Pith. sign in

REVIEW 37 cited by

Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs

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 2501.18585 v2 pith:Z2YITTOP submitted 2025-01-30 cs.CL

classification cs.CL
keywords reasoningllmso1-likeunderthinkingchallengingthoughtsincorrectmodels
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Large language models (LLMs) such as OpenAI's o1 have demonstrated remarkable abilities in complex reasoning tasks by scaling test-time compute and exhibiting human-like deep thinking. However, we identify a phenomenon we term underthinking, where o1-like LLMs frequently switch between different reasoning thoughts without sufficiently exploring promising paths to reach a correct solution. This behavior leads to inadequate depth of reasoning and decreased performance, particularly on challenging mathematical problems. To systematically analyze this issue, we conduct experiments on three challenging test sets and two representative open-source o1-like models, revealing that frequent thought switching correlates with incorrect responses. We introduce a novel metric to quantify underthinking by measuring token efficiency in incorrect answers. To address underthinking, we propose a decoding strategy with thought switching penalty TIP that discourages premature transitions between thoughts, encouraging deeper exploration of each reasoning path. Experimental results demonstrate that our approach improves accuracy across challenging datasets without requiring model fine-tuning. Our findings contribute to understanding reasoning inefficiencies in o1-like LLMs and offer a practical solution to enhance their problem-solving capabilities.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 37 Pith papers

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

  1. CUSUM-Shaped Inference-Time Monitoring and Targeted Re-Decoding for Quantized Small Language Model Reasoning

    cs.AI 2026-07 conditional novelty 6.0 of 10

    A monitoring-guided rollback controller produced a small and statistically uncertain accuracy gain on first-seen MATH-500 problems, and a larger exploratory gain on a historical-coverage set contaminated by manual thr...

  2. CAM: A Causality-based Analysis Framework for Multi-Agent Code Generation Systems

    cs.SE 2026-02 conditional novelty 6.0 of 10

    CAM ranks how much each intermediate output of a multi-agent code-generation system causally matters for final correctness, then uses that ranking to repair failures and prune features.

  3. Controllable LLM Reasoning via Sparse Autoencoder-Based Steering

    cs.AI 2026-01 conditional novelty 6.0 of 10

    SAE-Steering finds, via keyword-logit recall plus effectiveness ranking, sparse-autoencoder features that steer a reasoning model into a chosen reasoning strategy, beating baseline steering by ~15% on a judge-based me...

  4. Reconsidering Overthinking: Penalizing Internal and External Redundancy in CoT Reasoning

    cs.AI 2025-08 conditional novelty 6.0 of 10

    A dual-penalty RL method that compresses chain-of-thought traces by separately penalizing internal semantic stagnation and external post-answer continuation reduces reasoning length by about 40% while preserving accur...

  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. It's Not That Simple. An Analysis of Simple Test-Time Scaling

    cs.LG 2025-07 conditional novelty 6.0 of 10

    The apparent scaling behavior of the s1 model is mostly an artifact of truncating long outputs, while appending 'Wait' does not reliably improve answers.

  7. NaturalThoughts: Selecting and Distilling Reasoning Traces for General Reasoning Tasks

    cs.CL 2025-07 conditional novelty 6.0 of 10

    Selecting difficult, strategy-diverse reasoning traces from DeepSeek-R1 on NaturalReasoning questions improves supervised distillation of reasoning into 7B-70B student models.

  8. HCNQA: Enhancing 3D VQA with Hierarchical Concentration Narrowing Supervision

    cs.CV 2025-07 conditional novelty 6.0 of 10

    HCNQA supervises three intermediate reasoning phases (BoI, OoI, OoT) derived from ScanQA annotations, improving EM@1 from 25.94 to 27.01 over 3D-VisTA.

  9. 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.

  10. AdapThink: Adaptive Thinking Preferences for Reasoning Language Model

    cs.LG 2025-06 conditional novelty 6.0 of 10

    AdapThink is an RL post-training framework that adaptively reduces overthinking and underthinking in reasoning language models by rewarding confidence-appropriate reasoning depth and diverse training samples.

  11. Excessive Reasoning Attack on Reasoning LLMs

    cs.CR 2025-06 conditional novelty 6.0 of 10

    A 10-token adversarial suffix, optimized with three custom losses, raises reasoning length 3x to 9x on open reasoning LLMs without hurting accuracy, and partly transfers to closed models.

  12. Towards Understanding the Cognitive Habits of Large Reasoning Models

    cs.CL 2025-06 conditional novelty 6.0 of 10

    A new benchmark shows large reasoning models exhibit human-like cognitive habits in their chain-of-thought, and some habits correlate with unsafe responses.

  13. Overclocking LLM Reasoning: Monitoring and Controlling Thinking Path Lengths in LLMs

    cs.LG 2025-06 conditional novelty 6.0 of 10

    A fitted progress vector in LLM hidden states can monitor and, when added during decoding, shorten the thinking phase, often improving accuracy and cutting tokens.

  14. Unlocking Recursive Thinking of LLMs: Alignment via Refinement

    cs.CL 2025-06 conditional novelty 6.0 of 10

    An offline alignment pipeline using reward-filtered self-refinement data and long chain-of-thought SFT raises an 8B model's AlpacaEval 2 win rate from 25.0% to 51.0% with roughly 14k training examples.

  15. Sample Complexity and Representation Ability of Test-time Scaling Paradigms

    cs.LG 2025-06 conditional novelty 6.0 of 10

    Best-of-n sampling provably needs about 1/Δ samples versus 1/Δ² for self-consistency, and a constructed Transformer can route among experts using verifier feedback to reach near-optimal final responses.

  16. EPiC: Towards Lossless Speedup for Reasoning Training through Edge-Preserving CoT Condensation

    cs.LG 2025-06 conditional novelty 6.0 of 10

    Keeping only the first and last parts of chain-of-thought training traces preserves math reasoning accuracy on MATH500 at half the sequence length, but not on harder benchmarks.

  17. ReasoningFlow: Semantic Structure of Complex Reasoning Traces

    cs.CL 2025-06 conditional novelty 6.0 of 10

    ReasoningFlow is a schema that labels the semantic roles and relationships in LLM reasoning traces as directed graphs, plus a small hand-annotated dataset and a pattern-query tool.

  18. AlphaOne: Reasoning Models Thinking Slow and Fast at Test Time

    cs.CL 2025-05 conditional novelty 6.0 of 10

    A stochastic scheduling framework that modulates slow-to-fast reasoning in large reasoning models at test time, improving accuracy while reducing token usage.

  19. What Makes a Good Reasoning Chain? Uncovering Structural Patterns in Long Chain-of-Thought Reasoning

    cs.AI 2025-05 reject novelty 6.0 of 10

    Reasoning trees parsed from long chain-of-thought text predict answer correctness better than length and expose failure patterns such as over-branching.

  20. Are Large Language Models Reliable AI Scientists? Assessing Reverse-Engineering of Black-Box Systems

    cs.LG 2025-05 conditional novelty 6.0 of 10

    LLMs struggle to use passive observations for reverse engineering, but active intervention improves performance, largely through the process of generating queries rather than the data obtained.

  21. Failure Cases Are Better Learned But Boundary Says Sorry: Facilitating Smooth Perception Change for Accuracy-Robustness Trade-Off in Adversarial Training

    cs.CV 2025-08 reject novelty 5.0 of 10

    A paper whose abstract describes new adversarial training experiments, but whose full text is a different paper on CoT compression, leaving the claims unsupported.

  22. REST: Stress Testing Large Reasoning Models by Asking Multiple Problems at Once

    cs.CL 2025-07 conditional novelty 5.0 of 10

    Asking a reasoning model several problems at once reveals large accuracy drops and exposes differences that single-question benchmarks miss.

  23. OpenCodeReasoning-II: A Simple Test Time Scaling Approach via Self-Critique

    cs.CL 2025-07 conditional novelty 5.0 of 10

    A 2.5M-example code reasoning dataset with critique traces enables Qwen2.5-based models to surpass prior open-weight distilled models on LiveCodeBench via test-time self-critique selection.

  24. Adaptive Termination for Multi-round Parallel Reasoning: An Universal Semantic Entropy-Guided Framework

    cs.CL 2025-07 conditional novelty 5.0 of 10

    A semantic entropy-guided stopping rule for multi-round parallel LLM reasoning improves accuracy while reducing inference steps on five benchmarks.

  25. Activation Steering for Chain-of-Thought Compression

    cs.AI 2025-07 conditional novelty 5.0 of 10

    A single steering vector extracted from paired verbose and concise rationales compresses chain-of-thought output at inference time without retraining.

  26. Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs

    cs.AI 2025-07 conditional novelty 5.0 of 10

    The survey's L1/L2 taxonomy and benchmark show that current reasoning models waste compute on easy problems and underthink hard ones, motivating more adaptive inference.

  27. Large Reasoning Models are not thinking straight: on the unreliability of thinking trajectories

    cs.LG 2025-07 conditional novelty 5.0 of 10

    Large reasoning models frequently ignore explicitly injected ground-truth solutions mid-trajectory and continue to overthink, often ending on incorrect answers.

  28. Efficient Reasoning Through Suppression of Self-Affirmation Reflections in Large Reasoning Models

    cs.CL 2025-06 conditional novelty 5.0 of 10

    Suppressing low-probability 'wait' tokens that trigger self-affirmation reflections shortens reasoning-model output by 8-18% with roughly unchanged accuracy.

  29. Atomic-to-Compositional Generalization for Mobile Agents with A New Benchmark and Scheduling System

    cs.CL 2025-06 conditional novelty 5.0 of 10

    AI phone agents succeed on single-step commands but fail on multi-step compositions; a new benchmark measures this gap and a scheduler that decomposes tasks recovers most of it.

  30. One Missing Piece for Open-Source Reasoning Models: A Dataset to Mitigate Cold-Starting Short CoT LLMs in RL

    cs.CL 2025-06 conditional novelty 5.0 of 10

    A pipeline converts short-CoT LLM outputs into o1-style long chain-of-thought rationales using 1K seed reasoning flows, and SFT on the resulting dataset improves downstream RLVR cold-start.

  31. Control-R: Towards controllable test-time scaling

    cs.AI 2025-05 conditional novelty 5.0 of 10

    Control-R-32B is a fine-tuned 32B model that takes 11 control-field scores in the prompt, claiming to let users trade reasoning depth for efficiency and reaching SOTA on AIME2024 and MATH500 at that scale.

  32. Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training

    cs.AI 2025-05 conditional novelty 5.0 of 10

    Reinforcing the two experts most correlated with thinking tokens improves reasoning accuracy and efficiency in MoE large reasoning models, with gains of up to 10 points on AIME benchmarks.

  33. Self-Reasoning Language Models: Unfold Hidden Reasoning Chains with Few Reasoning Catalyst

    cs.CL 2025-05 conditional novelty 5.0 of 10

    Self-Reasoning Language Models use 1,000 seed catalyst rationales to teach a model to expand its own reasoning chains and iteratively self-improve, gaining about 2.5 points on average across five benchmarks.

  34. From Long to Short: LLMs Excel at Trimming Own Reasoning Chains

    cs.AI 2025-09 conditional novelty 4.0 of 10

    EDIT searches across step-count prompts to find the shortest reasoning path the model consistently answers correctly.

  35. A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Models

    cs.CL 2025-09 conditional novelty 4.0 of 10

    A structured literature survey concluding that reasoning capabilities do not automatically make LLMs more trustworthy and can introduce new vulnerabilities in safety, robustness, and privacy.

  36. First Finish Search: Efficient Test-Time Scaling in Large Language Models

    cs.CL 2025-05 conditional novelty 4.0 of 10

    First Finish Search launches n parallel reasoning traces and returns the shortest one, improving accuracy on AIME benchmarks while cutting token use.

  37. A Survey on Large Language Models for Mathematical Reasoning

    cs.AI 2025-06 conditional novelty 1.0 of 10

    Recent advances in LLM mathematical reasoning are organized into comprehension and generation phases, covering methods from prompting to test-time scaling.

Pith tools