Pith. sign in

REVIEW 16 cited by

AdaptThink: Reasoning Models Can Learn When to Think

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2505.13417 v1 pith:74QHZJ7T submitted 2025-05-19 cs.CL cs.AIcs.LG

classification cs.CLcs.AIcs.LG
keywords adaptthinkthinkingreasoningmodelsperformanceefficiencymodelnothinking
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Recently, large reasoning models have achieved impressive performance on various tasks by employing human-like deep thinking. However, the lengthy thinking process substantially increases inference overhead, making efficiency a critical bottleneck. In this work, we first demonstrate that NoThinking, which prompts the reasoning model to skip thinking and directly generate the final solution, is a better choice for relatively simple tasks in terms of both performance and efficiency. Motivated by this, we propose AdaptThink, a novel RL algorithm to teach reasoning models to choose the optimal thinking mode adaptively based on problem difficulty. Specifically, AdaptThink features two core components: (1) a constrained optimization objective that encourages the model to choose NoThinking while maintaining the overall performance; (2) an importance sampling strategy that balances Thinking and NoThinking samples during on-policy training, thereby enabling cold start and allowing the model to explore and exploit both thinking modes throughout the training process. Our experiments indicate that AdaptThink significantly reduces the inference costs while further enhancing performance. Notably, on three math datasets, AdaptThink reduces the average response length of DeepSeek-R1-Distill-Qwen-1.5B by 53% and improves its accuracy by 2.4%, highlighting the promise of adaptive thinking-mode selection for optimizing the balance between reasoning quality and efficiency. Our codes and models are available at https://github.com/THU-KEG/AdaptThink.

Discussion (0). Sign in to comment.

Forward citations

Cited by 16 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. AdaThinkV: Adaptive Thinking for Token-Efficient Video Reasoning

    cs.CV 2026-08 conditional novelty 7.0 of 10

    A video reasoning model learns per question whether to reason aloud or answer directly, improving accuracy by about 3 points over the best adaptive baseline while using about 23% fewer output tokens.

  2. Length Value Model: Scalable Value Pretraining for Token-Level Length Modeling

    cs.CL 2026-04 unverdicted novelty 6.0 of 10

    LenVM trains a token-level value head to predict discounted remaining length, enabling length control and efficiency steering on LLMs and VLMs.

  3. ConPress: Learning Efficient Reasoning from Multi-Question Contextual Pressure

    cs.CL 2026-02 conditional novelty 6.0 of 10

    Multi-question prompts elicit shorter chain-of-thought traces, and fine-tuning on those traces transfers the compression to single-question reasoning.

  4. R-4B: Incentivizing General-Purpose Auto-Thinking Capability in MLLMs via Bi-Mode Annealing and Reinforce Learning

    cs.CV 2025-08 conditional novelty 6.0 of 10

    R-4B uses bi-mode annealing plus a GRPO variant with forced thinking and non-thinking rollouts so a 4B multimodal model adaptively picks when to reason, reportedly matching 16B thinking models on several reasoning benchmarks.

  5. LAPO: Internalizing Reasoning Efficiency via Length-Adaptive Policy Optimization

    cs.AI 2025-07 conditional novelty 6.0 of 10

    LAPO is a two-stage RL method that reduces reasoning token usage by up to 40.9% while slightly improving accuracy, by learning per-problem length targets from the model's own correct answers and prompting the model to...

  6. Squeeze the Soaked Sponge: Efficient Off-policy Reinforcement Finetuning for Large Language Model

    cs.LG 2025-07 conditional novelty 6.0 of 10

    ReMix cuts reinforcement finetuning rollout volume by 30x to 450x on math reasoning by mixing historical and on-policy data with a convex KL constraint and a mid-training switch to on-policy updates.

  7. SmartThinker: Learning to Compress and Preserve Reasoning by Step-Level Length Control

    cs.AI 2025-07 conditional novelty 6.0 of 10

    SmartThinker shortens AI reasoning chains by estimating the importance of each step and penalizing the unimportant ones, preserving accuracy while using far fewer tokens.

  8. Do Thinking Tokens Help or Trap? Towards More Efficient Large Reasoning Model

    cs.CL 2025-06 conditional novelty 6.0 of 10

    Suppressing 'thinking tokens' in a 1.5B reasoning model preserves accuracy while cutting tokens, and the proposed DuP-PO RL method improves both accuracy and efficiency over GRPO.

  9. How Far Are We from Optimal Reasoning Efficiency?

    cs.CL 2025-06 conditional novelty 6.0 of 10

    The authors define a reasoning efficiency frontier and a gap metric (REG), then train models with REO-RL to shrink the gap by at least 50% with only small accuracy losses.

  10. Is Your Model Thinking or Just Stagnating? PUMA: Diagnosing Reasoning Pathology via Phase-Momentum Alignment

    cs.AI 2026-07 conditional novelty 5.0 of 10

    PUMA detects when a reasoning model's entropy drop aligns with hidden-state momentum, truncates at that point, and reports improved accuracy-efficiency on 1.5B-32B reasoning models.

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

    cs.CL 2025-08 conditional novelty 5.0 of 10

    An open training recipe that teaches a 32B LLM to reason in three adjustable modes, cutting token use by half or three quarters with small accuracy losses.

  12. Think in Blocks: Adaptive Reasoning from Direct Response to Deep Reasoning

    cs.AI 2025-08 unverdicted novelty 5.0 of 10

    A training framework that makes LLMs predict an integer reasoning budget as a number of blocks and adapt reasoning depth to task difficulty.

  13. KAT-V1: Kwai-AutoThink Technical Report

    cs.CL 2025-07 conditional novelty 5.0 of 10

    KAT-V1-40B is a 40B language model that switches between deep reasoning and direct answering per query, reporting matching-or-better benchmark scores with lower token use.

  14. Reasoning as a Resource: Optimizing Fast and Slow Thinking in Code Generation Models

    cs.SE 2025-06 conditional novelty 4.0 of 10

    Reasoning depth in code LLMs should be managed as a controllable resource across synthetic data generation, benchmarking, and deployment, rather than left implicit.

  15. Reinforcement Learning Meets Large Language Models: A Survey of Advancements and Applications Across the LLM Lifecycle

    cs.CL 2025-09 conditional novelty 3.0 of 10

    A survey that maps reinforcement learning methods, datasets, benchmarks, and open-source tools across the full training lifecycle of large language models, focusing on verifiable-reward reasoning.

  16. Towards Concise and Adaptive Thinking in Large Reasoning Models: A Survey

    cs.AI 2025-07 conditional novelty 3.0 of 10

    A comprehensive review that categorizes methods for shortening and adaptively triggering chain-of-thought reasoning in large language models.

Pith tools