Tag: Reinforcement Learning

All the articles with the tag "Reinforcement Learning".

• Learning When to Think: Shaping Adaptive Reasoning in R1-Style Models via Multi-Stage RL

  Published: 20 May, 2025 |  at  11:10 AM

  

  92.09 🤔

  本文提出 *AutoThink*，通过省略号提示和多阶段强化学习框架，使 R1 风格大型推理模型根据问题复杂性自适应地决定是否进行显式推理，在五个数学基准上实现了准确性和效率的优越权衡。

• Temporal Sampling for Forgotten Reasoning in LLMs

  Published: 28 May, 2025 |  at  11:20 AM

  

  92.01 🤔

  本文揭示了大型语言模型微调中的'Temporal Forgetting'现象，并提出'Temporal Sampling'方法，通过从多个训练检查点采样答案显著提升推理性能（Pass@k提升4-19个百分点），并通过LoRA适配降低存储成本。

• Learning to Think: Information-Theoretic Reinforcement Fine-Tuning for LLMs

  Published: 17 May, 2025 |  at  11:02 AM

  

  91.74 🤔

  This paper introduces Learning to Think (L2T), an information-theoretic reinforcement fine-tuning framework for LLMs that uses a universal dense process reward to optimize reasoning effectiveness and efficiency, achieving significant accuracy and token efficiency gains on math reasoning benchmarks.

• ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models

  Published: 3 Jun, 2025 |  at  11:45 AM

  

  91.52 🤔

  本文提出ProRL方法，通过长时间强化学习结合KL散度惩罚和参考策略重置，在多样化任务上训练Nemotron-Research-Reasoning-Qwen-1.5B模型，显著扩展了大型语言模型的推理边界，尤其在基础模型表现较差的领域和分布外任务上表现出色。

• Not All Thoughts are Generated Equal: Efficient LLM Reasoning via Multi-Turn Reinforcement Learning

  Published: 24 May, 2025 |  at  11:13 AM

  

  91.13 🤔

  本文提出Long⊗Short框架，通过长思维和短思维LLM协作推理，利用自动思维分块、冷启动SFT和多轮RL优化，显著提升推理效率，在多个基准上使Qwen2.5-7B和Llama3.1-8B性能接近蒸馏模型，同时减少token长度超80%。