Research map / Large language models / Alignment and preferences
Linguistic Relative Policy Optimization for Video Anomaly Reasoning
Jiaxu Leng, Jiankang Zheng, Mengjingcheng Mo, Zhanjie Wu, Haosheng Chen, Ji Gan, Xinbo Gao
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 changed | Where | Measured | Effect |
|---|---|---|---|
| inject LRPO-learned linguistic experience into context (ablation) | InternVL3 5-8B · XD-Violence | AP (%): 59.93 -> 73.17 | Improved |
| learn experience from group of trajectories instead of single trajectory (ablation) | LRPO · XD-Violence | AP (%): 66.78 -> 73.17 | Improved |
| add scenario experience on top of general experience (accuracy reward) (ablation) | LRPO · XD-Violence | AP (%): 68.48 -> 70.58 | Improved |
| add anomaly preference reward (ablation) | LRPO · XD-Violence | AP (%): 68.48 -> 69.78 | Improved |
| add anomaly temporal dependency reward (ablation) | LRPO · XD-Violence | AP (%): 68.48 -> 70.06 | Improved |
| combine all three rewards (ablation) | LRPO · XD-Violence | AP (%): 68.48 -> 71.91 | Improved |
| random scenario experience injection (ablation) | LRPO · XD-Violence | AP (%): 71.91 -> 71.61 | Hurt |
| visual-semantic retrieval instead of visual-only retrieval (ablation) | LRPO scenario injection · XD-Violence | AP (%): 72.54 -> 73.17 | Improved |
| progressively refine experience during optimization (ablation) | LRPO · XD-Violence | AP (%): 67.63 -> 73.17 | Improved |
| increase scenario experience ratio at inference (ablation) | LRPO · XD-Violence | AP (%): 73.17 -> 74.52 | Improved |
| scale scenario experience during experience learning (ablation) | LRPO · XD-Violence | AP (%): 73.17 -> 71.20 | Hurt |
| use full training set instead of 100 videos (ablation) | LRPO · XD-Violence | AP (%): 73.17 -> 74.09 | No clear effect |
Closest papers by meaning
- Vad-R1: Towards Video Anomaly Reasoning via Perception-to-Cognition Chain-of-Thought (NeurIPS 2025 · similarity 0.88)
- Harnessing Large Language Models for Training-free Video Anomaly Detection (CVPR 2024 · similarity 0.87)
- PANDA: Towards Generalist Video Anomaly Detection via Agentic AI Engineer (NeurIPS 2025 · similarity 0.87)
- VERA: Explainable Video Anomaly Detection via Verbalized Learning of Vision-Language Models (CVPR 2025 · similarity 0.87)
- AD-FM: Multimodal LLMs for Anomaly Detection via Multi-Stage Reasoning and Fine-Grained Reward Optimization (AAAI 2026 · similarity 0.86)
- MoniTor: Exploiting Large Language Models with Instruction for Online Video Anomaly Detection (NeurIPS 2025 · similarity 0.86)
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.
