Research map / Label-efficient and robust learning / OOD and anomaly detection

ConfDet: Learning Reliable Confidence for MLLM-based Detection

NeurIPS 2026 · Label-efficient and robust learning › OOD and anomaly detection · approach: Uncertainty and misclassification detection

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes ConfDet, a method for learning reliable confidence scores for MLLM-based object detection outputs, rather than relying on implicit confidence from self-assessment, generation probabilities, or external proxy scores.

Problem

MLLM-based detectors produce detection results without reliable confidence estimates, and detection data lacks human-annotated confidence labels, so existing methods derive confidence implicitly from prompted self-assessment, generation probabilities, or external proxy scores.

Approach

Learns explicit confidence for MLLM-generated detections instead of deriving it implicitly; the abstract does not describe the specific mechanism beyond this motivation.

Closest papers by meaning

• Linking Perception, Confidence and Accuracy in MLLMs (CVPR 2026 · similarity 0.85)
• Pearls from Pebbles: Improved Confidence Functions for Auto-labeling (NeurIPS 2024 · similarity 0.84)
• Rethinking Confidence Scores and Thresholds in Pseudolabeling-based SSL (ICML 2025 · similarity 0.84)
• Confidence Difference Reflects Various Supervised Signals in Confidence-Difference Classification (ICML 2025 · similarity 0.84)
• Enhanced Sample Selection with Confidence Tracking: Identifying Correctly Labeled Yet Hard-to-Learn Samples in Noisy Data (AAAI 2025 · similarity 0.83)
• Quantifying Uncertainty in Error Consistency: Towards Reliable Behavioral Comparison of Classifiers (NeurIPS 2025 · similarity 0.83)