홈 › Papers

Robust Trajectory Distillation: Hybrid Reweighting Meets Teacher-Inspired Targets

2026-06-29 · Kaifeng Chen, Lechao Cheng, Jiyang Li, Shengeng Tang, Fan Zhang, Yantao Pan, Yaxiong Wang, Tuanrui Hui, Zhun Zhong arxiv

Dataset distillation (DD) condenses large corpora into compact, information-rich subsets for efficient training and reuse. However, under noisy supervision, DD risks condensing corrupted associations together with useful signals, degrading robustness. Conventional noisy-label remedies (sample selection, loss weighting, label correction) tightly couple noise estimation with model optimization, often require clean anchors, and can amplify confirmation bias-assumptions that are misaligned with DD's goal of compact, plug-and-play supervision. We therefore propose a trajectory-based DD framework that jointly suppresses noise and preserves transferable knowledge without relabeling or clean subsets. It comprises two complementary components: Selective Guidance Reweighting (SGR), which fuses global forgetting patterns (second-split forgetting) with local neighborhood consistency into a progressive reweighting scheme that prioritizes clean supervision along the teacher trajectory; and Teacher-Inspired Auxiliary Targets (TIAT), which inject auxiliary residual guidance distilled from intermediate teacher dynamics to reinforce informative signals while remaining internally consistent. Together, SGR and TIAT produce distilled datasets with cleaner and richer representations under noisy supervision. The framework is robust, label-preserving, computationally lightweight, and broadly applicable, yielding consistent gains over state-of-the-art DD baselines across symmetric, asymmetric, and real-world noise.

📄 PDF Abstract BibTeX arXiv:2606.29837

Code (0)

등록된 구현이 없습니다.

Tasks

Noise Estimation

Trajectory-Refined Distillation

2026-06-07 · Li Jiang, Haoran Xu, Yichuan Ding, Amy Zhang arxiv

On-policy distillation (OPD) has become a central post-training tool for large language models (LLMs), providing dense per-token teacher supervision along the student's own rollouts. In this work, we identify a common st…

GEAR: Granularity-Adaptive Advantage Reweighting for LLM Agents via Self-Distillation

2026-05-12 · Sijia Li, Yuchen Huang, Zifan Liu, Yanping Li 외 arxiv

Reinforcement learning has become a widely used post-training approach for LLM agents, where training commonly relies on outcome-level rewards that provide only coarse supervision. While finer-grained credit assignment i…

Reinforcement Learning Mathematical Reasoning

H^2SD: Hybrid Hindsight Self-Distillation

2026-07-21 · Qiye Cai, Yichuan Ma, Linyang Li, Peiji Li 외 arxiv

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. However, most RLVR methods…

Reinforcement Learning Mathematical Reasoning Code Generation

ATOD: Annealed Turn-Aware On-Policy Distillation for Multi-Turn Agentic Tasks

2026-06-26 · Qitai Tan, Zefang Zong, Yang Li, Peng Chen arxiv

Training small language-model agents for long-horizon interactive tasks requires both fast imitation and reward-driven improvement. On-policy distillation (OPD) provides dense teacher guidance and typically improves rapi…

Reinforcement Learning

Visual-Advantage On-Policy Distillation for Vision-Language Models

2026-05-21 · Ruiqi Liu, Xiaolei Lv, Gengsheng Li, Ximo Zhu 외 arxiv

On-policy knowledge distillation has proven effective for language models, yet its application to vision-language models (VLMs) remains underexplored. We observe that standard on-policy distillation can improve a student…

Knowledge Distillation Mathematical Reasoning