New simulators and frameworks enhance LLM training, inference, and fine-tuning
作者PulseAugur 编辑部·[114 个来源]·
Researchers have developed several new tools and frameworks to improve the efficiency and accuracy of large language model (LLM) operations. Charon and Frontier are simulators designed to predict LLM training and inference performance with high accuracy, aiding in optimization efforts. FT-Dojo provides a benchmark environment for autonomous LLM fine-tuning, while rePIRL offers an inverse RL-inspired framework for learning process reward models. Additionally, PALS focuses on power-aware LLM serving for Mixture-of-Experts models, and LlamaWeb enables memory-efficient LLM inference in web browsers using WebGPU.
AI
影响
New simulators and frameworks promise more efficient, accurate, and power-aware LLM operations, potentially accelerating research and deployment.
排序理由
Multiple research papers introducing new simulators, frameworks, and techniques for LLM training, inference, and fine-tuning.
arXiv:2605.24879v1 Announce Type: new Abstract: Large language models (LLMs) are trained on vast datasets that may contain sensitive information. Differential privacy (DP), the de facto standard for formal privacy guarantees, provides a principled framework for training LLMs with…
arXiv:2605.25704v1 Announce Type: new Abstract: In contemporary large language models (LLMs), the swish-gated linear unit (SwiGLU) activation function is widely adopted to regulate the information flow and introduce non-linearity. For large positive inputs, SwiGLU approximates th…
arXiv:2605.24956v1 Announce Type: new Abstract: Standard next-token prediction (NTP) supervises language models solely through discrete labels in the output logit space. We argue that this sparse one-hot supervision leaves the latent representation space under-constrained, allowi…
arXiv:2605.12906v2 Announce Type: replace-cross Abstract: Data selection during supervised fine-tuning (SFT) can critically change the behavior of large language models (LLMs). Although existing work has studied the effect of selecting data based on heuristics such as perplexity,…
arXiv:2603.18363v2 Announce Type: replace-cross Abstract: Unsupervised Reinforcement Learning from Internal Feedback (RLIF) has emerged as a promising paradigm for eliciting the latent capabilities of Large Language Models (LLMs) without external supervision. However, current met…
arXiv:2510.02361v2 Announce Type: replace-cross Abstract: Transformer-based large models excel in natural language processing and computer vision, but face severe computational inefficiencies due to the self-attention's quadratic complexity with input tokens. Recently, researcher…
arXiv cs.AI
TIER_1English(EN)·Muyu Pan, Shu Zhao, Nan Zhang, Philip Shin, Varun Parekh, Vijaykrishnan Narayanan, Rui Zhang·
arXiv:2605.25850v1 Announce Type: cross Abstract: This paper investigates large language model (LLM) abstention learning, specifically using ternary reward, which incentivize truthfulness in large language models. This paper extends that idea by moving from a ternary reward to a …
arXiv cs.AI
TIER_1English(EN)·Haoran Gu, Handing Wang, Yi Mei, Mengjie Zhang·
arXiv:2605.25658v1 Announce Type: cross Abstract: Expensive optimization tasks are ubiquitous in real-world applications, demanding highly specialized solvers. While LLM-driven automated solver generation shows promise, current paradigms face three critical issues when tackling e…
arXiv cs.AI
TIER_1English(EN)·Xiangtian Ji, Yuxin Chen, Zhengzhou Cai, Xiang Wang, An Zhang, Tat-Seng Chua·
arXiv:2605.24846v1 Announce Type: cross Abstract: Large language models (LLMs) display strong comprehensive abilities, yet the internal mechanisms that support these behaviors remain insufficiently understood. In this work, we show that across a wide range of open-weight Transfor…
arXiv cs.AI
TIER_1English(EN)·Jaeung Lee, Dohyun Kim, Jaemin Jo·
arXiv:2605.24614v1 Announce Type: cross Abstract: Large language model (LLM) unlearning has emerged as a crucial post-hoc mechanism for privacy protection and AI safety, yet auditing whether target knowledge is truly erased remains challenging. Existing output-level metrics fail …
arXiv:2605.24613v1 Announce Type: cross Abstract: Post-hoc repair of LLM mathematical reasoning introduces an asymmetric risk: fixing an incorrect reasoning trace is useful, but replacing a trace that was already correct can be harmful. We study this problem under a selective rep…
arXiv cs.AI
TIER_1English(EN)·Jo\~ao Sedoc, Baotong Zhang, Dean Foster·
arXiv:2605.25133v1 Announce Type: new Abstract: Reliably knowing when a language model is correct is almost as important as being correct. We introduce prover-verifier deliberation (PVD), an inference-time protocol grounded in interactive proof theory, as a mechanism for selectiv…
arXiv cs.AI
TIER_1English(EN)·Jingchu Gai, Guanning Zeng, Christina Baek, Chen Wu, J. Zico Kolter, Andrej Risteski, Aditi Raghunathan·
arXiv:2605.24396v1 Announce Type: new Abstract: Long chains of thought (CoT) from current language models frequently contain logical gaps and unjustified leaps, limiting the gains from additional test-time compute. Improving reasoning quality directly would require process reward…
arXiv cs.AI
TIER_1English(EN)·Ashok Chandrasekar, Jason Kramberger·
arXiv:2605.24217v1 Announce Type: new Abstract: As Large Language Models (LLMs) transition from research environments to production deployments, evaluating their performance against strict Service Level Objectives (SLOs) has become critical. However, current evaluation methodolog…
arXiv cs.AI
TIER_1English(EN)·Minwei Kong, Chonghe Jiang, Ao Qu, Wenbin Ouyang, Zhaoming Zeng, Xiaotong Guo, Zhekai Li, Junyi Li, Yi Fan, Xinshou Zheng, Xi Jing, Yikai Zhang, Zhiwei Liang, Seonghoo Kim, Runqing Yang, Zijian Zhou, Sirui Li, Han Zheng, Wangyang Ying, Ou Zheng, Chonghua…·
arXiv:2605.25246v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used for optimization modeling and solver-code generation, yet practical operations research and optimization problems often require a harder capability: designing scalable algorithms th…
arXiv:2604.00499v2 Announce Type: replace Abstract: To schedule LLM inference, the \textit{shortest job first} (SJF) principle is favorable by prioritizing requests with short output lengths to avoid head-of-line (HOL) blocking. Existing methods usually predict a single output le…
arXiv cs.LG
TIER_1English(EN)·Daniel Barley, Jonathan Leis, Benjamin Klenk, Holger Fr\"oning·
arXiv:2605.24006v1 Announce Type: cross Abstract: Pipeline parallelism is a key technique for distributed training of large language models because it reduces per-device parameter and activation memory. However, comparing pipeline schedules is difficult: analytical models expose …
arXiv:2605.25451v1 Announce Type: new Abstract: Training multimodal large language models (MLLMs) is challenged by both model and data heterogeneity. Existing systems redesign the training pipeline to address these challenges, but remain bound by a Pareto frontier between compute…
arXiv:2605.24331v1 Announce Type: new Abstract: Context or prompt-level reweighting has emerged as a central algorithmic lever in Reinforcement Learning with Verified Rewards (RLVR) for improving the reasoning capability of large language models, yet the principle determining wha…
arXiv cs.AI
TIER_1English(EN)·Zhuchen Cao, Sven Apel, Adish Singla, Vera Demberg·
arXiv:2502.15835v5 Announce Type: replace-cross Abstract: Pragmatic reasoning helps interlocutors infer intended meaning from ambiguous or underspecified messages by considering shared context and counterfactual alternatives. Similar challenges arise in natural language-to-code g…
arXiv:2510.14925v4 Announce Type: replace Abstract: High-confidence errors in large language models are often treated as fragile failures. We study an alternative: some errors may be false fixed points, locally stable, internally coherent, and confidently wrong. This separates ro…
arXiv:2605.26046v1 Announce Type: cross Abstract: Customizing an LLM judge to a specific task or domain often involves optimizing its prompt across multiple evaluation criteria simultaneously. Textual gradient methods automate this for a single judge criterion, however they produ…
Customizing an LLM judge to a specific task or domain often involves optimizing its prompt across multiple evaluation criteria simultaneously. Textual gradient methods automate this for a single judge criterion, however they produce natural-language critiques, not numerical vecto…
This paper investigates large language model (LLM) abstention learning, specifically using ternary reward, which incentivize truthfulness in large language models. This paper extends that idea by moving from a ternary reward to a Trajectory-Informed advantage reweighting, dynamic…
In contemporary large language models (LLMs), the swish-gated linear unit (SwiGLU) activation function is widely adopted to regulate the information flow and introduce non-linearity. For large positive inputs, SwiGLU approximates the quadratic function $x^2$, providing strong non…
Expensive optimization tasks are ubiquitous in real-world applications, demanding highly specialized solvers. While LLM-driven automated solver generation shows promise, current paradigms face three critical issues when tackling expensive optimization: factual hallucinations due …
arXiv cs.AI
TIER_1English(EN)·Zihao Jing, Qiuhao Zeng, Ruiyi Fang, Yan Yi Li, Yan Sun, Boyu Wang, Pingzhao Hu·
arXiv:2602.02780v3 Announce Type: replace Abstract: Large language models (LLMs) are enabling reasoning over 2D and 3D structures, yet existing methods remain modality-specific and typically compress structural inputs through sequence-based tokenization or fixed-length query conn…
arXiv:2605.23170v1 Announce Type: cross Abstract: Position-controlled evaluation is standard for retrieval tasks such as Needle-in-a-Haystack and RULER, but mainstream reasoning benchmarks do not control positional placement of target tasks in long contexts. We audit 11 long-cont…
arXiv:2605.23168v1 Announce Type: cross Abstract: When practitioners fine-tune LLMs on unvetted datasets, an adversary can exploit the data supply chain through task-level poisoning: inserting a small number of crafted instruction-response pairs that cause the model to embed atta…
arXiv:2602.20102v2 Announce Type: replace-cross Abstract: Despite the strong performance of large language models (LLMs) across diverse tasks, their susceptibility to adversarial attacks and unsafe content generation remains a significant obstacle to deployment, particularly in h…
arXiv:2601.17261v4 Announce Type: replace Abstract: Zeroth-Order (ZO) optimization has emerged as a promising solution for fine-tuning LLMs under strict memory constraints, as it avoids the prohibitive memory cost of storing activations for backpropagation. However, existing ZO m…
arXiv cs.AI
TIER_1English(EN)·Sixing Chen, Ji-An Li, Saner Cakir, Sinan Akcali, Kayla Lee, Marcelo G. Mattar·
arXiv:2605.06840v5 Announce Type: replace Abstract: Large language models (LLMs), especially reasoning models, generate extended chain-of-thought (CoT) reasoning that often contains explicit deliberation over future outcomes. Yet whether this deliberation constitutes genuine plan…
arXiv cs.LG
TIER_1English(EN)·Mohammad R. Rezaei, Rahul G. Krishnan·
arXiv:2605.22897v1 Announce Type: new Abstract: A persistent challenge in machine learning for scientific applications is jointly achieving prediction and understanding. Statistical models excel on structured data but operate as black boxes, while existing interpretability method…
arXiv:2605.11215v2 Announce Type: replace-cross Abstract: Pre-training large language models on massive GPU clusters has made hardware faults routine rather than rare, driving the need for resilient training systems. Yet existing frameworks either focus on specific parallelism sc…
arXiv cs.AI
TIER_1English(EN)·Yiwen Duan, Jing Ye, Xinpei Zhao·
arXiv:2602.05472v2 Announce Type: replace Abstract: The quest for expert-level reasoning in Large Language Models (LLMs) has been hampered by a persistent \textit{reward bottleneck}: traditional reinforcement learning (RL) relies on scalar rewards that are \textbf{costly} to scal…
arXiv:2605.21851v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards has become the standard recipe for improving LLM reasoning, but the dominant algorithm GRPO assigns a single trajectory-level advantage to every token, diluting the signal at pivotal re…
arXiv cs.AI
TIER_1English(EN)·Akshay Manglik (Emily), Apaar Shanker (Emily), Kaustubh Deshpande (Emily), Jason Qin (Emily), Yash Maurya (Emily), Veronica Chatrath (Emily), Vijay S. Kalmath (Emily), Levi Lentz (Emily), Yuan (Emily), Xue·
arXiv:2605.21347v2 Announce Type: new Abstract: Diagnosing failures in LLM agents remains largely manual. Practitioners inspect a small subset of execution traces, form ad-hoc hypotheses, and iterate. This process misses patterns that only emerge across trace populations and does…
arXiv:2605.21427v1 Announce Type: new Abstract: Large language model (LLM) inference has become a dominant workload in modern data centers, driving significant GPU utilization and energy consumption. While prior systems optimize throughput and latency by batching, scheduling, and…
arXiv:2605.20194v1 Announce Type: cross Abstract: Large language models (LLMs) have been increasingly used to analyze text. However, they are often plagued with contextual reasoning limitations when analyzing long documents. When long documents are processed sequentially, early o…
arXiv:2605.20706v1 Announce Type: cross Abstract: Running language models in the browser presents a unique opportunity to build efficient, private, and portable AI applications, but requires contending with constrained memory availability and heterogeneous hardware targets. To re…
arXiv:2605.21312v1 Announce Type: cross Abstract: Modern LLM serving is no longer homogeneous or monolithic. Production systems now combine disaggregated execution, complex parallelism, runtime optimizations, and stateful workloads such as reasoning, agents, and RL rollouts. Simu…
arXiv cs.AI
TIER_1English(EN)·Jaemin Kim, Hangeol Chang, Hyunmin Hwang, Choonghan Kim, Jong Chul Ye·
arXiv:2505.19075v3 Announce Type: replace Abstract: Large Language Models (LLMs) have demonstrated remarkable general capabilities, but enhancing skills such as reasoning often demands substantial computational resources and may compromise generalization. While Parameter-Efficien…
arXiv:2603.01712v2 Announce Type: replace Abstract: Fine-tuning large language models for vertical domains remains labor-intensive, requiring practitioners to curate data, configure training, and iteratively diagnose model behavior. Despite growing interest in autonomous machine …
arXiv:2602.07832v2 Announce Type: replace-cross Abstract: Process rewards have been widely used in deep reinforcement learning to improve training efficiency, reduce variance, and prevent reward hacking. In LLM reasoning, existing works also explore various solutions for learning…
arXiv:2605.17164v2 Announce Type: replace-cross Abstract: Deploying large-scale LLM training and inference with optimal performance is exceptionally challenging due to a complex design space of parallelism strategies, system optimizations, and hardware configurations. Accurate an…
arXiv:2605.19362v2 Announce Type: replace-cross Abstract: Users often interpret and select agent skills through their SKILL markdown specifications. To protect users, existing audits mainly focus on malicious or unsafe skills. We study the complementary question of whether specif…
arXiv cs.CL
TIER_1English(EN)·Zhenwei Tang, Zhaoyan Liu, Rasa Hosseinzadeh, Tongzi Wu, Keyvan Golestan, Jesse C. Cresswell·
arXiv:2605.21748v1 Announce Type: new Abstract: As interactive LLM-based applications are created and refined, model developers need to evaluate the quality of generated text along many possible axes. For simpler systems, human evaluation may be practical, but in complicated syst…
arXiv:2605.22389v1 Announce Type: new Abstract: Effectively training Large Language Models (LLMs) for complex, long-CoT reasoning is often bottlenecked by the need for massive high-quality reasoning data. Existing methods are either computationally expensive or fail to reliably d…
arXiv cs.CL
TIER_1English(EN)·Arip Asadulaev, Daniil Ognev, Karim Salta, Martin Takac·
arXiv:2605.21654v1 Announce Type: cross Abstract: Reinforcement learning substantially improves pretrained language models, but it remains understudied why critic-free methods such as PPO and GRPO work as well as they do, and when they should provide the largest gains. We develop…
arXiv:2605.15588v2 Announce Type: replace Abstract: As large language models (LLMs) are deployed in consequential settings such as medical question answering and legal reasoning, the ability to estimate when their outputs are likely to be correct is essential for safe and reliabl…
arXiv:2603.27355v2 Announce Type: replace-cross Abstract: We present a readiness harness for LLM and RAG applications that turns evaluation into a deployment decision workflow. The system combines automated benchmarks, OpenTelemetry observability, and CI quality gates under a min…
arXiv cs.LG
TIER_1English(EN)·Andy Han, Kristina Fujimoto, Avidan Shah, Kiet Nguyen, Kai Xu, Chen Yueh-Han, Ilia Sucholutsky, Rico Angell·
arXiv:2605.21834v1 Announce Type: new Abstract: Aligned models can misbehave in several ways: they are often sycophantic, fall victim to jailbreaks, or fail to include appropriate safety warnings. Consistency training is a promising new alignment paradigm to mitigate such failure…
arXiv:2605.21856v1 Announce Type: new Abstract: Large language models (LLMs) have demonstrated impressive reasoning abilities across a wide range of tasks, but data contamination undermines the objective evaluation of these capabilities. This problem is further exacerbated by mal…
arXiv:2605.21975v1 Announce Type: new Abstract: Financial markets are characterized by extreme non-stationarity, low signal-to-noise ratios, and strong dependence on external information such as news, company fundamentals, and macroeconomic signals. Yet, existing approaches eithe…
arXiv:2605.22156v1 Announce Type: new Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has become a promising paradigm for scaling reasoning capabilities of Large Language Models (LLMs). However, the sparsity of binary verifier rewards often leads to low efficiency…
arXiv cs.LG
TIER_1English(EN)·Manuel Noah Riesen, Peter Alfred von Niederh\"ausern·
arXiv:2605.22195v1 Announce Type: new Abstract: Graph of Thoughts (GoT), a generalized form of recent prompting paradigms for large language models (LLMs), has been shown to be useful for elaborate problem solving. By executing a graph of operations, thoughts of the LLM are struc…
arXiv cs.LG
TIER_1English(EN)·Hongbin Zhang, Chaozheng Wang, Kehai Chen, Youcheng Pan, Yang Xiang, Jinpeng Wang, Min Zhang·
arXiv:2605.22263v1 Announce Type: new Abstract: On-policy self-distillation (OPSD) is an emerging LLM post-training paradigm in which the model serves as its own teacher: conditioned on privileged information such as a reference trace or hint, the same policy provides dense token…
arXiv:2605.22297v1 Announce Type: new Abstract: Learning rate configuration is a fundamental aspect of modern deep learning. The prevailing practice of applying a uniform learning rate across all layers overlooks the structural heterogeneity of Transformers, potentially limiting …
arXiv:2506.16659v3 Announce Type: replace Abstract: Training large language models (LLMs) relies on adaptive optimizers such as Adam, which introduce extra operations and require significantly more memory to maintain first- and second-order moments than SGD. While recent works su…
arXiv:2602.00688v2 Announce Type: replace Abstract: Fine-tuning large language models (LLMs) on sensitive datasets raises privacy concerns, as training data extraction (TDE) attacks can expose highly confidential information. Existing defenses against such attacks either lack for…
arXiv:2602.12506v3 Announce Type: replace Abstract: Reinforcement learning (RL) finetuning has become a key technique for enhancing large language models (LLMs) on reasoning-intensive tasks, motivating its extension to vision-language models (VLMs). While RL-tuned VLMs improve on…
arXiv:2605.10067v3 Announce Type: replace Abstract: Red teaming is critical for uncovering vulnerabilities in Large Language Models (LLMs). While automated methods have improved scalability, existing approaches often rely on static heuristics or stochastic search, rendering them …
Position-controlled evaluation is standard for retrieval tasks such as Needle-in-a-Haystack and RULER, but mainstream reasoning benchmarks do not control positional placement of target tasks in long contexts. We audit 11 long-context benchmarks and find none jointly controls task…
Effectively training Large Language Models (LLMs) for complex, long-CoT reasoning is often bottlenecked by the need for massive high-quality reasoning data. Existing methods are either computationally expensive or fail to reliably distinguish high- from low-quality reasoning samp…
Self-evolving skill libraries, pioneered by Voyager, let frozen LLM agents accumulate reusable knowledge without weight updates, yet recent evaluation shows that LLM-authored skills deliver $+0.0$pp over no-skill baselines while human-curated ones deliver $+16.2$pp: the bottlenec…
Self-evolving skill libraries, pioneered by Voyager, let frozen LLM agents accumulate reusable knowledge without weight updates, yet recent evaluation shows that LLM-authored skills deliver $+0.0$pp over no-skill baselines while human-curated ones deliver $+16.2$pp: the bottlenec…
A black-box detection method called Zero-CoT Probe is introduced to identify data contamination in large language models by truncating reasoning processes and comparing performance on original and perturbed datasets.
Reinforcement learning with verifiable rewards (RLVR) has become a dominant paradigm for improving reasoning in large language models (LLMs), yet the underlying geometry of the resulting parameter trajectories remains underexplored. In this work, we demonstrate that RLVR weight t…
Large language model (LLM) inference has become a dominant workload in modern data centers, driving significant GPU utilization and energy consumption. While prior systems optimize throughput and latency by batching, scheduling, and parallelism, they largely treat GPU power as a …
Large language model (LLM) inference has become a dominant workload in modern data centers, driving significant GPU utilization and energy consumption. While prior systems optimize throughput and latency by batching, scheduling, and parallelism, they largely treat GPU power as a …
Diagnosing failures in LLM agents remains largely manual. Practitioners inspect a small subset of execution traces, form ad-hoc hypotheses, and iterate. This process misses patterns that only emerge across trace populations and does not scale to production corpora where individua…
Diagnosing failures in LLM agents remains largely manual. Practitioners inspect a small subset of execution traces, form ad-hoc hypotheses, and iterate. This process misses patterns that only emerge across trace populations and does not scale to production corpora where individua…
Modern LLM serving is no longer homogeneous or monolithic. Production systems now combine disaggregated execution, complex parallelism, runtime optimizations, and stateful workloads such as reasoning, agents, and RL rollouts. Simulation is attractive for exploring this growing de…
Running language models in the browser presents a unique opportunity to build efficient, private, and portable AI applications, but requires contending with constrained memory availability and heterogeneous hardware targets. To realize this opportunity, we present Llamas on the W…
Diffusion Large Language Models (dLLMs) have emerged as a competitive alternative to autoregressive (AR) models, offering better hardware utilization and bidirectional context through parallel block-level decoding. However, as dLLMs continue to scale up with mixture-of-experts (M…
Tool-augmented reasoning has emerged as a promising direction for enhancing the reasoning capabilities of multimodal large language models (MLLMs). However, existing studies mainly focus on enabling models to perform tool invocation, while neglecting the necessity of invoking too…
LLM discovery and optimization systems are increasingly applied across domains, implementing a common propose-evaluate-revise loop. Such optimization or discovery progresses via context conditioning on received feedback from an environment. However, as modern LLM agents are incre…
LLM discovery and optimization systems are increasingly applied across domains, implementing a common propose-evaluate-revise loop. Such optimization or discovery progresses via context conditioning on received feedback from an environment. However, as modern LLM agents are incre…
Evaluating large language models (LLMs) on natural-language logical reasoning is essential because rule-governed tasks require conclusions to follow strictly from stated premises. Many existing logical-reasoning benchmarks are generated by templating natural-language items from s…
Large language models (LLMs) have achieved remarkable success in complex reasoning tasks via long chain-of-thought (CoT), yet their immense computational overhead hinders real-world deployment. LLM reasoning distillation addresses this by transferring reasoning capabilities from …
Entropy-based deep reasoning has emerged as a promising direction for improving the reasoning capabilities of Large Language Models (LLMs), but existing methods often either increase response length indiscriminately or shorten responses at the cost of accuracy. To better balance …
Large Language Models have achieved strong performance on reasoning tasks with objective answers by generating step-by-step solutions, but diagnosing where a multi-step reasoning trace might fail remains difficult. Confidence estimation offers a diagnostic signal, yet existing me…
arXiv cs.AI
TIER_1English(EN)·Pascal Van Hentenryck·
Optimization models developed by operations research (OR) experts are often deployed as decision-support systems in industrial settings. However, real-world environments are dynamic, with evolving business rules, previously overlooked constraints, and unforeseen perturbations. In…
Large Language Models (LLMs) are increasingly deployed as scientific AI as- sistants, and a growing body of benchmarks evaluates their capabilities across knowledge retrieval, reasoning, code generation, and tool use. These evaluations, however, typically assume the scientific pr…
Supporting long-context LLMs is challenging due to the substantial memory demands of the key-value (KV) cache. Existing offloading systems store the full cache in host memory and selectively fetch critical entries during decoding, but this strategy quickly hits a ceiling: sparsit…
Vectorization via Single Instruction, Multiple Data (SIMD) architectures is a cornerstone of high-performance computing. To fully exploit hardware potential, developers often resort to explicit vectorization using intrinsics, as compiler-based auto-vectorization frequently yields…
Whether machines can originate novel content has been debated for nearly two centuries, from Lovelace's assertion that no engine can "originate anything" to Turing's question of whether a machine can amplify ideas brought in from outside. Multi-large language model (LLM) systems,…
Second-order methods offer an attractive path toward more sample-efficient LLM training, but their practical use is often blocked by the systems cost of maintaining and updating large matrix-based optimizer states. We introduce \textbf{Asteria}, a runtime system designed to remov…
Rule2DRC introduces a large-scale benchmark for DRC script synthesis with 1,000 rule-to-script tasks and 13,921 evaluation layouts, along with SplitTester which improves program selection through execution-based feedback.
arXiv:2605.25571v1 Announce Type: new Abstract: Post-training via Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) is crucial for enhancing reasoning in Multimodal Large Language Models (MLLMs), yet existing paradigms often reach a performance bottleneck due to the li…
arXiv:2605.23362v1 Announce Type: cross Abstract: Evaluating large language models increasingly relies on LLM-as-a-judge protocols, but such evaluations remain costly: different judges have different prices and reliabilities, and the difficulty of each prompt-response pair can va…
Context or prompt-level reweighting has emerged as a central algorithmic lever in Reinforcement Learning with Verified Rewards (RLVR) for improving the reasoning capability of large language models, yet the principle determining what constitutes an optimal weighting remains poorl…
Evaluating large language models increasingly relies on LLM-as-a-judge protocols, but such evaluations remain costly: different judges have different prices and reliabilities, and the difficulty of each prompt-response pair can vary substantially. This raises a basic allocation q…
arXiv:2605.20270v1 Announce Type: cross Abstract: A local specialist LLM, fine-tuned with reinforcement learning from verifiable rewards (RLVR) on operator-local data, is installed in a regulated organization with per-deployment error budget $\alpha$. The operator needs a safety …
arXiv:2504.07347v3 Announce Type: replace Abstract: As demand for Large Language Models (LLMs) and AI agents grows rapidly, optimizing systems for efficient LLM inference becomes critical. While significant efforts have targeted system-level engineering, little has been explored …
A local specialist LLM, fine-tuned with reinforcement learning from verifiable rewards (RLVR) on operator-local data, is installed in a regulated organization with per-deployment error budget $α$. The operator needs a safety certificate for this deployment's stream at every round…
arXiv stat.ML
TIER_1English(EN)·Ruicheng Ao, Gan Luo, David Simchi-Levi, Xinshang Wang·
arXiv:2504.11320v3 Announce Type: replace-cross Abstract: Large language models now serve millions of users daily, with providers incurring costs exceeding $700,000 per day. Each request requires token-by-token inference, making GPU scheduling central to latency, capacity, and co…
New research shows LLMs can optimize database query execution plans—achieving up to 4.78x speedups by correcting the cardinality estimation errors that statistical heuristics miss.
<p><em>A daily deep dive into llm topics, coding problems, and platform features from <a href="https://pixelbank.dev" rel="noopener noreferrer">PixelBank</a>.</em></p> <h2> Topic Deep Dive: Applications of LLMs </h2> <p><em>From the Introduction to LLMs chapter</em></p> <h2> Intr…
<blockquote> <p>Cross-posted from <a href="https://carrick.tools/blog/benchmarking-llm-structured-outputs/" rel="noopener noreferrer">carrick.tools</a>.</p> </blockquote> <p>When you read the API documentation for OpenAI, Anthropic, or Google Gemini, the feature called "structure…
<h2> Introduction to LLM Inference Caching: Why It Matters? </h2> <p>When working with Large Language Models (LLMs), especially as you start using them in production environments, one of the first major challenges you'll face is the delicate balance between cost and latency. LLMs…
dev.to — LLM tag
TIER_1English(EN)·Nishkarsh Sahu·
<p>Recently I’ve been experimenting with integrating local AI runtimes into Rails applications using tools like Ollama and LM Studio.</p> <p>At first, the integration looked straightforward:<br /> make an HTTP request, stream the response, and return the generated text.</p> <p>Bu…
dev.to — LLM tag
TIER_1English(EN)·Kotcherla Murali Krishna·
<p>Why vLLM, TensorRT-LLM, and llama.cpp each solve only part of the problem — and how I built inferx to fill the gap. Runs on any laptop, no GPU needed.</p> <p>I spent the last few months building inferx — an open-source LLM inference optimization library that runs on any machin…
[Перевод] Масштабирование LLM: от одного чипа до ЦОДа. Глава 2. Шардинг Это продолжение цикла статей о масштабировании тренировки и инференса LLM. Предыдущая глава находится по этой ссылке . Итак, с основами разобрались, давайте теперь разбираться с тем, как распихать матрицы по …
