5 papers
Athanor: An Agent-Native System for Autonomous Post-Training Research
Yufa Zhou, Yichen Wang, Jiakun Fan +5
Recursive self-improvement, in which AI systems build stronger AI systems, is one promising route to superintelligence. We study a weaker form that can be measured today: frontier…
Factorization-Guided Training and Inference for Diffusion-Drafted Speculative Decoding
Jiakun Fan, Min Xiao, Qiang Wang +5
Speculative decoding accelerates large language model inference through parallel verification of tokens proposed by a drafter. Diffusion drafters further parallelize proposal gener…
Where Does Staleness Accumulate? Pool Aware Effective Staleness Control for Asynchronous RL in LLM Post-Training
Chenliang Li, Neiwen Ling, Zijun Wei +1
Fully asynchronous reinforcement learning (RL) improves resource utilization in large language model post-training by overlapping rollout generation with policy optimization, but i…
Diagnosing Training Inference Mismatch in LLM Reinforcement Learning via a Zero-Mismatch Reference
Tianle Zhong, Neiwen Ling, Yifan Pi +5
Modern LLM RL systems separate rollout generation from policy optimization. These two stages are expected to produce token probabilities that match exactly. However, implementation…
TimelyLLM: Segmented LLM Serving System for Time-sensitive Robotic Applications
Neiwen Ling, Guojun Chen, Lin Zhong
Large Language Models (LLMs) such as GPT-4 and Llama3 can already comprehend complex commands and process diverse tasks. This advancement facilitates their application in controlli…