PulseAugur
EN
LIVE 06:10:16

New Semi-CoT Framework Enhances LLM Reasoning with Pseudo-Supervision

By PulseAugur Editorial · [3 sources] ·

Researchers have introduced Semi-CoT, a novel framework for Semi-supervised Chain-of-Thought Learning that leverages unlabeled questions to generate pseudo reasoning supervision. This method refines the self-training approach for CoT by selecting reliable reasoning chains based on estimated answer-level semantic entropy. While experiments show promise in selecting high-precision pseudo-CoTs, effective utilization still requires improved demonstration selection or student training strategies. AI

IMPACT This research could lead to more efficient training of LLMs by utilizing unlabeled data for improved reasoning capabilities.

RANK_REASON The cluster contains two academic papers discussing novel methods and datasets for Chain-of-Thought reasoning in LLMs and MLLMs.

Read on Hugging Face Daily Papers →

AI-generated summary · Google Gemini · from 3 sources. How we write summaries →

New Semi-CoT Framework Enhances LLM Reasoning with Pseudo-Supervision

COVERAGE [3]

  1. arXiv cs.AI TIER_1 English(EN) · Hongyang He, Jiuming Liu, Victor Sanchez ·

    Revisiting Chain-of-Thought Reasoning under Limited Supervision: Semi-supervised Chain-of-Thought Learning

    arXiv:2607.01511v1 Announce Type: new Abstract: Chain-of-thought (CoT) reasoning has emerged as an effective approach for activating latent reasoning capabilities in large language models. However, most existing CoT methods use reasoning chains mainly as inference-time prompts, w…

  2. arXiv cs.LG TIER_1 English(EN) · Lingxiao Li, Yifan Wang, Xinyan Gao, Chen Tang, Xiangyu Yue, Chenyu You ·

    VisReason: A Large-Scale Dataset for Visual Chain-of-Thought Reasoning

    arXiv:2511.17731v2 Announce Type: replace-cross Abstract: Chain-of-Thought (CoT) prompting has proven remarkably effective for eliciting complex reasoning in large language models (LLMs). Yet, its potential in multimodal large language models (MLLMs) remains largely untapped, hin…

  3. Hugging Face Daily Papers TIER_1 English(EN) ·

    Revisiting Chain-of-Thought Reasoning under Limited Supervision: Semi-supervised Chain-of-Thought Learning

    Chain-of-thought (CoT) reasoning has emerged as an effective approach for activating latent reasoning capabilities in large language models. However, most existing CoT methods use reasoning chains mainly as inference-time prompts, while the generated reasoning traces are rarely r…

RELATED ENTITIES

RELATED TOPICS