Researchers propose a new method for asynchronous reinforcement learning in large language models, addressing the issue of stale rollouts generated by earlier policies. Their method, GMC-GRPO, provides improved convergence guarantees and better performance in experiments.

Researchers propose an alternative reconstruction loss function (RMSE) to improve post-training quantization of large language models (LLMs) by decoupling optimization strength from the reconstruction loss scale.

Researchers propose a post-training method for multi-token prediction heads in language models, achieving similar speedup to joint pre-training with significantly fewer tokens. They also introduce a relaxation of draft token verification and an adaptive controller for dynamic MTP head engagement.

A self-evolving framework, RuleEvolve, for coding rules in AI coding agents uses an LLM-powered mutator module to generate variants and a judge module to evaluate and update the pool with the best-performing ones, outperforming manual engineering and existing prompt optimization baselines in functional correctness, code length, and generation cost.

Researchers have developed a method for optimizing CPU speech synthesis in serverless architectures by reducing idle inference state and improving concurrency, resulting in significant cost savings and improved performance.

The paper proposes OSCAR, a framework for verifying improvements and allocating attempts across large language models (LLMs) for optimization modeling. OSCAR uses a certified simulator to compare candidates and continues searching beyond feasibility. It achieves high accuracy and reduces costs compared to other LLMs like Codex and Claude Code.

MoMHa is a system that optimizes large language model (LLM) harnesses by considering multiple objectives: accuracy, safety, and token cost. It uses an agentic proposer to search for optimal harness design and outperforms alternative approaches in various domains.