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

Hugging Face Daily Papers Papers

Summary

This paper introduces Experiential Learning (EL), a method that repurposes an LLM-as-a-Judge into an LLM-as-a-Coach to provide rich textual feedback instead of scalar rewards, improving performance and generalization on open-ended non-verifiable tasks.

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.
Original Article
View Cached Full Text

Cached at: 07/21/26, 06:35 AM

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

Source: https://huggingface.co/papers/2607.18110 Published on Jul 20

·

Submitted byhttps://huggingface.co/ytz20

ytzon Jul 21

Abstract

Reinforcementlearning(RL)onopen-endedtaskscompressesanLLM’srubric-basedevaluationintoascalarreward,discardingrichtextualfeedbackandconflatingresponseswithdistinctqualityprofiles.WeproposeExperientialLearning(EL),whichrepurposesthefeedbackmodelfromanLLM-as-a-JudgeintoanLLM-as-a-Coach.Thecoachdistillsitsassessmentofeachon-policyresponseintotransferableexperientialknowledge,whichconditionsateachermodelandisinternalizedbythepolicythroughon-policycontextdistillation.Comparedwithscalarrewards,thishigher-bandwidthfeedbackchannelprovidesdensesupervisionandpreservesfine-grainedpreferencesamonghigh-qualityresponses.Acrosstwopolicyfamilies,withfeedbackfromthepolicyitselforaproprietarymodel,ELconsistentlyoutperformsrubric-basedRLonheld-outandunseenopen-endedtasks.Notably,ELgeneralizesbetterbeyondthetrainingdistribution,andmitigatesrewardhacking.Thesefindingsestablishexperientialknowledgeasaricherandmoregeneralizablelearningsignalforpost-trainingonnon-verifiabletasks.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2607\.18110

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2607.18110 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2607.18110 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2607.18110 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

LLM-as-a-Tutor: Policy-Aware Prompt Adaptation for Non-Verifiable RL

Hugging Face Daily Papers

LLM-as-a-Tutor introduces a framework that extends LLM's role from judge to tutor by dynamically adjusting prompt difficulty through pairwise comparison and constraint addition, improving instruction-following performance in reinforcement learning.

LLM-as-Judge in Education: A Curriculum-Grounded Marking Pipeline

arXiv cs.AI

This paper presents a curriculum-grounded LLM-as-Judge pipeline for automated question-level marking in high-stakes exam preparation, using syllabus artefacts and marking guidelines to improve consistency and transparency, with preliminary evaluation showing outcomes comparable to human tutors.