Aleph Alpha released Kolibri 1 on October 3, 2026, an open-weight language model built for German and English. It has 78.1 billion parameters, but only 3.46 billion do work on each token, which keeps it fast for its size. The weights are free to download and …
🧠 Researchers introduce RealtimeBench, a benchmark with 180 tasks that evaluates language model agents while simulated environments continue running in real time rather than pausing for inference. They propose TORSI, a training method where a slower agent learns from failures and…
A zero-training memory harness for open chat models (Qwen and friends). The same model that reads and writes text also classifies its own context as long-term / short-term memory while decoding; a small deterministic eviction engine then enforces the token budget. Design inspired…