arXiv:2602.12429v2 Announce Type: replace Abstract: Foundation models have achieved remarkable success, yet their growing parameter counts pose significant computational and memory challenges. Low-rank factorization offers a promising route to reduce training and inference costs,…
arXiv:2607.14618v1 Announce Type: new Abstract: CPUs are the most universal target for on-device LLM inference, but existing low-bit quantization methods offer either coarse operating points or fine-grained mixed precision that is difficult to execute efficiently on CPUs. We pres…
arXiv:2607.14118v1 Announce Type: new Abstract: Large language models (LLMs) can generate research ideas that appear novel to expert reviewers, but recent work also shows that such ideas often lack diversity, are difficult for LLMs to evaluate reliably, and may fail to translate …
arXiv cs.AI
TIER_1English(EN)·Jiayi Zhang, Simon Yu, Derek Chong, Anthony Sicilia, Michael R. Tomz, Christopher D. Manning, Weiyan Shi·
arXiv:2510.01171v4 Announce Type: replace-cross Abstract: Post-training alignment often reduces LLM diversity, leading to a phenomenon known as mode collapse. Unlike prior work that attributes this effect to algorithmic limitations, we identify a fundamental, pervasive data-level…
arXiv cs.AI
TIER_1English(EN)·Jimmy T. H. Smith, Tarek Dakhran, Alberto Cabrera, Simon S. Lee, Paul Pak, Aditya Tadimeti, Tim Seyde, Maxime Labonne, Alexander Amini, Mathias Lechner·
arXiv:2607.15232v1 Announce Type: cross Abstract: A tokenizer fixed at the start of pre-training allocates vocabulary in proportion to the pre-training corpus, reflecting the deployment priorities at that time. When those priorities shift, languages added later are split into man…
arXiv:2607.14111v1 Announce Type: cross Abstract: Can small language models detect and report on perturbations their own internal activations? We investigate this question through the lens of activation steering: injecting concept vectors into a model's residual stream and measur…
arXiv:2607.14306v1 Announce Type: new Abstract: In this paper, we study the connection between an LLM's output distribution and the data used to train it. Specifically, we study the degree to which an LLM's next-token distribution agrees with the empirical next-token distribution…
arXiv:2607.14109v1 Announce Type: cross Abstract: Probing the capabilities of Large Language Models (LLMs) and building robust solutions for Multiple-Choice Question Answering (MCQA) remain central challenges in natural language understanding. Furthermore, the rapid proliferation…
A tokenizer fixed at the start of pre-training allocates vocabulary in proportion to the pre-training corpus, reflecting the deployment priorities at that time. When those priorities shift, languages added later are split into many more tokens per word, which can raise latency, c…
CPUs are the most universal target for on-device LLM inference, but existing low-bit quantization methods offer either coarse operating points or fine-grained mixed precision that is difficult to execute efficiently on CPUs. We present PolyQ, a CPU-oriented compiler/quantization …
arXiv:2607.13425v1 Announce Type: cross Abstract: Learning effectively from limited data is critical in domains like security where labeled examples are scarce. Large language models (LLMs) have demonstrated some capabilities for data-efficient learning, especially through parame…
arXiv:2510.26707v2 Announce Type: replace-cross Abstract: As LLMs occupy an increasingly important role in society, they are more and more confronted with questions that require them not only to draw on their general knowledge but also to align with certain human value systems. T…
arXiv:2601.20379v2 Announce Type: replace Abstract: Large language models (LLMs) struggle with complex, long-horizon reasoning due to instability caused by their frozen policy assumption. Current test-time scaling methods treat execution feedback merely as an external signal for …
arXiv:2607.13069v1 Announce Type: new Abstract: Large language models produce chain-of-thought (CoT) reasoning that appears logically sound yet may not genuinely depend on its stated premises. We introduce interventional grounding audits, a black-box, step-level test of premise d…
arXiv:2607.13205v1 Announce Type: cross Abstract: Attention-based KV cache eviction (H2O and its descendants) compresses the memory-constrained state of a long-context model by ranking tokens on accumulated attention mass, treated here as signal energy, and keeping the heaviest. …
Learning effectively from limited data is critical in domains like security where labeled examples are scarce. Large language models (LLMs) have demonstrated some capabilities for data-efficient learning, especially through parameter-efficient adaptation methods, but continue to …
Attention-based KV cache eviction (H2O and its descendants) compresses the memory-constrained state of a long-context model by ranking tokens on accumulated attention mass, treated here as signal energy, and keeping the heaviest. On schema-dense input streams such as nested JSON,…
As LLM technology advances, the space of model families, compute hardware, quantization schemes, parallelization strategies, and specialized optimization kernels continues to expand, sharply increasing the code complexity and maintenance cost of general-purpose inference framewor…
arXiv:2607.10139v1 Announce Type: cross Abstract: Selecting the correct answer from a pool of candidate reasoning chains is the engine of test-time scaling, yet the standard selectors each carry a cost: self-consistency inherits the errors of the single model it resamples, and tr…
arXiv:2607.09693v1 Announce Type: cross Abstract: Reinforcement learning (RL) has become the dominant paradigm for improving the reasoning capabilities of large language models, but it requires expensive training, curated data, and reward signals. Recent work shows that sampling …
arXiv:2607.11614v1 Announce Type: cross Abstract: Extending the context length of large language models (LLMs) is critical for many real-world applications, yet standard transformers remain constrained by quadratic compute and linear memory scaling. In this work, we investigate t…
arXiv cs.AI
TIER_1English(EN)·Arjun Ashok, Andrew Robert Williams, Vincent Zhihao Zheng, Irina Rish, Nicolas Chapados, \'Etienne Marcotte, Valentina Zantedeschi, Alexandre Drouin·
arXiv:2508.09904v3 Announce Type: replace-cross Abstract: Real-world forecasting requires models to integrate not only historical data but also relevant contextual information provided in textual form. While large language models (LLMs) show promise for context-aided forecasting,…
arXiv cs.CL
TIER_1English(EN)·Renuka Oladri, Mohan Vamsi Varadaraju Priya, Jerry Wu·
arXiv:2607.09999v1 Announce Type: new Abstract: We show that post-training quantization can silently alter how large language models reason even when task accuracy is preserved. Using a six-category failure taxonomy validated by two independent human annotators (Cohen's $\kappa$ …
arXiv cs.LG
TIER_1English(EN)·Yuchen Zhu, Wei Guo, Jaemoo Choi, Petr Molodyk, Bo Yuan, Molei Tao, Yongxin Chen·
arXiv:2510.08233v3 Announce Type: replace Abstract: Diffusion large language models (dLLMs) are promising alternatives to autoregressive large language models (AR-LLMs), as they potentially allow higher inference throughput. Reinforcement learning (RL) is crucial to enabling dLLM…
arXiv cs.LG
TIER_1English(EN)·Liangqi Yuan, Dong-Jun Han, Shiqiang Wang, Christopher G. Brinton·
arXiv:2502.11007v5 Announce Type: replace Abstract: Compared to traditional machine learning models, recent large language models (LLMs) can exhibit multi-task-solving capabilities through multi-modal data sources and multi-turn conversations. These unique characteristics of LLMs…
arXiv:2607.11505v1 Announce Type: cross Abstract: Post-training is essential for refining the domain-specific capabilities of large language models (LLMs), yet existing reward optimization and distribution matching methods tightly couple policy exploration with distribution align…
arXiv:2607.11207v1 Announce Type: cross Abstract: Table-based reasoning with large language models (LLMs), which requires reasoning based on natural language questions and structured tabular data, has gained widespread attention. However, a series of issues still constrain the ap…
Extending the context length of large language models (LLMs) is critical for many real-world applications, yet standard transformers remain constrained by quadratic compute and linear memory scaling. In this work, we investigate the Associative Recurrent Memory Transformer (ARMT)…
Post-training is essential for refining the domain-specific capabilities of large language models (LLMs), yet existing reward optimization and distribution matching methods tightly couple policy exploration with distribution alignment. This coupling forces expensive exploration d…
Post-training is essential for refining the domain-specific capabilities of large language models (LLMs), yet existing reward optimization and distribution matching methods tightly couple policy exploration with distribution alignment. This coupling forces expensive exploration d…
Table-based reasoning with large language models (LLMs), which requires reasoning based on natural language questions and structured tabular data, has gained widespread attention. However, a series of issues still constrain the application of this task. The previous approaches su…
arXiv:2604.00130v2 Announce Type: replace Abstract: Chain-of-Thought (CoT) prompting has significantly improved the reasoning capabilities of large language models (LLMs). However, conventional CoT often relies on unstructured, flat reasoning chains that suffer from redundancy an…
arXiv:2607.08774v1 Announce Type: new Abstract: Reliability in large language model (LLM) systems is typically framed as a function of model capability. We challenge this by demonstrating that reliability is significantly influenced by \emph{inference-time control} -- the computa…
arXiv cs.AI
TIER_1English(EN)·Duy Nguyen, Archiki Prasad, Elias Stengel-Eskin, Mohit Bansal·
arXiv:2507.18043v2 Announce Type: replace-cross Abstract: Inference-time steering methods offer a lightweight alternative to fine-tuning large language models (LLMs) and vision-language models (VLMs) by modifying internal activations at test time without updating model weights. H…
Large language models (LLMs) increasingly improve their reasoning at test time via additional computation, yet most existing works treat each problem in isolation. When problems arrive sequentially, accumulating reusable experience across them can further improve performance. Exi…
Recent LLM-based agent systems continuously accumulate context across multi-turn interactions, tool invocations, and cross-session workflows. Replaying the full history for every request quickly becomes impractical: long contexts increase prefill cost, may exceed context limits, …
arXiv:2511.23231v2 Announce Type: replace Abstract: Large Language Models (LLMs) and Large Vision-Language Models (LVLMs) demonstrate strong reasoning capabilities, yet their performance in English significantly outperforms that in low-resource languages, raising fairness concern…
Medium — fine-tuning tag
TIER_1English(EN)·Neha Khan • AI & Software Engineer·
<p>You've got 200k tokens. So why do you keep running out of room halfway through your API call?</p> <p>Most developers treat context like a gas tank—fill it up and hope you don't run empty. That's the wrong mental model. Context is inventory. You need to <em>manage</em> it.</p> …
<h2> What Changed </h2> <p>The emergence of Chain-of-Thought (CoT) reasoning has significantly advanced the capability of large language models (LLMs) to handle complex, multi-step tasks. However, a persistent challenge in human-AI interaction with these models has been the ineff…
<p>From data preparation and tokenizer selection to pretraining, LoRA, RLHF, evaluation, and production monitoring, this guide covers the major stages involved in training an AI model.</p> <p>Training an artificial intelligence model is not simply a matter of loading a dataset on…
<h2> Part 2: The Attention Sink Detector </h2> <p><em>Part 2 of a 4-part series on system-level LLM inference internals. Part 1 built the entropy tracker; this one looks one step earlier in the pipeline — at prefill, before a single token is generated.</em></p> <h2> Where This Fi…
<h2> What Changed </h2> <p>Prism ML has introduced Bonsai-27B, a 27B-class language model that leverages binary transformer weights, achieving a deployed footprint of approximately 3.9 GB. This represents a significant reduction in size, roughly 14.2 times smaller than its FP16 c…
<h2> What Changed </h2> <p>GnLOLot has introduced the MiniCPM5-1B-Claude-Opus-Fable5-Thinking model, a specialized 1-billion parameter language model designed to enhance coding and instruction-following performance. This new model is a fine-tuned version of the <code>openbmb/Mini…
<h2> Part 1: The Entropy Tracker </h2> <p><em>Part 1 of a 4-part series on system-level LLM inference internals.</em></p> <h2> What This Series Builds </h2> <p>Most LLM tooling treats inference as a black box. Hosted APIs make this worse; they strip away logits, attention weights…
