AI Paper Atlas

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

ADSeeker: A Knowledge-Grounded Reasoning Framework for Industry Anomaly Detection and Reasoning

Kai Zhang, Zekai Zhang, Xihe Sun, Anpeng Wang, Jingmeng Nie, Qinghui Chen, Han Hao, Jianyuan Guo, Jinglin Zhang

CVPR 2026 · Label-efficient and robust learning › OOD and anomaly detection · approach: Unified multi-class anomaly 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 Q2K RAG (ablation)ADSeeker baseline · MMAD two subtasksaccuracy (%): 76.6 -> 78.6Improved
add AD Expert module (ablation)ADSeeker baseline · MMAD two subtasksaccuracy (%): 76.6 -> 77.9Improved
combine Q2K RAG and AD Expert (ablation)ADSeeker baseline · MMAD two subtasksaccuracy (%): 76.6 -> 82.8Improved
LoRA fine-tune for 5 epochs on 43K instruction data (ablation)baseline MLLM · MMAD two subtasksaccuracy (%): 76.6 -> 80.0Improved
extend LoRA fine-tuning from 5 to 20 epochs (ablation)baseline MLLM · MMAD two subtasksaccuracy (%): 80.0 -> 40.9Hurt
use Q2K RAG framework instead of LoRA fine-tuningbaseline MLLM · MMAD two subtasksaccuracy (%): 80.0 -> 82.8Improved
apply Seek-Setting (Q2K RAG + AD Expert) to Qwen2.5-VL 7BQwen2.5-VL 7B · MMAD 1-shotaverage accuracy (%): 66.62 -> 69.90Improved
apply Seek-Setting to LLaVA-NeXT 7BLLaVA-NeXT 7B · MMAD 1-shotaverage accuracy (%): 59.32 -> 65.96Improved
apply Seek-Setting to LLaVA-OV 7BLLaVA-OV 7B · MMAD 1-shot defect descriptionaccuracy (%): 62.19 -> 57.17Hurt
apply Seek-Setting to Qwen2.5-VL 7BQwen2.5-VL 7B · MMAD 1-shot object classificationaccuracy (%): 92.07 -> 89.91Hurt
apply Seek-Setting to LLaVA-NeXT 7BLLaVA-NeXT 7B · MMAD 1-shot defect classificationaccuracy (%): 33.79 -> 53.29Improved
apply Seek-Setting without extra training to LLaVA-OV vs trained Anomaly-OVLLaVA-OV · MMAD subset averageaccuracy (%): 69.80 -> 73.54Improved

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