AI Paper Atlas

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

ANTS: Adaptive Negative Textual Space Shaping for OOD Detection via Test-Time MLLM Understanding and Reasoning

Wenjie Zhu, Yabin Zhang, Xin Jin, Wenjun Zeng, Lei Zhang

CVPR 2026 · Label-efficient and robust learning › OOD and anomaly detection · approach: Spurious-aware OOD 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
add negative image mining before ENS generation (ablation)ANTS · OpenOOD far-OODFPR95: 43.87 -> 19.22Improved
add negative image mining before ENS generation (ablation)ANTS · OpenOOD near-OODFPR95: 74.48 -> 73.70No clear effect
use ENS with mined negative images vs NegLabel (ablation)ANTS · OpenOOD far-OODFPR95: 27.34 -> 19.22Improved
add visually similar ID-class mining before VSNL (ablation)ANTS · OpenOOD near-OODFPR95: 74.36 -> 63.11Improved
add visually similar ID-class mining before VSNL (ablation)ANTS · OpenOOD far-OODFPR95: 53.82 -> 23.44Improved
combine ENS and VSNL with fixed lambda 0.5 (ablation)ANTS · OpenOOD near-OODFPR95: 73.70 -> 62.05Improved
use adaptive lambda instead of 0.5 (ablation)ANTS · OpenOOD far-OODFPR95: 21.65 -> 15.38Improved
use adaptive lambda instead of 0.5 (ablation)ANTS · OpenOOD near-OODFPR95: 62.05 -> 60.98Improved
generate visually similar labels for all ID classes (delta=1) (ablation)ANTS · ImageNet-1k near-OODFPR95Hurt
lengthen generated negative sentences excessively (ablation)ANTS · ImageNet-1kFPR95: best at average length 8.4Mixed
use inexpressive or dissimilar MLLM prompts (ablation)ANTS · ImageNet-1kFPR95Hurt
ANTS vs strongest zero-shot baseline CSPANTS · ImageNet-1k avg of 4 OOD setsFPR95: 17.51 -> 11.20Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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