9030 club reading "Reinforcement Learning via Self-Distillation": #63 - ML Paper Reading Group

​☕️📝 Paper Link 📝☕️

​https://arxiv.org/abs/2601.20802

Can an LLM turn feedback on why it failed into its own dense training signal—and beat standard RLVR without a teacher or reward model?

​Abstract; Large language models are increasingly post-trained with reinforcement learning in verifiable domains such as code and math. Yet, current methods for reinforcement learning with verifiable rewards (RLVR) learn only from a scalar outcome reward per attempt, creating a severe credit-assignment bottleneck. Many verifiable environments actually provide rich textual feedback, such as runtime errors or judge evaluations, that explain why an attempt failed. We formalize this setting as reinforcement learning with rich feedback and introduce Self-Distillation Policy Optimization (SDPO), which converts tokenized feedback into a dense learning signal without any external teacher or explicit reward model. SDPO treats the current model conditioned on feedback as a self-teacher and distills its feedback-informed next-token predictions back into the policy. In this way, SDPO leverages the model's ability to retrospectively identify its own mistakes in-context. Across scientific reasoning, tool use, and competitive programming on LiveCodeBench v6, SDPO improves sample efficiency and final accuracy over strong RLVR baselines. Notably, SDPO also outperforms baselines in standard RLVR environments that only return scalar feedback by using successful rollouts as implicit feedback for failed attempts. Finally, applying SDPO to individual questions at test time accelerates discovery on difficult binary-reward tasks, achieving the same discovery probability as best-of-k sampling or multi-turn conversations with 3x fewer attempts.

​Event Schedule;

​7pm to 8pm --> quiet reading time, grab and snack and read! (optional)

​8pm to 9pm --> open discussion about reading 📝

​9pm --> we have our space for a bit longer, stay to socialize or network!

​Our event is hosted within Mox SF, the gracious donors of the space we will meet.