# Detector Confidence Signals Presence Rather Than Occlusion in Cluttered Manipulation

arXiv: https://arxiv.org/abs/2607.13361 · HTML: https://mlsanity.com/paper/2607.13361/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 21 references that are in the map.
- First posted on arXiv 15 Jul 2026.

## Builds on (most cited first, 20 of 20)

- [Segment Anything](https://mlsanity.com/paper/2304.02643/) (2023)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection](https://mlsanity.com/paper/2303.05499/) (2023)
- [OpenVLA: An Open-Source Vision-Language-Action Model](https://mlsanity.com/paper/2406.09246/) (2024)
- [Evaluating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2305.10355/) (2023)
- [Grounded Language-Image Pre-training](https://mlsanity.com/paper/2112.03857/) (2021)
- [The 2017 DAVIS Challenge on Video Object Segmentation](https://mlsanity.com/paper/1704.00675/) (2017)
- [SAPIEN: A SimulAted Part-based Interactive ENvironment](https://mlsanity.com/paper/2003.08515/) (2020)
- [YOLO-World: Real-Time Open-Vocabulary Object Detection](https://mlsanity.com/paper/2401.17270/) (2024)
- [Scaling Open-Vocabulary Object Detection](https://mlsanity.com/paper/2306.09683/) (2023)
- [Simple Open-Vocabulary Object Detection with Vision Transformers](https://mlsanity.com/paper/2205.06230/) (2022)
- [MOSE: A New Dataset for Video Object Segmentation in Complex Scenes](https://mlsanity.com/paper/2302.01872/) (2023)
- [ManiSkill3: GPU Parallelized Robotics Simulation and Rendering for Generalizable Embodied AI](https://mlsanity.com/paper/2410.00425/) (2024)
- [Closed-Loop Next-Best-View Planning for Target-Driven Grasping](https://mlsanity.com/paper/2207.10543/) (2022)
- [VISO-Grasp: Vision-Language Informed Spatial Object-centric 6-DoF Active View Planning and Grasping in Clutter and Invisibility](https://mlsanity.com/paper/2503.12609/) (2025)
- [Score the Steps, Not Just the Goal: VLM-Based Subgoal Evaluation for Robotic Manipulation](https://mlsanity.com/paper/2509.19524/) (2025)
- [ActiveGrasp: Information-Guided Active Grasping with Calibrated Energy-based Model](https://mlsanity.com/paper/2511.12795/) (2025)
- [From Grounding to Manipulation: Case Studies of Foundation Model Integration in Embodied Robotic Systems](https://mlsanity.com/paper/2505.15685/) (2025)
- [Overcoming Visual Clutter in Vision Language Action Models via Concept-Gated Visual Distillation](https://mlsanity.com/paper/2603.10340/) (2026)
- [Sample-efficient Low-level Motion Planning for Robotic Manipulation Tasks via Zero-shot Transfer Learning](https://mlsanity.com/paper/2606.06041/) (2026)

## Abstract (arXiv)

Occlude a named object until about an eighth of it remains visible, and an open-vocabulary detector's confidence that the object is present barely changes; as the clutter around it grows the confidence can even rise. On real video the detector still reports the object present in 99% of occluded frames, on another instance of the same category. This matters because that confidence is widely read as a visibility signal, used to threshold detections, evaluate open-vocabulary detectors, ground language, retrieve instances, and gate active perception. We audit whether it reflects occlusion by pairing every view with a geometry-segmentation oracle that gives detector-free ground-truth visibility. As true visibility falls from every scene to one in eight, the confidence stays nearly constant and uncorrelated with visibility, and the detector reports the target present in about nine of ten scenes, firing on same-category distractors: it signals that the category is present somewhere, not that the specific target is visible. The failure holds across three detectors (Grounding DINO, OWLv2, and Segment Anything Model 3), nine object categories, two simulators with different renderers and object sets, built and natural occlusion, and real video. Two consequences follow: a confidence-based metric understates the value of resolving occlusion by about ten times (8 against 88 points in our active-perception setting), and a confidence-based gate fires exactly when the object is hidden. No single-view signal we tried, including a realizable localization check, flags the occlusion, because the occluders sit where the target is. We connect the effect to detector miscalibration and object hallucination, release the controlled benchmark, and recommend target-grounded signals for gating and evaluation.