AdapThink: Adaptive Thinking Preferences for Reasoning Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent research has highlighted a significant inefficiency associated with the slow thinking paradigm . models often overthink simple tasks while underthinking complex challenges . |
| Approach: | They propose a framework for adaptive reasoning preference control that dynamically adjusts reflection preferences based on group-level distributional statistics of reasoning length and reflection intensity. |
| Outcome: | The proposed framework reduces average response length by 17.1%-21.4% while improving performance by 6.12-6.59 points under 32K token budgets. |
Similar Papers
AdaptThink: Reasoning Models Can Learn When to Think (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in large reasoning models have demonstrated remarkable capabilities in tackling complex tasks. |
| Approach: | They propose an algorithm to teach reasoning models to choose the optimal thinking mode based on problem difficulty. |
| Outcome: | The proposed algorithm reduces the average response length and improves accuracy on three math datasets. |
ADaPT: Token-Level Decoupling for Efficient Large Reasoning Models (2026.findings-acl)
Copied to clipboard
Tingyun li, Zishang Jiang, Jinyi Han, Xinyi Wang, Sihang Jiang, Han Xia, Zhaoqian Dai, Ma Shuguang, Fei Yu, Jiaqing Liang, Yanghua Xiao
| Challenge: | Existing efficiency-oriented methods attempt to shorten or mix reasoning strategies, yet often degrade reasoning capability. |
| Approach: | They propose a token-level dual-process framework that explicitly decouples efficiency and correctness signals during training. |
| Outcome: | The proposed framework reduces inference cost while maintaining strong reasoning ability across multiple benchmarks. |
DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models (2025.emnlp-industry)
Copied to clipboard
Yi Shen, Jian Zhang, Jieyun Huang, Shuming Shi, Wenjing Zhang, Jiangze Yan, Ning Wang, Kai Wang, Zhaoxiang Liu, Shiguo Lian
| Challenge: | Recent advances in slow-thinking reasoning models have shown exceptional performance in complex reasoning tasks. |
| Approach: | They propose a framework that enables models to automatically adjust Chain-of-Thought (CoT) length based on problem difficulty. |
| Outcome: | The proposed framework penalizes inefficiency on simple problems while incentivizing deep reasoning for complex ones. |
SAT: Balancing Reasoning Accuracy and Efficiency with Stepwise Adaptive Thinking (2026.acl-long)
Copied to clipboard
| Challenge: | Large Reasoning Models (LRMs) produce excessively long Chains of Thought (COT) Existing solutions that improve token efficiency but sacrifice fine-grained control can disrupt the logical integrity of the reasoning process. |
| Approach: | They propose a framework that performs step-level, difficulty-aware pruning while preserving the core reasoning structure. |
| Outcome: | Experiments show that SAT reduces reasoning tokens by 40% while maintaining or improving accuracy. |
Adaption-of-Thought: Learning Question Difficulty Improves Large Language Models for Reasoning (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods do not differentiate question difficulty when designing prompting methods for them. |
| Approach: | They propose an adaptive method to improve large language models for reasoning problems by measuring question difficulty and tailoring demonstration set construction and difficulty-adapted retrieval strategies. |
| Outcome: | The proposed method shows an absolute improvement of up to 5.5% on arithmetic reasoning, 7.4% on symbolic reasoning, and 2.3% on commonsense reasoning. |
Beyond Templates: Dynamic Adaptation of Reasoning Demonstrations via Feasibility-Aware Exploration (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing reasoning datasets that are designed for powerful LLMs often lead to degraded performance when directly applied to weaker models. |
| Approach: | They propose a data adaptation framework that bridges the capability gap between expert reasoning trajectories and diverse SLMs by employing a selective imitation strategy guided by step-wise adaptability estimation via solution simulation. |
| Outcome: | The proposed framework improves generalization and data efficiency over static fine-tuning and can be applied to large models with limited model capacity. |
Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) increase test-time computation, often in the form of chain-of-thought (CoT) however, reasoning traces can become unnecessarily long, increasing computation costs without improving accuracy and sometimes even degrading performance. |
| Approach: | They propose a multi-stage efficient reasoning method that combines supervised fine-tuning with reinforcement learning using an adaptive length penalty. |
| Outcome: | The proposed method reduces response length by an average of 28% for 8B models and 40% for 32B models while incurring only minor performance drops of 1.6 and 2.5 points, respectively. |
Diffuse Thinking: Exploring Diffusion Language Models as Efficient Thought Proposers for Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) have demonstrated strong capabilities in complex reasoning tasks, yet their autoregressive generation paradigm makes it computationally prohibitive to explore diverse reasoning paths. |
| Approach: | They propose a framework that combines diffusion-based generation with autoregressive evaluation to efficiently generate diverse intermediate reasoning thoughts and employ LLMs as evaluators to assess and select candidates based on their plausibility and correctness. |
| Outcome: | The proposed framework improves inference efficiency while maintaining competitive or superior reasoning accuracy. |
CAT: Confidence-Adaptive Thinking for Efficient Reasoning of Large Reasoning Models (2026.acl-industry)
Copied to clipboard
| Challenge: | Existing compression methods for large reasoning models rely on uniform length reduction or coarse-grained difficulty estimation, often leading to performance degradation on difficult problems. |
| Approach: | They propose a framework that incorporates model’s intrinsic self-certainty signals as confidence into the preference optimization process, which autonomously modulates reasoning lengths based on problem difficulty. |
| Outcome: | The proposed framework outperforms state-of-the-art models on reasoning accuracy across multiple benchmarks on different base models. |
Meta-Reasoner: Dynamic Guidance for Optimized Inference-time Reasoning in Large Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent advances on prompting and post-training have enabled LLMs to perform step-wise reasoning tasks, but they tend to explore unproductive solution paths without effective backtracking or strategy adjustment. |
| Approach: | They propose a framework that empowers LLMs to “think about how to think” and dynamically adapts reasoning strategies in real-time. |
| Outcome: | The proposed framework outperforms previous SOTA methods by 9-12% in accuracy while reducing inference time by 28-35% under the same compute budget. |