# Generative Verification: Rethinking the Uncertainty Signal for Active Learning of Object Detection

arXiv: https://arxiv.org/abs/2609.20262 · HTML: https://mlsanity.com/paper/2609.20262/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 95 references that are in the map.
- First posted on arXiv 28 Jul 2026.

## Builds on (most cited first, 8 of 8)

- [Active Learning for Deep Object Detection](https://mlsanity.com/paper/1809.09875/) (2018)
- [ALBench: A Framework for Evaluating Active Learning in Object Detection](https://mlsanity.com/paper/2207.13339/) (2022)
- [Active Learning for Object Detection with Non-Redundant Informative Sampling](https://mlsanity.com/paper/2307.08414/) (2023)
- [Streamlining the Development of Active Learning Methods in Real-World Object Detection](https://mlsanity.com/paper/2508.19906/) (2025)
- [DeLR: Active Learning for Detection with Decoupled Localization and Recognition Query](https://mlsanity.com/paper/2312.16931/) (2023)
- [Performance-guided Reinforced Active Learning for Object Detection](https://mlsanity.com/paper/2601.15688/) (2026)
- [Active-SAOOD: Active Sparsely Annotated Oriented Object Detection in Remote Sensing Images](https://mlsanity.com/paper/2605.10162/) (2026)
- [Portable Active Learning for Object Detection](https://mlsanity.com/paper/2605.10349/) (2026)

## Abstract (arXiv)

Nearly every acquisition function for active object detection shares one arrangement, in that the model being improved is also the model being interrogated. We depart from it. In generative verification an independent generative model re-derives the label of a detection from the pixels inside its predicted box, and the disagreement between the two becomes the acquisition signal. Two properties follow from the arrangement itself rather than from any tuning. A displaced box, a box on background and a correct box carrying the wrong label all yield a crop that fails verification, so the failure modes arrive already combined in one scalar and the hand-weighted classification and localization terms of existing criteria are no longer needed. And because the verifier never observes the detector confidence, confidently wrong detections score highest, although a self-derived signal reads them as uninteresting and they are the costliest to leave unlabeled. We build the verifier as a conditional diffusion model whose diffusion target is a label representation rather than an image. Its reverse process is stochastic, so repeated generations return a distribution whose concentration reports how firmly the evidence determines the label, where a classifier returns a single point estimate. On PASCAL VOC and MS-COCO the signal outperforms output-uncertainty, feature-geometry, perturbation and ensemble criteria, gaining about one mAP50 point per round on MS-COCO, with its largest margins in the early rounds where confident detector errors are most common.