Tag: Curriculum Learning

All the articles with the tag "Curriculum Learning".

• Can Large Reasoning Models Self-Train?

  Published: 1 Jun, 2025 |  at  11:43 AM

  

  86.73 🤔

  本文提出Self-Rewarded Training (SRT) 方法，通过模型自一致性驱动强化学习实现无监督数学推理能力提升，初期性能媲美有监督方法，但因奖励黑客问题导致长期训练性能崩溃，并探索了提前停止和课程学习等缓解策略。

• ZeroSearch: Incentivize the Search Capability of LLMs without Searching

  Published: 8 May, 2025 |  at  06:16 PM

  

  86.09 🤔

  ZEROSEARCH introduces a reinforcement learning framework that enhances LLMs' search capabilities by simulating search engines with fine-tuned LLMs, achieving performance comparable to or better than real search engines without API costs through a curriculum-based rollout strategy.

• SATURN: SAT-based Reinforcement Learning to Unleash Language Model Reasoning

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

  

  85.35 🤔

  SATURN提出一个基于SAT问题的强化学习框架，通过课程学习和可控难度的SAT任务显著提升大型语言模型在SAT、数学和编程任务上的推理能力。

• Learning Like Humans: Advancing LLM Reasoning Capabilities via Adaptive Difficulty Curriculum Learning and Expert-Guided Self-Reformulation

  Published: 17 May, 2025 |  at  11:01 AM

  

  85.16 🤔

  This paper introduces Adaptive Difficulty Curriculum Learning (ADCL) and Expert-Guided Self-Reformulation (EGSR) to enhance LLM reasoning by dynamically adjusting training curricula and guiding models to reformulate expert solutions, achieving significant performance improvements over standard RL baselines on mathematical reasoning benchmarks.