Cover Image

The Extrapolation Cliff in On-Policy Distillation of Near-Deterministic Structured Outputs

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Li, Xin, Jiang, Hao, Wang, Annan, Zhang, Yichi, Yuen, Chau
Format:	Preprint
Published:	2026
Subjects:	Machine Learning Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Learning beyond Teacher: Generalized On-Policy Distillation with Reward Extrapolation
by: Yang, Wenkai, et al.
Published: (2026)

Near-Policy: Accelerating On-Policy Distillation via Asynchronous Generation and Selective Packing
by: Rang, Miao, et al.
Published: (2026)

WirelessMathBench: A Mathematical Modeling Benchmark for LLMs in Wireless Communications
by: Li, Xin, et al.
Published: (2025)

The Policy Cliff: A Theoretical Analysis of Reward-Policy Maps in Large Language Models
by: Xu, Xingcheng
Published: (2025)

Trust Region On-Policy Distillation
by: Xing, Xingrun, et al.
Published: (2026)

Deterministic Differentiable Structured Pruning for Large Language Models
by: Huang, Weiyu, et al.
Published: (2026)

DOGe: Defensive Output Generation for LLM Protection Against Knowledge Distillation
by: Li, Pingzhi, et al.
Published: (2025)

ROSD: Reflective On-Policy Self-Distillation for Language Model Reasoning across Domains
by: Zhao, Ziqi, et al.
Published: (2026)

Internalize the Temperature: On-Policy Self-Distillation as Policy Reheater for Reinforcement Learning
by: Yang, Xuewei, et al.
Published: (2026)

Thinking-Free Policy Initialization Makes Distilled Reasoning Models More Effective and Efficient Reasoners
by: Xu, Xin, et al.
Published: (2025)

Large Language Models as Interpolated and Extrapolated Event Predictors
by: Zhang, Libo, et al.
Published: (2024)

Quantifying the Impact of Structured Output Format on Large Language Models through Causal Inference
by: Yuan, Han, et al.
Published: (2025)

Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models
by: Zhao, Siyan, et al.
Published: (2026)

OmniOPD: Logit-Free On-Policy Distillation via Speculative Verification
by: Zhou, Yuhang, et al.
Published: (2026)

Learning Extrapolative Sequence Transformations from Markov Chains
by: Hager, Sophia, et al.
Published: (2025)

A Comment On "The Illusion of Thinking": Reframing the Reasoning Cliff as an Agentic Gap
by: Khan, Sheraz, et al.
Published: (2025)

AlignDistil: Token-Level Language Model Alignment as Adaptive Policy Distillation
by: Zhang, Songming, et al.
Published: (2025)

SODA: Semi On-Policy Black-Box Distillation for Large Language Models
by: Chen, Xiwen, et al.
Published: (2026)

Revisiting On-Policy Distillation: Empirical Failure Modes and Simple Fixes
by: Fu, Yuqian, et al.
Published: (2026)

Structured Agent Distillation for Large Language Model
by: Liu, Jun, et al.
Published: (2025)

A Survey of On-Policy Distillation for Large Language Models
by: Song, Mingyang, et al.
Published: (2026)

Scaling Reasoning Efficiently via Relaxed On-Policy Distillation
by: Ko, Jongwoo, et al.
Published: (2026)

Model Extrapolation Expedites Alignment
by: Zheng, Chujie, et al.
Published: (2024)

SLOT: Structuring the Output of Large Language Models
by: Wang, Darren Yow-Bang, et al.
Published: (2025)

Few-Step Diffusion Language Models via Trajectory Self-Distillation
by: Zhang, Tunyu, et al.
Published: (2026)

Rethinking Attention Output Projection: Structured Hadamard Transforms for Efficient Transformers
by: Aggarwal, Shubham, et al.
Published: (2026)

Real-Time Trustworthiness Scoring for LLM Structured Outputs and Data Extraction
by: Goh, Hui Wen, et al.
Published: (2026)

Steering Safely or Off a Cliff? Rethinking Specificity and Robustness in Inference-Time Interventions
by: Goyal, Navita, et al.
Published: (2026)

LongReD: Mitigating Short-Text Degradation of Long-Context Large Language Models via Restoration Distillation
by: Dong, Zican, et al.
Published: (2025)

Extracting Prompts by Inverting LLM Outputs
by: Zhang, Collin, et al.
Published: (2024)

From Interpolation to Extrapolation: Complete Length Generalization for Arithmetic Transformers
by: Duan, Shaoxiong, et al.
Published: (2023)

Enhancing Length Extrapolation in Sequential Models with Pointer-Augmented Neural Memory
by: Le, Hung, et al.
Published: (2024)

Variational Masked Diffusion Models
by: Zhang, Yichi, et al.
Published: (2025)

CPC-CMS: Cognitive Pairwise Comparison Classification Model Selection Framework for Document-level Sentiment Analysis
by: Li, Jianfei, et al.
Published: (2025)

SlimMoE: Structured Compression of Large MoE Models via Expert Slimming and Distillation
by: Li, Zichong, et al.
Published: (2025)

MAD-OPD: Breaking the Ceiling in On-Policy Distillation via Multi-Agent Debate
by: Wang, Jianze, et al.
Published: (2026)

ORPO-Distill: Mixed-Policy Preference Optimization for Cross-Architecture LLM Distillation
by: Singh, Aasheesh, et al.
Published: (2025)

Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe
by: Li, Yaxuan, et al.
Published: (2026)

BitNet Distillation
by: Wu, Xun, et al.
Published: (2025)

RLPR: Extrapolating RLVR to General Domains without Verifiers
by: Yu, Tianyu, et al.
Published: (2025)