{"slug": "confidence-calibrated-inference-expansion-for-evaluator-guided-test-time", "title": "Confidence-Calibrated Inference Expansion for Evaluator-Guided Test-Time Reasoning", "summary": "A compute-aware three-action controller for evaluator-guided test-time reasoning improved observed utility over calibration-selected fixed policies while cutting normalized online compute by about 30% on held-out width expansion and 16% on revision tasks, according to the arXiv paper 2610.20836v1. The controller scores terminal actions by predicted answer success and expansion by a rollout-based continuation-success surrogate, with compute cost priced explicitly, and learns action-specific confidence estimates with local uncertainty penalties for ambiguous decisions. Across four Qwen3-8B seeds, compute was lower than fitted Q in both settings while utility differences were small and mixed, framing evaluator-guided reasoning as resource allocation under uncertainty rather than a uniformly stronger reasoner.", "body_md": "arXiv:2610.20836v1 Announce Type: new \nAbstract: 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.", "url": "https://wpnews.pro/news/confidence-calibrated-inference-expansion-for-evaluator-guided-test-time", "canonical_source": "https://arxiv.org/abs/2610.20836", "published_at": "2026-10-16 04:00:00+00:00", "updated_at": "2026-10-16 05:06:02.601968+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "large-language-models", "ai-research"], "entities": ["Qwen3-8B", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/confidence-calibrated-inference-expansion-for-evaluator-guided-test-time", "markdown": "https://wpnews.pro/news/confidence-calibrated-inference-expansion-for-evaluator-guided-test-time.md", "text": "https://wpnews.pro/news/confidence-calibrated-inference-expansion-for-evaluator-guided-test-time.txt", "jsonld": "https://wpnews.pro/news/confidence-calibrated-inference-expansion-for-evaluator-guided-test-time.jsonld"}}