Popular repositories Loading

1. TAAC-0.829 TAAC-0.829 Public

   2026 TAAC solution (Score: 0.8290, Baseline: 0.8112). Featured with time-series extensions, semantic alignment, and social calendar features.

   Python 3

2. grpo-all-you-need-to-know grpo-all-you-need-to-know Public

   从“为什么能去掉 critic”讲到“训练为什么学不动”：手算组内优势，拆解奖励与归一化，串起 Dr.GRPO、DAPO、GSPO 的改进动机。中文推导、核心代码、训练排坑，含 37 道章末练习与参考答案。

   JavaScript 2

3. ppo-all-you-need-to-know ppo-all-you-need-to-know Public

   从概率论一步步推到 PPO：讲清公式的来由、近似与实现，拆开 GAE 与 clip 的细节，把 PyTorch 代码接到训练诊断与排坑。中文系统教程，含 31 道章末练习与参考答案。

   JavaScript 1

4. CreditORLM CreditORLM Public

   Consistency verification and object-level credit assignment for LLM optimization modeling.

   Python 1

5. Trip-RL Trip-RL Public

   Long-horizon travel planning agents with curriculum reinforcement learning and online policy distillation

   Python 1