ParaTempo: Efficient Parallel Reasoning via Temporal Confidence
2026-08-17 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors present ParaTempo, a method to make large reasoning models more efficient by managing multiple reasoning paths based on a new idea called temporal confidence. This measure checks how focused each reasoning branch is on a particular answer over time, allowing the system to stop unhelpful paths early and spend more effort on promising ones. Their approach does not need extra training or synced steps, and it reduces both the time and computational work needed while keeping accuracy similar. Tests show that ParaTempo is better at predicting which branches will succeed compared to older methods.
parallel reasoningtemporal confidenceasynchronous computationreasoning branchespruningtoken-level confidencereasoning trajectorylarge language modelsanswer-space convergencelatency reduction
Authors
Xuteng Zhang, Wenhao Zeng, Xiaodong Gu, Chao Hu, Haotian Lin, Yuling Shi, Min Wang, Beijun Shen
Abstract
Parallel reasoning improves the accuracy and robustness of large reasoning models by exploring multiple solution paths, but its computational cost grows with reasoning depth and branch count. Existing methods for managing these parallel paths typically rely on final-answer consensus, local token confidence, or isolated intermediate probes. However, these signals are often delayed, weakly tied to actual reasoning progress, or too noisy for dynamic, branch-level control. To address these limitations, we introduce ParaTempo, a training-free asynchronous parallel reasoning framework. ParaTempo is driven by temporal confidence, a branch-local measure of answer-space convergence. Each branch is periodically probed for a tentative answer probability distribution, and temporal confidence quantifies how sharply the recent intermediate probes concentrate on a dominant answer. Once sufficient evidence has accumulated, ParaTempo drives its entire control process from this single signal: low-confidence branches are pruned, branches that persistently commit to their dominant answer are retired early, freed computation is reallocated by forking new branches, and generation stops globally once the confidence-weighted vote concentrates. Without requiring synchronization among reasoning trajectories, ParaTempo adaptively allocates computation based on branch-level convergence. Experiments on challenging mathematical and scientific reasoning benchmarks show that ParaTempo reduces average latency by 21.8-32.2% and total token usage by 18.1-30.3% while maintaining competitive accuracy. Moreover, temporal confidence exhibits stronger temporal stability and predictive power for future branch convergence than token-level and instantaneous signals.