Research map / Large language models / Reasoning and chain of thought
When Simple Problems Wear Complex Costumes: Improving Efficiency in LRM's Adaptive Reasoning
Junnan Ren, Yan Zhang, Qian Chen, Yunhang Shen, Ke Li, Shengchuan Zhang, Liujuan Cao, Rongrong Ji
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Addresses overthinking in large reasoning models by improving adaptive reasoning, so simple problems get direct answers and complex ones get explicit reasoning (details limited to the abstract).
Problem
Large reasoning models overthink simple tasks, applying complex multi-step reasoning that wastes computation and adds latency; existing adaptive reasoning models that switch between explicit reasoning and direct answers are the focus of this work.
Approach
Improves adaptive reasoning in LRMs so the model decides when to reason explicitly versus answer directly; the specific mechanism is not stated in the available abstract.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| apply AdaRS two-stage training (SFT with redundancy augmentation + GRPO) | DeepSeek-R1-Distill-Qwen-1.5B Think · GSM8K, Math500, AIME 2024 average | accuracy: 62.4 -> 65.6 | Improved |
| apply AdaRS two-stage training | DeepSeek-R1-Distill-Qwen-1.5B Think · GSM8K, Math500, AIME 2024 average | tokens per instance: 7413 -> 3250 | Improved |
| add Cold Start SFT with concise and redundant augmented queries (ablation) | DeepSeek-R1-Distill-Qwen-1.5B · GSM8K | tokens per instance: 1634 -> 1047 | Improved |
| add Cold Start SFT (ablation) | DeepSeek-R1-Distill-Qwen-1.5B · ASDIV | accuracy: 89.3 -> 91.8 | Improved |
| use GRPO with naive reward that ignores reasoning cost (ablation) | Cold Start model · GSM8K | tokens per instance: 1047 -> 1809 | Hurt |
| use mode-aware reward (Enhancing Adaptive Policy) instead of naive reward (ablation) | GRPO stage · GSM8K | tokens per instance: 1809 -> 503 | Improved |
| use mode-aware reward instead of naive reward (ablation) | GRPO stage · ASDIV | tokens per instance: 1766 -> 308 | Improved |
| add mode-aware reward on top of Cold Start (ablation) | Cold Start model · GSM8K | accuracy: 82.7 -> 83.5 | Improved |
| train with redundancy-augmented data and custom reward vs prior adaptive models | adaptive reasoning models · ASDIV-redundancy | Think%: 58.5% (AdaptThink) -> 33.1% (AdaRS) | Improved |
| add redundant phrasing to simple queries | AdaRS · ASDIV | Think%: 4.3% -> 33.1% | Hurt |
| add redundant phrasing to simple queries | AdaptThink · ASDIV | Think%: 21.9% -> 58.5% | Hurt |
| use AdaRS adaptive switching instead of forced Nothink mode | DeepSeek-R1-Distill-Qwen-1.5B · GSM8K, Math500, AIME 2024 average | accuracy: 52.2 -> 65.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Thinkless — Thinkless: LLM Learns When to Think (NeurIPS 2025) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- How Far Are We from Optimal Reasoning Efficiency? (NeurIPS 2025 · similarity 0.89)
- Stop Unnecessary Reflection: Training LRMs for Efficient Reasoning with Adaptive Reflection and Length Coordinated Penalty (ICLR 2026 · similarity 0.89)
- Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning (NeurIPS 2025 · similarity 0.88)
- Think Only When You Need with Large Hybrid-Reasoning Models (NeurIPS 2025 · similarity 0.88)
- Learning When to Think: Shaping Adaptive Reasoning in R1-Style Models via Multi-Stage RL (NeurIPS 2025 · similarity 0.88)
- DyCon: Dynamic Reasoning Control via Evolving Difficulty Modeling (ICML 2026 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
