The Role of Feedback Alignment in Self-Distillation

• URL: http://arxiv.org/abs/2606.11173v1
• Date: Tue, 09 Jun 2026 17:50:09 GMT
• Title: The Role of Feedback Alignment in Self-Distillation
• Abstract summary: Self-distillation trains the model to retain this improvement when the context is not present.<n>We study context design for self-distillation by training a solver on feedback from a frozen critic.
• Score: 0.3580891736370874
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Conditioning a language model on additional context, such as feedback on a previous attempt, typically improves its response. Self-distillation trains the model to retain this improvement when the context is not present. The method works by matching the model's output distribution under two settings: a student that sees only the question, and a self-teacher that also sees the context. What the model learns therefore depends on what context the self-teacher receives, yet the design of this context remains largely unexplored. We study context design for self-distillation by training a solver on feedback from a frozen critic. We compare three conditions: (i) a binary reward (GRPO), (ii) the reference solution, and (iii) a step-by-step critique aligned to the solver's reasoning trace. Step-aligned critique yields the largest gains, outperforming GRPO by 16.11 points and reference-solution-conditioned self-distillation by 5.27 points (Avg@12). Per-token advantage analysis reveals why: step-aligned feedback targets only the tokens where reasoning fails, leaving correct behavior intact. Conditioning on the reference solution, by contrast, pressures the model to change its behavior at every token (even correct steps) because an alternative derivation inevitably differs in phrasing and approach. This suggests that structural alignment between feedback and the solver's reasoning is a key driver of self-distillation effectiveness.

Related papers

• H$^2$SD: Hybrid Hindsight Self-Distillation [61.71131241473028]
  Reinforcement learning with verifiable rewards (RLVR) provides reliable outcome supervision for language model reasoning.<n>Existing self-distillation methods add a privileged teacher but typically assign it a fixed role.<n>We introduce Hybrid Hindsight Self-Distillation ($mathrmH2mathrmSD$), which jointly adapts teacher context and update strategy to trajectory correctness.
  arXiv  Detail & Related papers  (2026-07-21T10:47:27Z)
• Purified OPSD: On-Policy Self-Distillation Without Losing How to Think [52.879387744920415]
  We find that on-policy self-distillation (OPSD) consistently fails on long chain-of-thought (long-CoT) reasoning models.<n>We propose a two-step solution: first, we construct a reference-only teacher to isolate the non-transferable component of the supervision signal.<n>Second, we use pointwise mutual information (PMI) as the mechanism to transform this residual into a well-formed PMI target distribution.
  arXiv  Detail & Related papers  (2026-07-02T14:33:07Z)
• Rethinking Reward Supervision: Rubric-Conditioned Self-Distillation [60.55792673956761]
  We propose textbfRubric-Conditioned Self-Distillation, a framework that incorporates rubrics as structured, fine-grained feedback for on-policy self-distillation.<n>We evaluate on a diverse suite of science reasoning benchmarks and results show that rubric-conditioned self-distillation effectively converts rubric-level criteria into token-level guidance.
  arXiv  Detail & Related papers  (2026-06-17T17:54:04Z)
• ROSD: Reflective On-Policy Self-Distillation for Language Model Reasoning across Domains [40.558397280256685]
  On-policy self-distillation (OPSD) improves the reasoning performance of large language models (LLMs)<n>Existing OPSD methods often yield limited gains on in-domain reasoning and generalize poorly to out-of-domain problems.<n>We propose Reflective On-policy Self-Distillation (ROSD), a framework that turns reference-solution imitation into targeted reasoning correction.
  arXiv  Detail & Related papers  (2026-05-27T06:09:29Z)
• Anti-Self-Distillation for Reasoning RL via Pointwise Mutual Information [22.436966302243565]
  We propose Anti-Self-Distillation, which ascends a divergence between student and teacher rather than descending it.<n>AntiSD reaches the GRPO baseline's accuracy in 2 to 10x fewer training steps and improves final accuracy by up to 11.5 points.
  arXiv  Detail & Related papers  (2026-05-12T06:40:43Z)
• PAINT: Partial-Solution Adaptive Interpolated Training for Self-Distilled Reasoners [9.324642081509756]
  Large language model (LLM) reasoning requires supervision that is both aligned with the model's own test-time states and informative at the token level.<n>Recent privileged on-policy self-distillation explores a middle ground by scoring student rollouts with the same model under verified solution context.<n>We propose PAINT, which masks the verified solution according to rollout-reference overlap and applies a small energy-space on a sparse set of entropy-mismatch token positions.
  arXiv  Detail & Related papers  (2026-04-29T11:56:07Z)
• Pando: Do Interpretability Methods Work When Models Won't Explain Themselves? [53.07826484214082]
  We introduce Pando, a model-organism benchmark that breaks this confound via an explanation axis.<n>Pando predicts held-out model decisions from $10$ labeled query-response pairs.<n>When explanations are faithful, black-box elicitation matches or exceeds all white-box methods.
  arXiv  Detail & Related papers  (2026-04-13T06:42:24Z)
• Learning to Focus: Causal Attention Distillation via Gradient-Guided Token Pruning [62.23671919314693]
  Large language models (LLMs) have demonstrated significant improvements in contextual understanding.<n>However, their ability to attend to truly critical information during long-context reasoning and generation still falls behind the pace.<n>We introduce a two-stage framework called Learning to Focus (LeaF) to mitigate confounding factors.
  arXiv  Detail & Related papers  (2025-06-09T15:16:39Z)
• STRIVE: Structured Reasoning for Self-Improvement in Claim Verification [30.15803409441136]
  We propose STRIVE: Structured Reasoning for Self-Improved Verification.<n>Our method introduces a structured reasoning design with Claim Decomposition, Entity Analysis, and Evidence Grounding Verification.<n>It is then applied to generate reasoning chains for all training examples, selecting only those that are correct and structurally sound for subsequent self-improvement training.
  arXiv  Detail & Related papers  (2025-02-17T16:07:07Z)
• Enhancing LLM Reasoning via Critique Models with Test-Time and Training-Time Supervision [120.40788744292739]
  We propose a two-player paradigm that separates the roles of reasoning and critique models. We first propose AutoMathCritique, an automated and scalable framework for collecting critique data. We demonstrate that the critique models consistently improve the actor's performance on difficult queries at test-time.
  arXiv  Detail & Related papers  (2024-11-25T17:11:54Z)
• Chain-of-Probe: Examining the Necessity and Accuracy of CoT Step-by-Step [81.50681925980135]
  We propose a method to probe changes in the mind during the model's reasoning.<n>By analyzing patterns in mind change, we examine the correctness of the model's reasoning.<n>Our validation reveals that many responses, although correct in their final answer, contain errors in their reasoning process.
  arXiv  Detail & Related papers  (2024-06-23T15:50:22Z)
• REFINER: Reasoning Feedback on Intermediate Representations [47.36251998678097]
  We introduce REFINER, a framework for finetuning language models to generate intermediate inferences. REFINER works by interacting with a critic model that provides automated feedback on the reasoning. Empirical evaluations show significant improvements over baseline LMs of comparable scale.
  arXiv  Detail & Related papers  (2023-04-04T15:57:28Z)
• Why do you think that? Exploring Faithful Sentence-Level Rationales Without Supervision [60.62434362997016]
  We propose a differentiable training-framework to create models which output faithful rationales on a sentence level. Our model solves the task based on each rationale individually and learns to assign high scores to those which solved the task best.
  arXiv  Detail & Related papers  (2020-10-07T12:54:28Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.