AI Paper Atlas

Research map / Video / Video-language understanding

WatchLog: Efficient and Interpretable Event Reasoning for Endpoint Detection and Response Logs with Multimodal LLMs

Hongyi Zhou, Jianfeng Pan, Min Peng, Shaomang Huang, Xuling Zhang

ICML 2026 · Video › Video-language understanding · approach: Event and anomaly reasoning in video

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 image-event alignment pre-training (ablation)WatchLog · EDR8M-20ROverall accuracy: 5.3 -> 88.8Improved
add video-log alignment pre-training on top of image-event alignment (ablation)WatchLog · EDR8M-20ROverall accuracy: 88.8 -> 90.3Improved
add alignment pre-training (ablation)WatchLog · EDR8M-20RFalse alarm: 100.0 -> 5.0Improved
add image-event alignment pre-training (ablation)WatchLog · EDR8M-20RRationale completeness: 32.1 -> 67.2Improved
add video-log alignment pre-training (ablation)WatchLog · EDR8M-20RRationale completeness: 67.2 -> 68.5Improved
scale backbone from 2B to 3B (ablation)WatchLog · EDR8M-20ROverall accuracy: 90.3 -> 91.3Improved
scale backbone from 2B to 7B (ablation)WatchLog · EDR8M-20ROverall accuracy: 90.3 -> 92.5Improved
scale backbone from 2B to 7B (ablation)WatchLog · EDR8M-20RRationale completeness: 68.5 -> 78.2Improved
scale backbone from 2B to 3B (ablation)WatchLog · EDR8M-20RFalse alarm: 5.0 -> 0.0Improved
replace long-context LLM with WatchLog video encoding (ablation)WatchLog vs Qwen2.5-1.5B-Instruct · 1024K-token inputGPU memory: 98.46 -> 48.61Improved
replace long-context LLM with WatchLog video encoding (ablation)WatchLog vs Qwen2.5-1.5B-Instruct · 1024K-token inputTTFT: 259.30 -> 1.11Improved
replace long-context LLM with WatchLog video encoding (ablation)WatchLog vs Qwen2.5-7B-Instruct · 1024K-token inputTTFT: 486.19 -> 1.11Improved

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