paper-with-me

홈 › Papers

Rethinking External Slow-Thinking: From Snowball Errors to Probability of Correct Reasoning

2025-01-26 · Zeyu Gan, Yun Liao, Yong liu

Test-time scaling, which is also often referred to as slow-thinking, has been demonstrated to enhance multi-step reasoning in large language models (LLMs). However, despite its widespread utilization, the mechanisms underlying slow-thinking methods remain poorly understood. This paper explores the mechanisms of external slow-thinking from a theoretical standpoint. We begin by examining the snowball error effect within the LLM reasoning process and connect it to the likelihood of correct reasoning using information theory. Building on this, we show that external slow-thinking methods can be interpreted as strategies to mitigate the error probability. We further provide a comparative analysis of popular external slow-thinking approaches, ranging from simple to complex, highlighting their differences and interrelationships. Our findings suggest that the efficacy of these methods is not primarily determined by the specific framework employed, and that expanding the search scope or the model's internal reasoning capacity may yield more sustained improvements in the long term. We open-source our code at https://github.com/ZyGan1999/Snowball-Errors-and-Probability.

📄 PDF Abstract BibTeX arXiv:2501.15602

Code (1)

zygan1999/snowball-errors-and-probability 공식 구현 pytorch

Similar Papers 제목 키워드 기반

VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning

2025-04-10 · Haozhe Wang, Chao Qu, Zuming Huang, Wei Chu 외

Recently, slow-thinking systems like GPT-o1 and DeepSeek-R1 have demonstrated great potential in solving challenging problems through explicit reflection. They significantly outperform the best fast-thinking models, such…

MathMultimodal Reasoning

From Hallucination to Structure Snowballing: The Alignment Tax of Constrained Decoding in LLM Reflection

2026-04-07 · Hongxu Zhou arxiv

Intrinsic self-correction in Large Language Models (LLMs) frequently fails in open-ended reasoning tasks due to ``hallucination snowballing,'' a phenomenon in which models recursively justify early errors during free-tex…

Retrieving, Rethinking and Revising: The Chain-of-Verification Can Improve Retrieval Augmented Generation

2024-10-08 · Bolei He, Nuo Chen, Xinran He, Lingyong Yan 외

Recent Retrieval Augmented Generation (RAG) aims to enhance Large Language Models (LLMs) by incorporating extensive knowledge retrieved from external sources. However, such approach encounters some challenges: Firstly, t…

Language ModelingLanguage ModellingRAGRetrieval+1

Rethinking with Retrieval: Faithful Large Language Model Inference

2022-12-31 · Hangfeng He, Hongming Zhang, Dan Roth

Despite the success of large language models (LLMs) in various natural language processing (NLP) tasks, the stored knowledge in these models may inevitably be incomplete, out-of-date, or incorrect. This motivates the nee…

Language ModelingLanguage ModellingLarge Language Modelmodel+2

MM-Snowball: Evaluating and Mitigating Hallucination Snowballing in Multimodal Multi-Turn Dialogue

2026-05-30 · Yue Jiang, Xue Jiang, Lihua Zhang, Zhiqiang Wang 외 arxiv

Multimodal large language models (MLLMs) demonstrate remarkable visual understanding, yet their reliability in interactive settings is severely undermined by hallucination snowballing: a phenomenon where initial errors a…

Visual Grounding