TCOD: Exploring Temporal Curriculum in On-Policy Distillation for Multi-turn Autonomous Agents

• URL: http://arxiv.org/abs/2604.24005v3
• Date: Wed, 29 Apr 2026 03:05:32 GMT
• Title: TCOD: Exploring Temporal Curriculum in On-Policy Distillation for Multi-turn Autonomous Agents
• Abstract summary: We identify a key limitation of vanilla OPD in multi-turn agent settings, which we term Trajectory-Level KL Instability.<n>We propose TCOD, a framework that controls the trajectory depth exposed to the student and progressively expands it with a curriculum schedule.<n> Experimental results across four student-teacher pairs show that TCOD mitigates KL escalation and enhances KL stability throughout training, improving agent performance by up to 18 points over vanilla OPD.
• Score: 55.27396165691312
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: On-policy distillation (OPD) has shown strong potential for transferring reasoning ability from frontier or domain-specific models to smaller students. While effective on static single-turn tasks, its behavior in multi-turn agent settings remains underexplored. In this work, we identify a key limitation of vanilla OPD in such settings, which we term Trajectory-Level KL Instability. Specifically, we observe that KL divergence increases together with a drop in success rate, and even after convergence, the KL remains high, leading to unstable training. This instability arises from inter-turn error compounding: as errors accumulate, the student is driven beyond the teacher's effective support, rendering the supervision signal unreliable. To address this, we propose TCOD (Temporal Curriculum On-Policy Distillation), a simple yet effective framework that controls the trajectory depth exposed to the student and progressively expands it from short to long with a curriculum schedule. Experimental results across four student-teacher pairs on three multi-turn agent benchmarks (ALFWorld, WebShop, ScienceWorld) show that TCOD mitigates KL escalation and enhances KL stability throughout training, improving agent performance by up to 18 points over vanilla OPD. Further evaluations show that TCOD can even surpass the teacher's performance and generalize to tasks on which the teacher fails. Our code is available at https://github.com/kokolerk/TCOD.

Related papers

• DemoPSD: Disagreement-Modulated Policy Self-Distillation [49.30809363804017]
  On-policy self-distillation has emerged as a practical method for training large language models.<n>DemoPSD steers the student toward a weighted geometric combination of the teacher and student distributions.
  arXiv  Detail & Related papers  (2026-07-02T17:58:29Z)
• AsyncOPD: How Stale Can On-Policy Distillation Be? [23.273855195545394]
  On-policy distillation (OPD) trains a student on its own rollouts guided by teacher feedback.<n>Asynchronous training pipelines can alleviate this bottleneck by decoupling rollout generation from learner updates.<n>We present the first systematic study of staleness in asynchronous OPD.
  arXiv  Detail & Related papers  (2026-06-23T04:50:49Z)
• On-Policy Distillation with Curriculum Turn-level Guidance for Multi-turn Agents [50.78688247330029]
  On-Policy Distillation (OPD) is a natural recipe for transferring such capabilities to smaller students.<n>We propose Guided On-Policy Distillation (Guided-OPD), a simple yet effective algorithm that mixes teacher- and student-generated turns within each rollout.<n>Guided-OPD improves Score by 21.1% and Success Rate by 25.5% over vanilla OPD on average.
  arXiv  Detail & Related papers  (2026-06-14T16:41:45Z)
• Your Teacher Can't Help You Here: Combating Supervision Fidelity Decay in On-Policy Distillation [81.10000755917712]
  On-policy distillation transfers reasoning capabilities by training a student model on its own generated trajectories using token-level feedback from a teacher.<n>As student-generated prefixes lengthen, the teacher's next-token distribution becomes less confident and less discriminative.<n>To mitigate SFD, we introduce textbfLookahead Group Reward (ours).
  arXiv  Detail & Related papers  (2026-05-29T04:39:20Z)
• Prefix Teach, Suffix Fade: Local Teachability Collapse in Strong-to-Weak On-Policy Distillation [49.117085054884676]
  On-policy distillation trains a student model on its own rollouts using dense feedback from a stronger teacher.<n>We operationalize this principle through a trajectory-specific release rule.<n> Experimental results across strong-to-weak distillation tasks indicate that this release rule consistently outperforms standard full-trajectory OPD.
  arXiv  Detail & Related papers  (2026-05-13T15:05:30Z)
• Prune-OPD: Efficient and Reliable On-Policy Distillation for Long-Horizon Reasoning [66.52232008796294]
  Prune-OPD dynamically aligns training budgets with supervision quality.<n>It reduces training time by 37.6%--68.0% while preserving, and often improving, performance on challenging benchmarks.
  arXiv  Detail & Related papers  (2026-05-08T14:38:53Z)
• SOD: Step-wise On-policy Distillation for Small Language Model Agents [32.49707795291693]
  Tool-integrated reasoning is difficult to scale to small language models.<n>Recently, on-policy distillation has gained popularity by supplying dense token-level supervision from a teacher.<n>We propose SOD, a step-wise on-policy distillation framework for small language model agents.
  arXiv  Detail & Related papers  (2026-05-08T13:30:42Z)
• MAD-OPD: Breaking the Ceiling in On-Policy Distillation via Multi-Agent Debate [17.31967380175203]
  We propose MAD-OPD (Multi-Agent Debate-driven On-Policy Distillation), which recasts the distillation teacher as a deliberative collective of teachers that debate over the student's on-policy state.<n>We also introduce On-Policy Agentic Distillation (OPAD), which adds step-level sampling to stabilize training under multi-step error compounding.
  arXiv  Detail & Related papers  (2026-05-02T09:41:37Z)
• Distillation Traps and Guards: A Calibration Knob for LLM Distillability [54.90137955363471]
  We propose a post-hoc calibration method that enables control over a teacher's distillability via reinforcement fine-tuning.<n>Our objective combines task utility, KL anchor, and across-tokenizer calibration reward.<n> Experiments show that students distilled from distillable teachers outperform SFT and KD baselines.
  arXiv  Detail & Related papers  (2026-04-21T01:22:35Z)
• Skill-SD: Skill-Conditioned Self-Distillation for Multi-turn LLM Agents [24.99615788156812]
  We introduce Skill-SD, a framework that turns the agent's own trajectories into dynamic training-only supervision.<n>We derive an importance-weighted reverse-KL loss to provide gradient-correct token-level distillation.<n> Experimental results on agentic benchmarks demonstrate that Skill-SD substantially outperforms the standard RL baseline.
  arXiv  Detail & Related papers  (2026-04-12T14:57:52Z)
• Demystifying OPD: Length Inflation and Stabilization Strategies for Large Language Models [35.70546523782139]
  On-policy distillation (OPD) trains student models under their own induced distribution while leveraging supervision from stronger teachers.<n>As training progresses, on-policy rollouts can undergo abrupt length inflation, causing truncated trajectories to dominate the training data.<n>This truncation collapse coincides with abrupt repetition saturation and induces biased gradient signals, leading to severe training instability and sharp degradation in validation performance.<n>We propose StableOPD, a stabilized OPD framework that combines a reference-based divergence constraint with rollout mixture distillation.
  arXiv  Detail & Related papers  (2026-04-09T17:58:02Z)
• Reinforcement-aware Knowledge Distillation for LLM Reasoning [63.53679456364683]
  Reinforcement learning (RL) post-training has recently driven gains in long chain-of-thought reasoning large language models (LLMs)<n>Most existing knowledge distillation methods are designed for supervised fine-tuning (SFT), relying on fixed teacher traces or teacher-student Kullback-Leibler (KL) divergence-based regularization.<n>We propose RL-aware distillation (RLAD), which performs selective imitation during RL -- guiding the student toward the teacher only when it improves the current policy update.
  arXiv  Detail & Related papers  (2026-02-26T00:20:39Z)
• "The Whole Is Greater Than the Sum of Its Parts": A Compatibility-Aware Multi-Teacher CoT Distillation Framework [16.96094045628127]
  Chain-of-Thought (CoT) reasoning empowers Large Language Models (LLMs) with remarkable capabilities but typically requires prohibitive parameter scales.<n>CoT distillation has emerged as a promising paradigm to transfer reasoning prowess into compact Student Models (SLMs)<n>We introduce COMPACT, a framework that adaptively fuses supervisions from different teachers by dynamically weighting teacher gradients.
  arXiv  Detail & Related papers  (2026-01-20T14:05:19Z)
• Stable On-Policy Distillation through Adaptive Target Reformulation [7.361248172930405]
  Veto is an objective-level reformulation that constructs a geometric bridge in the logit space.<n>Veto consistently outperforms supervised fine-tuning and existing on-policy baselines.
  arXiv  Detail & Related papers  (2026-01-12T02:57:39Z)
• AdaSwitch: Adaptive Switching Generation for Knowledge Distillation [58.647880811071495]
  Small language models (SLMs) are crucial for applications with strict latency and computational constraints.<n>We propose AdaSwitch, a novel approach that combines on-policy and off-policy generation at the token level.<n>AdaSwitch consistently improves accuracy, offering a practical and effective method for distilling SLMs with acceptable additional overhead.
  arXiv  Detail & Related papers  (2025-10-09T06:38:37Z)
• Co-Supervised Learning: Improving Weak-to-Strong Generalization with Hierarchical Mixture of Experts [81.37287967870589]
  We propose to harness a diverse set of specialized teachers, instead of a single generalist one, that collectively supervises the strong student. Our approach resembles the classical hierarchical mixture of experts, with two components tailored for co-supervision. We validate the proposed method through visual recognition tasks on the OpenAI weak-to-strong benchmark and additional multi-domain datasets.
  arXiv  Detail & Related papers  (2024-02-23T18:56:11Z)
• Solving Continual Offline Reinforcement Learning with Decision Transformer [78.59473797783673]
  Continuous offline reinforcement learning (CORL) combines continuous and offline reinforcement learning. Existing methods, employing Actor-Critic structures and experience replay (ER), suffer from distribution shifts, low efficiency, and weak knowledge-sharing. We introduce multi-head DT (MH-DT) and low-rank adaptation DT (LoRA-DT) to mitigate DT's forgetting problem.
  arXiv  Detail & Related papers  (2024-01-16T16:28:32Z)

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.