Saved in:
Bibliographic Details
Main Authors: Chen, Zizhao, Li, Yuying, Lin, Siting, Wang, Lianxi
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.11019
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918494862311424
author Chen, Zizhao
Li, Yuying
Lin, Siting
Wang, Lianxi
author_facet Chen, Zizhao
Li, Yuying
Lin, Siting
Wang, Lianxi
contents Although large language models rely on chain-of-thought for complex reasoning, the overthinking phenomenon severely degrades inference efficiency. Existing reinforcement learning methods compress reasoning chains by designing elaborate reward functions, which renders high-quality samples extremely sparse in the exploration space and creates a sampling bottleneck for the prior policy. Inspired by cognitive science, we theoretically prove that a posterior distribution guided by reference answers achieves higher expected utility than the prior distribution, thus capable of breaking through the sampling bottleneck of high-quality samples. However, the posterior distribution is unavailable during inference. To this end, we formalize efficient reasoning as a variational inference problem and introduce an efficiency-aware evidence lower bound as the theoretical foundation. Based on this, we propose the VPG-EA framework. It adopts a parameter-shared dual-stream architecture to instantiate both the posterior distribution and the prior policy; after filtering out pseudo-efficient paths via cross-view evaluation, it unidirectionally transfers the posterior's efficient patterns to the prior policy through variational distillation. Experiments on DeepSeek-R1-Distill-Qwen-1.5B and 7B scales demonstrate that VPG-EA improves the comprehensive efficiency metric epsilon cubed by 8.73% and 12.37% over the strongest baselines on each model size, respectively.
format Preprint
id arxiv_https___arxiv_org_abs_2605_11019
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Efficient LLM Reasoning via Variational Posterior Guidance with Efficiency Awareness
Chen, Zizhao
Li, Yuying
Lin, Siting
Wang, Lianxi
Machine Learning
Artificial Intelligence
Although large language models rely on chain-of-thought for complex reasoning, the overthinking phenomenon severely degrades inference efficiency. Existing reinforcement learning methods compress reasoning chains by designing elaborate reward functions, which renders high-quality samples extremely sparse in the exploration space and creates a sampling bottleneck for the prior policy. Inspired by cognitive science, we theoretically prove that a posterior distribution guided by reference answers achieves higher expected utility than the prior distribution, thus capable of breaking through the sampling bottleneck of high-quality samples. However, the posterior distribution is unavailable during inference. To this end, we formalize efficient reasoning as a variational inference problem and introduce an efficiency-aware evidence lower bound as the theoretical foundation. Based on this, we propose the VPG-EA framework. It adopts a parameter-shared dual-stream architecture to instantiate both the posterior distribution and the prior policy; after filtering out pseudo-efficient paths via cross-view evaluation, it unidirectionally transfers the posterior's efficient patterns to the prior policy through variational distillation. Experiments on DeepSeek-R1-Distill-Qwen-1.5B and 7B scales demonstrate that VPG-EA improves the comprehensive efficiency metric epsilon cubed by 8.73% and 12.37% over the strongest baselines on each model size, respectively.
title Efficient LLM Reasoning via Variational Posterior Guidance with Efficiency Awareness
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2605.11019