paper-with-me

홈 › Papers

ThinkDial: An Open Recipe for Controlling Reasoning Effort in Large Language Models

2025-08-26 · Qianyu He, Siyu Yuan, Xuefeng Li, Mingxuan Wang, Jiangjie Chen arxiv

Large language models (LLMs) with chain-of-thought reasoning have demonstrated remarkable problem-solving capabilities, but controlling their computational effort remains a significant challenge for practical deployment. Recent proprietary systems like OpenAI's gpt-oss series have introduced discrete operational modes for intuitive reasoning control, but the open-source community has largely failed to achieve such capabilities. In this paper, we introduce ThinkDial, the first open-recipe end-to-end framework that successfully implements gpt-oss-style controllable reasoning through discrete operational modes. Our system enables seamless switching between three distinct reasoning regimes: High mode (full reasoning capability), Medium mode (50 percent token reduction with <10 percent performance degradation), and Low mode (75 percent token reduction with <15 percent performance degradation). We achieve this through an end-to-end training paradigm that integrates budget-mode control throughout the entire pipeline: budget-mode supervised fine-tuning that embeds controllable reasoning capabilities directly into the learning process, and two-phase budget-aware reinforcement learning with adaptive reward shaping. Extensive experiments demonstrate that ThinkDial achieves target compression-performance trade-offs with clear response length reductions while maintaining performance thresholds. The framework also exhibits strong generalization capabilities on out-of-distribution tasks.

📄 PDF Abstract BibTeX arXiv:2508.18773

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

Teaching Thinking Models to Reason with Tools: A Full-Pipeline Recipe for Tool-Integrated Reasoning

2026-05-07 · Qianjia Cheng, Yuchen Zhang, Zhilin Wang, Yuxin Zuo 외 arxiv

Tool-integrated reasoning (TIR) offers a direct way to extend thinking models beyond the limits of text-only reasoning. Paradoxically, we observe that tool-enabled evaluation can degrade reasoning performance even when t…

Reinforcement Learning

OpenMMReasoner: Pushing the Frontiers for Multimodal Reasoning with an Open and General Recipe

2025-11-20 · Kaichen Zhang, Keming Wu, Zuhao Yang, Bo Li 외 arxiv

Recent advancements in large reasoning models have fueled growing interest in extending such capabilities to multimodal domains. However, despite notable progress in visual reasoning, the lack of transparent and reproduc…

Reinforcement LearningMultimodal ReasoningVisual Reasoning

VulnLLM-R: Specialized Reasoning LLM with Agent Scaffold for Vulnerability Detection

2025-12-08 · Yuzhou Nie, Hongwei Li, Chengquan Guo, Ruizhe Jiang 외 arxiv

We propose VulnLLM-R, the~\emph{first specialized reasoning LLM} for vulnerability detection. Our key insight is that LLMs can reason about program states and analyze the potential vulnerabilities, rather than simple pat…

Vulnerability Detection

Multi-modal Cooking Workflow Construction for Food Recipes

2020-08-20 · Liangming Pan, Jingjing Chen, Jianlong Wu, Shaoteng Liu 외

Understanding food recipe requires anticipating the implicit causal effects of cooking actions, such that the recipe can be converted into a graph describing the temporal workflow of the recipe. This is a non-trivial tas…

Common Sense ReasoningDecoder

Reasoning Curriculum: Bootstrapping Broad LLM Reasoning from Math

2025-10-30 · Bo Pang, Deqian Kong, Silvio Savarese, Caiming Xiong 외 arxiv

Reinforcement learning (RL) can elicit strong reasoning in large language models (LLMs), yet most open efforts focus on math and code. We propose Reasoning Curriculum, a simple two-stage curriculum that first elicits rea…

Reinforcement Learning