mlsanityData to 14 Oct 2026

Less is More: Improving LLM Reasoning with Minimal Test-Time Intervention

arXiv 2510.13940 ↗First posted 15 Oct 2025cs.CL cs.AIcited by 5

On the map: LLM reasoning → Test-time compute and efficient thinking

Abstract (arXiv)

Recent progress in large language models (LLMs) has focused on test-time scaling to improve reasoning via increased inference computation, but often at the cost of efficiency. We revisit test-time behavior and uncover a simple yet underexplored phenomenon: reasoning uncertainty is highly localized-only a small subset of high-entropy tokens dominantly affects output correctness. Motivated by this, we propose Minimal Test-Time Intervention (MTI), a training-free framework that enhances reasoning accuracy and stability with minimal overhead. MTI includes: (i) Selective CFG intervention, applying classifier-free guidance only at uncertain positions; and (ii) Lightweight negative-prompt guidance, reusing the main model's KV cache to approximate unconditional decoding efficiently. MTI yields consistent gains across general, coding, and STEM tasks-e.g., +9.28% average improvement on six benchmarks for DeepSeek-R1-7B and +11.25% on AIME2024 using Ling-mini-2.0-while remaining highly efficient.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 36 of its 54 references that are in the map. First posted on arXiv 15 Oct 2025.

Builds on

36 of 54 references.

Closest ancestors · 12

Same direction, newest first

  1. 2025Deep Think with Confidence21 Aug
  2. 2025Efficient Reasoning for Large Reasoning Language Models via Certainty-Guided Reflection Suppression7 Aug
  3. 2025Making Slow Thinking Faster: Compressing LLM Chain-of-Thought via Step Entropy5 Aug
  4. 2025R-Stitch: Dynamic Trajectory Stitching for Efficient Reasoning23 Jul
  5. 2025Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning3 Jun
  6. 2025Parallel Scaling Law for Language Models15 May
  7. 2025Dynamic Early Exit in Reasoning Models22 Apr
  8. 2025Reasoning Models Can Be Effective Without Thinking14 Apr
  9. 2025Entropy-Gated Branching for Efficient Test-Time Reasoning27 Mar
  10. 2025How Well do LLMs Compress Their Own Chain-of-Thought? A Token Complexity Approach3 Mar
  11. 2025TokenSkip: Controllable Chain-of-Thought Compression in LLMs17 Feb
  12. 2024Efficiently Scaling LLM Reasoning with Certaindex30 Dec

Other references · 24

  1. 2023GPT-4 Technical Reportcited by 18,396
  2. 2024The Llama 3 Herd of Modelscited by 14,715
  3. 2025Qwen3 Technical Reportcited by 8,820
  4. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  5. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  6. 2022Classifier-Free Diffusion Guidancecited by 5,847
  7. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  8. 2024Gemini 1.5: Unlocking multimodal understanding across millions of tokens of contextcited by 3,224
  9. 2019The Curious Case of Neural Text Degenerationcited by 2,916
  10. 2024LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Codecited by 2,004
  11. 2018Hierarchical Neural Story Generationcited by 1,193
  12. 2024Yi: Open Foundation Models by 01.AIcited by 719
12 more
  1. 2025Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoningcited by 559
  2. 2024StarCoder 2 and The Stack v2: The Next Generationcited by 502
  3. 2023DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Modelscited by 376
  4. 2024RoboPoint: A Vision-Language Model for Spatial Affordance Prediction for Roboticscited by 251
  5. 2024Turning Up the Heat: Min-p Sampling for Creative and Coherent LLM Outputscited by 150
  6. 2025RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Roboticscited by 127
  7. 2023Stay on topic with Classifier-Free Guidancecited by 91
  8. 2025Token Reduction Should Go Beyond Efficiency in Generative Models -- From Vision, Language to Multimodalitycited by 50
  9. 2025From Seeing to Doing: Bridging Reasoning and Decision for Robotic Manipulationcited by 35
  10. 2024Decoding Speculative Decodingcited by 34
  11. 2023Object-aware Inversion and Reassembly for Image Editingcited by 30
  12. 2023Draft & Verify: Lossless Large Language Model Acceleration via Self-Speculative Decodingcited by 8

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoningcited by 6
  2. 2026Directional Reasoning Trajectory Change (DRTC): Identifying Critical Trace Segments in Reasoning Modelscited by 1
  3. 2026Sample Where You Struggle: Sharpening Base Model Reasoning via Entropy-Guided Power Samplingcited by 1
  4. 2026Beyond Entropy: Learning from Token-Level Distributional Deviations for LLM Reasoningcited by 1
  5. 2026Multimodal Safety Evaluation Should Measure Controllability Beyond Classificationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Prompting Test-Time Scaling Is A Strong LLM Reasoning Data Augmentation
  2. 2025Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs
  3. 2026ThinkBooster: A Unified Framework for Seamless Test-Time Scaling of LLM Reasoning
  4. 2026Refining Over Resampling: Test-Time Self-Correction for LLM Reasoning
  5. 2025Sample Smart, Not Hard: Correctness-First Decoding for Better Reasoning in LLMs
  6. 2026Think Twice Before You Write -- an Entropy-based Decoding Strategy to Enhance LLM Reasoning
  7. 2026Beyond Confidence: Stability-Aware Test-Time Adaptation for LLM Reasoning
  8. 2024Inference Scaling vs Reasoning: An Empirical Analysis of Compute-Optimal LLM Problem-Solving