The paper introduces Self‑Steered Reasoner (SSR), a method that aligns self‑distillation with the reasoning actions a large language model can actually reach. SSR projects privileged teacher trajectories onto locally reachable actions, filters for correct paths, and iteratively refines sampling using diagnostic guidance. Experiments show that SSR recovers learnable trajectories on hard problems and boosts accuracy by 4.9 percentage points on benchmark datasets.
arXiv:2606. 10385v1 Announce Type: cross Abstract: On-policy distillation (OPD) has demonstrated strong empirical gains in enhancing complex reasoning in LLMs by aligning a student model with a teacher's predictive distribution over the student's own trajectories.
arXiv:2605. 12400v2 Announce Type: replace-cross Abstract: We study on-policy self-distillation (OPSD), where a language model improves its reasoning ability by distilling privileged teacher distributions along its own on-policy trajectories.
CataOPD introduces a new framework for improving large language model reasoning by combining reinforcement learning and on‑policy distillation. The method treats the teacher as a catalyst that expands the student’s reachability, using Self‑Rescue Routing to find correct trajectories through additional on‑policy sampling and Catalytic‑Guided Self‑Resolution to elicit verified student trajectories. Barrier‑Weighted Internalization further focuses updates on decisive tokens, leading to better performance on unseen problems and improved out‑of‑distribution generalization.
The paper introduces a recursive self-improvement framework for language models that replaces an external teacher with a frozen copy of the student, enabling dynamic co-evolution (DCE) and self-refined concise learning (SRCL). DCE allows the privileged teacher to evolve alongside the student, while SRCL trains on shorter, verified rewrites to reduce verbosity. Experiments show that the combined DCE+SRCL approach outperforms traditional on‑policy self‑distillation across multiple model sizes and math benchmarks, achieving significant accuracy gains and shorter outputs.
The paper introduces temporal self‑distillation for reinforcement learning with verifiable rewards (RLVR), proposing that a policy can learn from a stronger future checkpoint of itself. Two methods—Near‑Future Policy Optimization (NPO) and Near‑Future Policy Distillation (NPD)—use verified future‑self trajectories and token‑level transfer, respectively, while AutoNPO adaptively selects the optimal future checkpoint. Experiments on eight image‑text benchmarks show that near‑future teachers yield higher performance than far‑future ones, indicating that the balance between new capability and learner compatibility is key.
arXiv:2607. 02234v1 Announce Type: new Abstract: On-policy self-distillation (OPSD) has emerged as a promising paradigm for improving LLM reasoning, where a privileged teacher with access to reference solutions provides token-level supervision on the student's own generated trajectories.
ReTeach is a reflective self‑distillation framework that builds a self‑teacher through multiple rounds of reflection and retry, relying only on self‑generated attempts and outcome‑level verification. Starting from an unsuccessful student rollout, the teacher alternates explicit reflection with new attempts until success or a retry limit is reached, without needing reference answers, external feedback, or cross‑example memory. The method uses outcome‑aware weighting to distinguish correct, reflection‑corrected, and unresolved examples, and on‑policy distillation aligns the student’s token‑level predictions with the teacher’s context‑conditioned distributions, achieving a 1.39‑point accuracy gain over GRPO across six reasoning benchmarks.
The paper introduces Cognition-Aligned Reasoning Distillation (CARD), an RL-based method that converts solver trajectories from large reasoning models into teaching trajectories better suited for smaller student models. CARD uses two reward signals—a Distribution Alignment Reward and a Logical Coherence Reward—to align the teacher’s outputs with the student’s cognitive capabilities and reduce misleading reasoning steps. Experiments show that CARD improves both in-domain and out-of-distribution performance compared to distillation from raw teacher trajectories.
Activation-Conditioned Self-Distillation (ACSD) is a new on‑policy self‑distillation method that uses a frozen copy of the base model to extract a steering vector by contrasting activations from self‑generated trajectories that reach verified correct answers with all other trajectories. The student learns from next‑token distributions on its own prefixes, without needing reference text or teacher parameter updates, and is used alone at inference. Across five models, ACSD achieves the highest mean accuracy on four mathematical benchmarks, with notable gains on DeepSeek‑R1‑0528‑Qwen3‑8B and LiveCodeBench v6 compared to the OPSD baseline.
arXiv:2607. 18955v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) has substantially improved the reasoning capabilities of large language models on tasks such as mathematical reasoning and code generation.
Negative Self-Distillation (NSD) is a new framework for improving large language models by encouraging them to diverge from their own flawed reasoning rather than imitate privileged solutions. Unlike On-Policy Self-Distillation, which can suppress uncertainty and exploratory behavior, NSD generates a question‑specific negative condition (e.g., a careless reasoner) and uses a dynamic gating mechanism to target only reasoning‑critical tokens for penalization. This approach preserves foundational language capabilities while consistently outperforming OPSD and other label‑free self‑bootstrapping reinforcement learning baselines.
arXiv:2605.10194v2 Announce Type: replace Abstract: On-policy self-distillation (OPSD) uses a model as its own teacher under privileged context, providing token-level supervision on the model's own r...