AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

Fantastic Tractor-Dogs and How Not to Find Them With Open-Vocabulary Detectors

Frank Ruis, Gertjan J. Burghouts, Hugo Kuijf

ICLR 2026 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary object detection

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
append attention sink tokens to the promptMM-GDINO Swin-T · LVIS MiniVal (negative-class prompts)AP at FPRbg 0.01: 0.076 -> 0.214Improved
append attention sink tokens to the promptLLMDet Swin-T · LVIS MiniVal (negative-class prompts)AP at FPRbg 0.01: 0.045 -> 0.223Improved
append attention sink tokens to the promptLLMDet Swin-T · LVIS MiniVal (negative-class prompts)AP at FPRbg 0.05: 0.140 -> 0.326Improved
append attention sink tokens to the promptLLMDet Swin-B · LVIS MiniVal (negative-class prompts)AP at FPRbg 0.05: 0.125 -> 0.349Improved
append attention sink tokens to the promptGLIP Swin-T · LVIS MiniVal (negative-class prompts)standard AP (no FPR filter): 0.333 -> 0.302Hurt
append attention sink tokens to the promptLLMDet Swin-L · LVIS MiniVal (negative-class prompts)standard AP (no FPR filter): 0.488 -> 0.502No clear effect
append attention sink tokens to the promptGDINO Swin-T · POPE detection FPRbg 0.05AP: 0.413 -> 0.413No clear effect
append attention sink tokens to the promptLLMDet-T · POPE detection FPRbg 0.05AP: 0.278 -> 0.417Improved
learn the attention sink embeddings by textual inversion instead of training-free init (ablation)attention sinks · VOC-based sink trainingAP at low FPRbgNo clear effect
disable problematic attention heads instead of using attention sinks (ablation)early-fusion OVD · background false positivesAP at low FPRbgMixed
use more than about 8 attention sinks (24 optimal) (ablation)LLMDet-T · LVIS MiniValAP at low FPRbgImproved
add vision sinks to vision features in addition to language sinks (ablation)attention sinks · LVIS MiniValAP at low FPRbgNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy