PulseAugur
中
实时 06:35:17
English(EN) Think Deep, Not Just Long: Measuring LLM Reasoning Effort via Deep-Thinking Tokens

新研究探索先进的大语言模型推理技术，提升性能和效率

作者 PulseAugur 编辑部 · [4 个来源] ·

研究人员正在探索新颖的方法来增强大语言模型（LLM）的推理能力，超越传统方法。一个框架ILR整合了动态交互策略和感知校准，以提高LLM的独立解决问题能力，相比基线提高了高达5%。另一种方法，自进化后训练（SePT），表明LLM可以在没有外部奖励的情况下，通过训练自己的采样响应来提高推理能力。此外，R$^2$PO将用于生成训练数据的策略与用于推理的策略分离开来，从而在MATH-500等基准测试中提高了准确性。另一项独立研究引入了“深度思考代币”的概念，认为它是比原始代币数量更能可靠地指示推理质量的指标，并提出了一种名为Think@n的新缩放策略，以降低推理成本。 AI

影响 这些研究进展可能带来更高效、更强大的大语言模型，提高它们在复杂推理任务上的性能，并降低计算成本。

排序理由 该集群包含多篇学术论文，详细介绍了LLM推理方面的新方法和发现。

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →

新研究探索先进的大语言模型推理技术，提升性能和效率

报道来源 [4]

  1. arXiv cs.AI TIER_1 English(EN) · Hehai Lin, Shilei Cao, Sudong Wang, Haotian Wu, Minzhi Li, Linyi Yang, Juepeng Zheng, Chengwei Qin ·

    Interactive Learning for LLM Reasoning

    arXiv:2509.26306v5 Announce Type: replace Abstract: Existing multi-agent learning approaches have developed interactive training environments to explicitly promote collaboration among multiple Large Language Models (LLMs), thereby constructing stronger multi-agent systems (MAS). …

  2. arXiv cs.AI TIER_1 English(EN) · Mengqi Li, Lei Zhao, Anthony Man-Cho So, Ruoyu Sun, Xiao Li ·

    A Model Can Help Itself: Reward-Free Self-Training for LLM Reasoning

    arXiv:2510.18814v4 Announce Type: replace-cross Abstract: Can language models improve their reasoning performance without external rewards, using only their own sampled responses for training? We show that they can. We propose Self-evolving Post-Training (SePT), a simple post-tra…

  3. arXiv cs.AI TIER_1 English(EN) · Jingchu Wang, Bingbing Xu, Yige Yuan, Dan Zhang, Bin Xie, Xiaoqian Sun, Huawei Shen ·

    R$^2$PO: Decoupling Rollout and Inference Policies for LLM Reasoning

    arXiv:2601.11960v3 Announce Type: replace-cross Abstract: Existing reinforcement learning methods for LLM reasoning implicitly assume that the policy generating training trajectories should coincide with the one producing inference responses. We argue that this is a misleading in…

  4. arXiv cs.CL TIER_1 English(EN) · Wei-Lin Chen, Liqian Peng, Tian Tan, Chao Zhao, Blake JianHang Chen, Ziqian Lin, Alec Go, Yu Meng ·

    Think Deep, Not Just Long: Measuring LLM Reasoning Effort via Deep-Thinking Tokens

    arXiv:2602.13517v2 Announce Type: replace Abstract: Large language models (LLMs) have demonstrated impressive reasoning capabilities by scaling test-time compute via long Chain-of-Thought (CoT). However, recent findings suggest that raw token counts are unreliable proxies for rea…

相关实体

相关话题