Confidence-Calibrated Inference Expansion for Evaluator-Guided Test-Time Reasoning

arXiv:2610.20836v1 Announce Type: new Abstract: Evaluator-guided reasoning presents a local control problem at every step: trust the current answer, return a fallback, or spend more inference. We formulate this choice as a compute-aware three-action controller, scoring terminal actions by predicted answer success and expansion by a rollout-based continuation-success surrogate, with compute cost priced explicitly. A local action-gap analysis characterizes when score errors can change a decision, while an expected regret bound accounts for errors along the realized expansion path. Guided by this analysis, we learn and calibrate action-specific confidence estimates, with local uncertainty penalties for ambiguous decisions. On held-out width expansion and revision tasks, the controller improves observed utility over calibration-selected fixed policies while reducing normalized online compute by about 30% and 16%, respectively. Across four Qwen3-8B seeds, compute is lower than fitted Q in both settings, while utility differences are small and mixed. The results frame evaluator-guided reasoning as resource allocation under uncertainty, rather than a uniformly stronger reasoner.

Related

• Beyond Confidence: Stability-Aware Test-Time Adaptation for LLM Reasoning
• ThinkFuse: Trajectory-Aware Test-Time Fusion for Small Reasoning Models
• Aligning Large Vision-Language Models at Test Time: A Trajectory-Guided Structured Sampling Approach
• Do I Need the Cloud? Uncertainty-Aware Step-Level Handoff for Small Language Model Agents
• CADER: Confidence-Aware Dynamic Evidence Reasoning for Long-Video Understanding

Source: arXiv cs.AI | 2026-10-16