Speculative Thinking: Enhancing Small-Model Reasoning with Large Model Guidance at Inference Time

• URL: http://arxiv.org/abs/2504.12329v1
• Date: Sat, 12 Apr 2025 21:25:32 GMT
• Title: Speculative Thinking: Enhancing Small-Model Reasoning with Large Model Guidance at Inference Time
• Authors: Wang Yang, Xiang Yue, Vipin Chaudhary, Xiaotian Han,
• Abstract summary: We introduce Speculative Thinking, a training-free framework that enables large reasoning models to guide smaller ones during inference at the reasoning level.<n>By strategically delegating reflective steps to a more capable model, our method significantly boosts the reasoning accuracy of reasoning models while shortening their output.
• Score: 14.963731580129892
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recent advances leverage post-training to enhance model reasoning performance, which typically requires costly training pipelines and still suffers from inefficient, overly lengthy outputs. We introduce Speculative Thinking, a training-free framework that enables large reasoning models to guide smaller ones during inference at the reasoning level, distinct from speculative decoding, which operates at the token level. Our approach is based on two observations: (1) reasoning-supportive tokens such as "wait" frequently appear after structural delimiters like "\n\n", serving as signals for reflection or continuation; and (2) larger models exhibit stronger control over reflective behavior, reducing unnecessary backtracking while improving reasoning quality. By strategically delegating reflective steps to a more capable model, our method significantly boosts the reasoning accuracy of reasoning models while shortening their output. With the assistance of the 32B reasoning model, the 1.5B model's accuracy on MATH500 increases from 83.2% to 89.4%, marking a substantial improvement of 6.2%. Simultaneously, the average output length is reduced from 5439 tokens to 4583 tokens, representing a 15.7% decrease. Moreover, when applied to a non-reasoning model (Qwen-2.5-7B-Instruct), our framework boosts its accuracy from 74.0% to 81.8% on the same benchmark, achieving a relative improvement of 7.8%.

Related papers

• Stepwise Penalization for Length-Efficient Chain-of-Thought Reasoning [66.22060690012512]
  Large reasoning models improve with more test-time computation, but often overthink, producing unnecessarily long chains-of-thought that raise cost without improving accuracy.<n>We propose Step-wise Adaptive Penalization (SWAP), a fine-grained framework that allocates length reduction across steps based on intrinsic contribution.
  arXiv  Detail & Related papers  (2026-02-27T20:23:59Z)
• Stop Unnecessary Reflection: Training LRMs for Efficient Reasoning with Adaptive Reflection and Length Coordinated Penalty [42.57318973226598]
  ARLCP is a reinforcement learning framework designed to balance reasoning efficiency and solution accuracy.<n>We evaluate our method on five mathematical reasoning benchmarks using DeepSeek-R1-Distill-Qwen-1.5B and DeepSeek-R1-Distill-Qwen-7B models.
  arXiv  Detail & Related papers  (2026-02-12T16:04:00Z)
• Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning [11.179446105672461]
  We propose a multi-stage efficient reasoning method that combines supervised fine-tuning and reinforcement learning.<n>Our approach reduces response length by an average of 28% for 8B models and 40% for 32B models.<n>It achieves a superior trade-off compared to more complex state-of-the-art efficient reasoning methods.
  arXiv  Detail & Related papers  (2026-01-06T12:31:51Z)
• Logit Arithmetic Elicits Long Reasoning Capabilities Without Training [21.054461373109522]
  We show that ThinkLogit can tune a target large non-reasoning model for long reasoning using a substantially smaller reasoning model as the guider.<n>Experiments demonstrate that ThinkLogit and ThinkLogit-DPO achieve a relative improvement in average accuracy by 24.5% and 29.1%, respectively.
  arXiv  Detail & Related papers  (2025-10-10T13:07:14Z)
• Your Models Have Thought Enough: Training Large Reasoning Models to Stop Overthinking [50.97239453902612]
  Large Reasoning Models (LRMs) have achieved impressive performance on challenging tasks, yet their deep reasoning often incurs substantial computational costs.<n>Inspired by Evidence Accumulation Models, we find that LRMs have accumulated sufficient information early in reasoning, making further reasoning steps redundant.<n>We propose Just-Enough Thinking (JET), which trains models to proactively terminate unnecessary reasoning.
  arXiv  Detail & Related papers  (2025-09-27T16:25:06Z)
• Logit Arithmetic Elicits Long Reasoning Capabilities Without Training [14.015546463427732]
  Large reasoning models (LRMs) can do complex reasoning via long chain-of-thought (CoT) involving cognitive strategies such as backtracking and self-correction.<n>Recent studies suggest that some models inherently possess these long reasoning abilities, which may be unlocked via extra training.<n>We propose a decoding-time approach, ThinkLogit, to tune a target large LM for long reasoning using a substantially smaller model as guider.
  arXiv  Detail & Related papers  (2025-07-17T03:31:36Z)
• ConciseHint: Boosting Efficient Reasoning via Continuous Concise Hints during Generation [53.149817480019834]
  Recent advancements in large reasoning models (LRMs) have achieved notable performance enhancements on complex reasoning tasks by scaling up the generation length by Chain-of-Thought (CoT)<n>We propose a framework dubbed ConciseHint, which continuously encourages the reasoning model to speak concisely by injecting the textual hint during the token generation of the reasoning process.<n>Experiments on the state-of-the-art LRMs, including DeepSeek-R1 and Qwen-3 series, demonstrate that our method can effectively produce concise reasoning processes while maintaining performance well.
  arXiv  Detail & Related papers  (2025-06-23T16:20:44Z)
• Reasoning Models Are More Easily Gaslighted Than You Think [85.84943447589511]
  We evaluate three state-of-the-art reasoning models, including OpenAI's o4-mini, Claude-3.7-Sonnet and Gemini-2.5-Flash.<n>Our evaluation reveals significant accuracy drops following gaslighting negation prompts.<n>We introduce GaslightingBench-R, a new diagnostic benchmark designed to evaluate reasoning models' susceptibility to defend their belief.
  arXiv  Detail & Related papers  (2025-06-11T12:52:25Z)
• Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lens [51.90059610606049]
  This paper revisits the efficiency of such reasoning processes through an information-theoretic lens.<n>We propose two metrics, InfoBias and InfoGain, to quantify divergence from ideal reasoning paths and stepwise information contribution.<n>Motivated by these findings, we introduce an entropy-based Adaptive Think strategy that dynamically halts reasoning once confidence is sufficiently high.
  arXiv  Detail & Related papers  (2025-05-23T13:38:56Z)
• Scaling Reasoning can Improve Factuality in Large Language Models [7.184302333801519]
  We thoroughly examine large language model (LLM) reasoning within complex open-domain question-answering (QA) scenarios.<n>To enrich reasoning traces, we introduce factual information from knowledge graphs in the form of paths into our reasoning traces.<n>Our findings indicate that, within a single run, smaller reasoning models achieve noticeable improvements in factual accuracy compared to their original instruction-tuned counterparts.
  arXiv  Detail & Related papers  (2025-05-16T11:39:33Z)
• Phi-4-reasoning Technical Report [42.508165017775]
  We introduce Phi-4-reasoning, a 14-billion parameter reasoning model that achieves strong performance on complex reasoning tasks. We develop Phi-4-reasoning-plus, a variant enhanced through a short phase of outcome-based reinforcement learning. Both models outperform significantly larger open-weight models such as DeepSeek-R1-Distill-Llama-70B model and approach the performance levels of full DeepSeek-R1 model.
  arXiv  Detail & Related papers  (2025-04-30T05:05:09Z)
• Quantization Hurts Reasoning? An Empirical Study on Quantized Reasoning Models [48.98109982725689]
  We conduct the first systematic study on quantized reasoning models, evaluating the open-sourced DeepSeek-R1-Distilled Qwen and LLaMA families.<n>Our investigation covers weight, KV cache, and activation quantization using state-of-the-art algorithms at varying bit-widths.<n>We identify model size, model origin, and task difficulty as critical determinants of performance.
  arXiv  Detail & Related papers  (2025-04-07T08:22:45Z)
• SEAL: Steerable Reasoning Calibration of Large Language Models for Free [58.190800043449336]
  Large Language Models (LLMs) have demonstrated compelling capabilities for complex reasoning tasks via the extended chain-of-thought (CoT) reasoning mechanism.<n>Recent studies reveal substantial redundancy in the CoT reasoning traces, which negatively impacts model performance.<n>We introduce SEAL, a training-free approach that seamlessly calibrates the CoT process, improving accuracy while demonstrating significant efficiency gains.
  arXiv  Detail & Related papers  (2025-04-07T02:42:07Z)
• Retro-Search: Exploring Untaken Paths for Deeper and Efficient Reasoning [84.2749507577386]
  We introduce Retro-Search, an MCTS-inspired search algorithm, for distilling higher quality reasoning paths from large models.<n>Retro-Search retrospectively revises reasoning paths to discover better, yet shorter traces, which can lead to student models with enhanced reasoning capabilities.<n>Our approach can enable two use cases: self-improvement, where models are fine-tuned on their own Retro-Search-ed traces, and weak-to-strong improvement.
  arXiv  Detail & Related papers  (2025-04-06T06:23:27Z)
• Benchmarking Reasoning Robustness in Large Language Models [76.79744000300363]
  We find significant performance degradation on novel or incomplete data.<n>These findings highlight the reliance on recall over rigorous logical inference.<n>This paper introduces a novel benchmark, termed as Math-RoB, that exploits hallucinations triggered by missing information to expose reasoning gaps.
  arXiv  Detail & Related papers  (2025-03-06T15:36:06Z)
• LLMs Can Easily Learn to Reason from Demonstrations Structure, not content, is what matters! [53.84130385074551]
  Large reasoning models (LRMs) tackle complex reasoning problems by following long chain-of-thoughts (Long CoT) We find that a Large Language model (LLM) can effectively learn Long CoT reasoning through data-efficient supervised fine-tuning (SFT) and parameter-efficient low-rank adaptation (LoRA) With just 17k long CoT training samples, the Qwen2.5-32B-Instruct model achieves significant improvements on a wide range of math and coding benchmarks.
  arXiv  Detail & Related papers  (2025-02-11T08:48:48Z)
• Exploring the Limit of Outcome Reward for Learning Mathematical Reasoning [65.2421542320293]
  Reasoning abilities are crucial components of general intelligence.<n>Recent advances by proprietary companies, such as o-series models of OpenAI, have made remarkable progress on reasoning tasks.<n>This paper proposes a new RL framework, termed OREAL, to pursue the performance limit that can be achieved through textbfOutcome textbfREwtextbfArd-based reinforcement textbfLearning for mathematical reasoning tasks.
  arXiv  Detail & Related papers  (2025-02-10T18:57:29Z)
• Entropy Adaptive Decoding: Dynamic Model Switching for Efficient Inference [0.0]
  We present Entropy Adaptive Decoding (EAD), a novel approach for efficient language model inference.<n>EAD switches between different-sized models based on prediction uncertainty.<n>We show remarkable efficiency gains across different model families.
  arXiv  Detail & Related papers  (2025-02-05T22:15:21Z)
• Weak-to-Strong Preference Optimization: Stealing Reward from Weak Aligned Model [28.569089876442682]
  This work is inspired by weak-to-strong generalization, where a strong LM fine-tuned on labels generated by a weaker model can consistently outperform its weak supervisor.<n>We propose Weak-to-Strong Preference Optimization (WSPO), which achieves strong model alignment by learning the distribution differences before and after the alignment of the weak model.
  arXiv  Detail & Related papers  (2024-10-24T11:06:29Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.