Efficient Reasoning via Thought-Training and Thought-Free Inference

2025/11/05 by Canhui Wu, Wu, Canhui, Cao, Qiong +6
Computer Science · #Multimodal Machine Learning Applications #Topic Modeling #Explainable Artificial Intelligence (XAI)

paper · pdf · doi:10.48550/arxiv.2511.03408

Abstract

Recent advances in large language models (LLMs) have leveraged explicit Chain-of-Thought (CoT) prompting to improve reasoning accuracy. However, most existing methods primarily focus on compressing verbose reasoning outputs. These Long-to-Short transformations aim to improve efficiency, but require a large amount of short CoT data. In this work, we introduce 3TF (Thought-Training and Thought-Free inference), a framework for efficient reasoning that takes a Short-to-Long perspective. We first train a hybrid model that can operate in both reasoning and non-reasoning modes, and then further train it on CoT-annotated data to internalize structured reasoning, while enforcing concise, thought-free outputs at inference time using the no-reasoning mode. Unlike compression-based approaches, 3TF improves the reasoning quality of non-reasoning outputs, enabling models to perform rich internal reasoning implicitly while keeping external outputs short. Empirically, 3TF-trained models obtain large improvements on reasoning benchmarks under thought-free inference, demonstrating that high quality reasoning can be learned and executed implicitly without explicit step-by-step generation.

Citations

• Not All Correct Answers Are Equal: Why Your Distillation Source Matters
• ShorterBetter: Guiding Reasoning Models to Find Optimal Inference Length for Efficient Reasoning
• Between Underthinking and Overthinking: An Empirical Study of Reasoning Length and correctness in LLMs
• Unlocking Efficient Long-to-Short LLM Reasoning with Model Merging
• Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching
• DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models
• Chain of Draft: Thinking Faster by Writing Less
• CoT-Valve: Length-Compressible Chain-of-Thought Tuning
• The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks
• Training Language Models to Reason Efficiently
• O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning
• C3oT: Generating Shorter Chain-of-Thought without Compromising Effectiveness
• SWIFT:A Scalable lightWeight Infrastructure for Fine-Tuning
• Distilling System 2 into System 1
• Let's Verify Step by Step
• Large Language Models are Zero-Shot Reasoners
• Selection-Inference: Exploiting Large Language Models for Interpretable Logical Reasoning
• Self-Consistency Improves Chain of Thought Reasoning in Language Models
• Training Verifiers to Solve Math Word Problems

Related

• ThinkLess: A Training-Free Inference-Efficient Method for Reducing Reasoning Redundancy
• Efficient Reasoning Models: A Survey
• Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought
• SCOUT: Teaching Pre-trained Language Models to Enhance Reasoning via Flow Chain-of-Thought
• Fractured Chain-of-Thought Reasoning