What Shared Prefixes Hide: Trajectory Dropout for On-Policy Distillation

arXiv:2609.33455 · cs.AI · Submitted 2026-09-27 · Read on arXiv

cs.AI

Submitted: 2026-09-27

Updated: 2026-09-27

Terminology

• Distillation — 35× in this paper · explained in 2 episodes
• Benchmark — 32× in this paper · explained in 2 episodes
• On-Policy Distillation — 22× in this paper · explained in 3 episodes
• Trajectory Dropout — 14× in this paper
• Ablation — 8× in this paper · explained in 1 episode
• Prefix-Induced Supervision Attenuation — 7× in this paper
• Accuracy — 5× in this paper · explained in 1 episode
• Alignment — 4× in this paper · explained in 1 episode
• Self-distillation — 4× in this paper · explained in 6 episodes
• Consistency — 3× in this paper · explained in 3 episodes
• Thinking mode — 3× in this paper · explained in 2 episodes
• Core — 2× in this paper · explained in 2 episodes
• GRPO — 2× in this paper · explained in 6 episodes
• Generalization — 2× in this paper · explained in 13 episodes
• Actor — 1× in this paper · explained in 3 episodes
• Adam — 1× in this paper · explained in 2 episodes
• Computational overhead — 1× in this paper · explained in 2 episodes
• Data Augmentation — 1× in this paper · explained in 7 episodes
• Fidelity — 1× in this paper · explained in 2 episodes
• Generative Model — 1× in this paper · explained in 2 episodes
• Generative Models — 1× in this paper · explained in 2 episodes
• Knowledge Distillation — 1× in this paper · explained in 11 episodes
• Regularization — 1× in this paper · explained in 3 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes
• Robustness — 1× in this paper · explained in 5 episodes
• Sparsity — 1× in this paper · explained in 4 episodes
• Teacher-Student Gap — 1× in this paper · explained in 2 episodes
• Trace — 1× in this paper · explained in 2 episodes

Sources

• On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes
• Demystifying Training-Time Augmentation for Data-Constrained Language Model Pretraining
• Trajectory-Refined Distillation
• Bridging Reasoning Trajectories in On-Policy Distillation via Near-Future Guidance
• Filter, Then Reweight: Rethinking Optimization Granularity in On-Policy Distillation
• ReNIO: Reweighting Negative Trajectory Importance for LLM On-Policy Distillation
• Same Evidence, Different Answers: Canonical-Context On-Policy Distillation for Multi-Turn Language Models
• Prefix Teach, Suffix Fade: Local Teachability Collapse in Strong-to-Weak On-Policy Distillation
• Your Teacher Can't Help You Here: Combating Supervision Fidelity Decay in On-Policy Distillation
• GPQA: A Graduate-Level Google-Proof Q&A Benchmark
• Not All Disagreement Is Learnable: Token Teachability in On-Policy Distillation
• Qwen3 Technical Report
• On-Policy Context Distillation for Language Models
• Dense Supervision, Sparse Updates: On the Sparsity and Geometry of On-Policy Distillation
• Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
• VSAL: A Vision Solver with Adaptive Layouts for Graph Property Detection