mlsanityData to 7 Oct 2026

Parallel-Probe: Towards Efficient Parallel Thinking via 2D Probing

arXiv 2602.03845 ↗First posted 3 Feb 2026cs.CLcited by 13

Not on the map. Why?

Abstract (arXiv)

Parallel thinking has emerged as a promising paradigm for reasoning, yet it imposes significant computational burdens. Existing efficiency methods primarily rely on local, per-trajectory signals and lack principled mechanisms to exploit global dynamics across parallel branches. We introduce 2D probing, an interface that exposes the width-depth dynamics of parallel thinking by periodically eliciting intermediate answers from all branches. Our analysis reveals three key insights: non-monotonic scaling across width-depth allocations, heterogeneous reasoning branch lengths, and early stabilization of global consensus. Guided by these insights, we introduce textbfParallel-Probe, a training-free controller designed to optimize online parallel thinking. Parallel-Probe employs consensus-based early stopping to regulate reasoning depth and deviation-based branch pruning to dynamically adjust width. Extensive experiments across three benchmarks and multiple models demonstrate that Parallel-Probe establishes a superior Pareto frontier for test-time scaling. Compared to standard majority voting, it reduces sequential tokens by up to 35.8% and total token cost by over 25.8% while maintaining competitive accuracy.

Key facts. Cited by 13 papers in the map (data through 7 Oct 2026). Builds on 36 of its 39 references that are in the map. First posted on arXiv 3 Feb 2026.

Builds on

36 of 39 references.

References · 36

  1. 2025Qwen3 Technical Reportcited by 8,195
  2. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  3. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,655
  4. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,928
  5. 2025MathArena: Evaluating LLMs on Uncontaminated Math Competitionscited by 275
  6. 2022Generating Sequences by Learning to Self-Correctcited by 271
  7. 2025Dynamic Early Exit in Reasoning Modelscited by 217
  8. 2025Deep Think with Confidencecited by 199
  9. 2025Reasoning Models Know When They're Right: Probing Hidden States for Self-Verificationcited by 154
  10. 2023Let's Sample Step by Step: Adaptive-Consistency for Efficient Reasoning and Coding with LLMscited by 111
  11. 2025Confidence Improves Self-Consistency in LLMscited by 111
  12. 2024Forest-of-Thought: Scaling Test-Time Compute for Enhancing LLM Reasoningcited by 110
24 more
  1. 2024Escape Sky-high Cost: Early-stopping Self-Consistency for Multi-step Reasoningcited by 105
  2. 2025Efficient Test-Time Scaling via Self-Calibrationcited by 74
  3. 2024Reasoning Aware Self-Consistency: Leveraging Reasoning Paths for Efficient LLM Samplingcited by 63
  4. 2025Sampling-Efficient Test-Time Scaling: Self-Estimating the Best-of-N Sampling in Early Decodingcited by 62
  5. 2024Make Every Penny Count: Difficulty-Adaptive Self-Consistency for Cost-Efficient Reasoningcited by 61
  6. 2025Parallel-R1: Towards Parallel Thinking via Reinforcement Learningcited by 59
  7. 2025SETS: Leveraging Self-Verification and Self-Correction for Improved Test-Time Scalingcited by 54
  8. 2025Answer Convergence as a Signal for Early Stopping in Reasoningcited by 38
  9. 2025ParaThinker: Native Parallel Thinking as a New Paradigm to Scale LLM Test-time Computecited by 36
  10. 2025Hogwild! Inference: Parallel LLM Generation via Concurrent Attentioncited by 32
  11. 2025Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lenscited by 32
  12. 2025Multiverse: Your Language Models Secretly Decide How to Parallelize and Merge Generationcited by 30
  13. 2025AlphaOne: Reasoning Models Thinking Slow and Fast at Test Timecited by 27
  14. 2025Entropy After </Think> for reasoning model early exitingcited by 24
  15. 2025Group Think: Multiple Concurrent Reasoning Agents Collaborating at Token Level Granularitycited by 22
  16. 2025Think Just Enough: Sequence-Level Entropy as a Confidence Signal for LLM Reasoningcited by 22
  17. 2025Early Stopping Chain-of-thoughts in Large Language Modelscited by 19
  18. 2025Multi-Crit: Benchmarking Multimodal Judges on Pluralistic Criteria-Followingcited by 14
  19. 2025Slim-SC: Thought Pruning for Efficient Scaling with Self-Consistencycited by 12
  20. 2025DeepPrune: Parallel Scaling without Inter-trace Redundancycited by 10
  21. 2026Hidden States as Early Signals: Step-level Trace Evaluation and Pruning for Efficient Test-Time Scalingcited by 10
  22. 2024Divide, Reweight, and Conquer: A Logit Arithmetic Approach for In-Context Learningcited by 9
  23. 2025Iterative Deepening Sampling as Efficient Test-Time Scalingcited by 8
  24. 2026SyncThink: A Training-Free Strategy to Align Inference Termination with Reasoning Saturationcited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 13.

13 papers.

Sort:

By area:

  1. 2026Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perceptioncited by 47
  2. 2026LLMs Improving LLMs: Agentic Discovery for Test-Time Scalingcited by 7
  3. 2026Unsafe by Reciprocity: How Generation-Understanding Coupling Undermines Safety in Unified Multimodal Modelscited by 2
  4. 2026Share More, Search Less: Collaborative Parallel Thinking for Efficient Test-Time Scalingcited by 2
  5. 2026MARS: Margin-Adversarial Risk-controlled Stopping for Parallel LLM Test-time Scalingcited by 2
  6. 2026Mitigating Factual Hallucination in Large Reasoning Models via Mixed-Mode Advantage Regularizationcited by 2
  7. 2026Small RL Controller, Large Language Model: RL-Guided Adaptive Sampling for Test-Time Scalingcited by 1
  8. 2026ParaTempo: Efficient Parallel Reasoning via Temporal Confidencecited by 1
  9. 2026Process Rewards with Learned Reliabilitycited by 0
  10. 2026Understanding Evaluation Illusion in Diffusion Large Language Modelscited by 0
  11. 2026LatentSift: Policy-State Filtering for Token-Efficient Verification of Software Engineering Agentscited by 0
  12. 2026Align Then Reason: A Multimodal Lip-Sync Judge for Dubbingcited by 0
1 more
  1. 2026From Pareto to Preference: Personalized Test-Time Scaling via Amortized Agentic Policy Discoverycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025ThreadWeaver: Adaptive Threading for Efficient Parallel Reasoning in Language Models
  2. 2025Let Me Think! A Long Chain-of-Thought Can Be Worth Exponentially Many Short Ones
  3. 2025Adaptive Termination for Multi-round Parallel Reasoning: An Universal Semantic Entropy-Guided Framework
  4. 2026Planned Test-Time Scaling with Coordinated Reasoning Paths
  5. 2025ParallelMuse: Agentic Parallel Thinking for Deep Information Seeking
  6. 2026PaCoRe: Learning to Scale Test-Time Compute with Parallel Coordinated Reasoning
  7. 2025Scaling over Scaling: Exploring Test-Time Scaling Plateau in Large Reasoning Models
  8. 2025A2R: An Asymmetric Two-Stage Reasoning Framework for Parallel Reasoning