Calibration Is Not Control: Intervention Value for LLM-Agent Oversight
cs.AI
Submitted: 2026-06-19
Updated: 2026-10-06
Comments: NeurIPS 2026
Code: https://github.com/bennidict23/calibration-is-not-control
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Training Verifiers to Solve Math Word Problems
- Calibrate-Then-Act: Cost-Aware Exploration in LLM Agents
- The Llama 3 Herd of Models
- Language Models (Mostly) Know What They Know
- Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
- ATLAS: Agentic Test-time Learning-to-Allocate Scaling
- Qwen2.5 Technical Report
- Causal LLM Routing: End-to-End Regret Minimization from Observational Data
- Solving math word problems with process- and outcome-based feedback
- Accurate Failure Prediction in Agents Does Not Imply Effective Failure Prevention
- Fail-Fast, Restart-Smart: Early Failure Prediction and Restart for SWE Agentic Tasks
- Self-Consistency Improves Chain of Thought Reasoning in Language Models
- The Confidence Dichotomy: Analyzing and Mitigating Miscalibration in Tool-Use Agents
- Tree of Thoughts: Deliberate Problem Solving with Large Language Models
- Tracing Agentic Failure from the Flow of Success
- Agentic Confidence Calibration
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems