Two is better than one: A Collapse-free Multi-Reward RLIF Training Framework
cs.LG, cs.CL
Submitted: 2026-05-21
Updated: 2026-10-31
Code: https://github.com/huggingface/lighteval
Terminology
Sources
- The Unreasonable Effectiveness of Entropy Minimization in LLM Reasoning
- Exploration vs Exploitation: Rethinking RLVR through Clipping, Entropy, and Spurious Reward
- Training Verifiers to Solve Math Word Problems
- The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- Scaling Laws for Reward Model Overoptimization
- CRUXEval: A Benchmark for Code Reasoning, Understanding and Execution
- How Far Can Unsupervised RLVR Scale LLM Training?
- Measuring Mathematical Problem Solving With the MATH Dataset
- R-Zero: Self-Evolving Reasoning LLM from Zero Data
- LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code
- GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization
- Maximizing Confidence Alone Improves Reasoning
- Can Large Reasoning Models Self-Train?
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- MiMo: Unlocking the Reasoning Potential of Language Model -- From Pretraining to Posttraining
- Qwen2.5 Technical Report
- SSR-Zero: Simple Self-Rewarding Reinforcement Learning for Machine Translation
- Right Question is Already Half the Answer: Fully Unsupervised LLM Reasoning Incentivization
- No Free Lunch: Rethinking Internal Feedback for LLM Reasoning
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis