Publications (7)
Reward Modeling from Natural Language Human Feedback
Zongqi Wang, Rui Wang, Yuchuan Wu +5
Reinforcement Learning with Verifiable reward (RLVR) on preference data has become the mainstream approach for training Generative Reward Models (GRMs). Typically in pairwise rewar…
Improve LLM-as-a-Judge Ability as a General Ability
Jiachen Yu, Shaoning Sun, Xiaohui Hu +3
LLM-as-a-Judge leverages the generative and reasoning capabilities of large language models (LLMs) to evaluate LLM responses across diverse scenarios, providing accurate preference…
S2J: Bridging the Gap Between Solving and Judging Ability in Generative Reward Models
Shaoning Sun, Jiachen Yu, Zongqi Wang +3
With the rapid development of large language models (LLMs), generative reward models (GRMs) have been widely adopted for reward modeling and evaluation. Previous studies have prima…
Internalize the Temperature: On-Policy Self-Distillation as Policy Reheater for Reinforcement Learning
Xuewei Yang, Jiachen Yu, Jie Wu +3
Reinforcement learning from verifiable rewards improves the reasoning ability of large language models, but often suffers from entropy collapse, in which increasingly concentrated…
Distributional Clarity: The Hidden Driver of RL-Friendliness in Large Language Models
Shaoning Sun, Mingzhu Cai, Huang He +5
Language model families exhibit striking disparity in their capacity to benefit from reinforcement learning: under identical training, models like Qwen achieve substantial gains, w…
ProFit: Leveraging High-Value Signals in SFT via Probability-Guided Token Selection
Tao Liu, Taiqiang Wu, Runming Yang +3
Supervised fine-tuning (SFT) is a fundamental post-training strategy to align Large Language Models (LLMs) with human intent. However, traditional SFT often ignores the one-to-many…
Kimi K3: Open Frontier Intelligence
Kimi Team, Tongtong Bai, Yifan Bai +398
We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window. Kimi K3 is…