ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning

Hugging Face Daily Papers Papers
reinforcement-learning reasoning llm negative-trajectories reflection on-policy-training

Summary

ReflectRL is a framework that learns from 'golden negative trajectories' (failed reasoning attempts by expert models) by reflecting on them, then transfers this reflective reasoning back to direct reasoning, improving LLM performance across benchmarks.

On-policy training has emerged as a powerful post-training paradigm for improving the reasoning capabilities of large language models, and is often enhanced by golden trajectories from stronger expert models. However, when the expert fails on harder problems, existing trajectory-guided methods lose their main source of supervision, and these failed trajectories are typically discarded as negative samples. We argue that such failures, which we call Golden Negative Trajectories, can still provide valuable reasoning signals when treated not as demonstrations to imitate, but as flawed trajectories to reflect upon. We identify a Reflection Advantage: for hard problems, reflecting on a flawed trajectory can be easier and more effective than solving the problem directly from scratch. Motivated by this, we propose ReflectRL, a lightweight plug-and-play framework that learns from Golden Negative Trajectories during on-policy training. ReflectRL first uses these trajectories to elicit Reflective Reasoning, then applies Reflective-to-Direct Policy Transition to transfer the acquired reasoning behavior back to Direct Reasoning. Experiments across 9 benchmarks, 4 LLM backbones, and 4 on-policy training methods show that ReflectRL consistently improves reasoning performance with minimal overhead.
Original Article
View Cached Full Text

Cached at: 08/05/26, 05:46 PM

Paper page - ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning

Source: https://huggingface.co/papers/2608.03972 Authors:

,

,

,

,

,

,

,

,

,

,

,

Abstract

On-policytraininghasemergedasapowerfulpost-trainingparadigmforimprovingthereasoningcapabilitiesoflargelanguagemodels,andisoftenenhancedbygoldentrajectoriesfromstrongerexpertmodels.However,whentheexpertfailsonharderproblems,existingtrajectory-guidedmethodslosetheirmainsourceofsupervision,andthesefailedtrajectoriesaretypicallydiscardedasnegativesamples.Wearguethatsuchfailures,whichwecallGoldenNegativeTrajectories,canstillprovidevaluablereasoningsignalswhentreatednotasdemonstrationstoimitate,butasflawedtrajectoriestoreflectupon.WeidentifyaReflectionAdvantage:forhardproblems,reflectingonaflawedtrajectorycanbeeasierandmoreeffectivethansolvingtheproblemdirectlyfromscratch.Motivatedbythis,weproposeReflectRL,alightweightplug-and-playframeworkthatlearnsfromGoldenNegativeTrajectoriesduringon-policytraining.ReflectRLfirstusesthesetrajectoriestoelicitReflectiveReasoning,thenappliesReflective-to-DirectPolicyTransitiontotransfertheacquiredreasoningbehaviorbacktoDirectReasoning.Experimentsacross9benchmarks,4LLMbackbones,and4on-policytrainingmethodsshowthatReflectRLconsistentlyimprovesreasoningperformancewithminimaloverhead.

View arXiv pageView PDFGitHub1Add to collection

Get this paper in your agent:

hf papers read 2608\.03972

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/2608.03972 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.03972 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.03972 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