English

LLM-as-a-Coach: Experiential Learning for Non-Verifiable Tasks

Machine Learning 2026-07-20 v1 Computation and Language

Abstract

Reinforcement learning (RL) on open-ended tasks compresses an LLM's rubric-based evaluation into a scalar reward, discarding rich textual feedback and conflating responses with distinct quality profiles. We propose Experiential Learning (EL), which repurposes the feedback model from an LLM-as-a-Judge into an LLM-as-a-Coach. The coach distills its assessment of each on-policy response into transferable experiential knowledge, which conditions a teacher model and is internalized by the policy through on-policy context distillation. Compared with scalar rewards, this higher-bandwidth feedback channel provides dense supervision and preserves fine-grained preferences among high-quality responses. Across two policy families, with feedback from the policy itself or a proprietary model, EL consistently outperforms rubric-based RL on held-out and unseen open-ended tasks. Notably, EL generalizes better beyond the training distribution, and mitigates reward hacking. These findings establish experiential knowledge as a richer and more generalizable learning signal for post-training on non-verifiable tasks.

Cite

@article{arxiv.2607.18110,
  title  = {LLM-as-a-Coach: Experiential Learning for Non-Verifiable Tasks},
  author = {Tianzhu Ye and Li Dong and Guanheng Chen and He Zhu and Xun Wu and Shaohan Huang and Furu Wei},
  journal= {arXiv preprint arXiv:2607.18110},
  year   = {2026}
}