Entity · technique

Experiential Learning

techniqueactiveprovisionalexperiential-learning-35518cba·1 events·first seen 9h ago

Aliases: Experiential Learning

Co-occurring entities

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

More like this (12)

LLM-as-a-Coach: Experiential Learning for Non-Verifiable Tasks Personalized Evaluation as Learning in-context learning Curiosity-Driven Learning active learning Judgement-of-Learning Continual Learning Applied Intuition Planning Experience Exploration and Utilization curriculum learning Hindsight Experience Replay closed-loop experimentation

Recent events (1)

6arXiv · cs.CL·9h ago·source ↗

Experiential Learning proposes LLM-as-a-Coach to replace scalar RL rewards with textual feedback distillation

Researchers introduce Experiential Learning (EL), a post-training method that replaces scalar RL rewards with rich textual feedback from an LLM-as-a-Coach, which distills per-response assessments into transferable experiential knowledge used to condition a teacher model and update the policy via context distillation. Evaluated across two policy families with both self-feedback and proprietary model feedback, EL consistently outperforms rubric-based RL on open-ended tasks and shows better out-of-distribution generalization while mitigating reward hacking. The work targets a known weakness of RLHF-style training on non-verifiable tasks where scalar rewards lose fine-grained preference signal.

Evaluation and Benchmarking Alignment and RLHF LLM-as-a-Judge LLM-as-a-Coach: Experiential Learning for Non-Verifiable Tasks Experiential Learning