ReflectRL repurposes failed expert reasoning traces as reflective scaffolding during RL and distillation training, then transitions the policy to direct reasoning, improving math and science benchmark scores.
Process reinforcement through implicit rewards.Transactions on Machine Learning Research,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
baseline 1
citation-polarity summary
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1roles
baseline 1polarities
baseline 1representative citing papers
citing papers explorer
-
ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning
ReflectRL repurposes failed expert reasoning traces as reflective scaffolding during RL and distillation training, then transitions the policy to direct reasoning, improving math and science benchmark scores.