Research map / Video / Video-language understanding
AD-MIR: Bridging the Gap from Perception to Persuasion in Advertising Video Understanding via Structured Reasoning
Binxiao Xu, Junyu Feng, Xiaopeng Lin, Haodong Li, ZhiYuan Feng, Bohan Zeng, Ruichuan An, Ming Lu, Qi She, Wentao Zhang
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 |
|---|---|---|---|
| wrap o1 in AD-MIR agent framework | o1 · AdsQA | accuracy: 33.1 -> 38.1 strict; 52.4 -> 60.0 relaxed | Improved |
| wrap Qwen2.5-VL-7B in AD-MIR agent framework | Qwen2.5-VL-7B · AdsQA | accuracy: +4.5 strict; +3.1 relaxed | Improved |
| use AD-MIR instead of DVD agent | o1 backbone · AdsQA | accuracy: +1.8 strict; +9.5 relaxed | Improved |
| training-free AD-MIR instead of RL-trained ReAd-R | Qwen2.5-VL-7B · AdsQA | accuracy: +5.7 strict; +0.6 relaxed | Improved |
| increase lexical weight beta in hybrid retrieval (ablation) | AD-MIR (Qwen2.5-VL-7B) · AdsQA | accuracy | Improved |
| raise max reasoning steps Tmax beyond 8 (ablation) | AD-MIR (Qwen2.5-VL-7B) · AdsQA | accuracy | No clear effect |
| add Hybrid Indexing (ablation) | AD-MIR (Qwen2.5-VL-7B) · AdsQA | accuracy | Improved |
| add Subject Registry (ablation) | AD-MIR (Qwen2.5-VL-7B) · AdsQA | accuracy | Improved |
| add Communication Expert (ablation) | AD-MIR (Qwen2.5-VL-7B) · AdsQA persuasion dimensions | accuracy | Improved |
| add Visual Anchor verification (ablation) | AD-MIR (Qwen2.5-VL-7B) · AdsQA | accuracy | Improved |
| use AD-MIR agent loop instead of DVD (cost) | o1 backbone · AdsQA | latency / tokens: 45.5 -> 93.1 s latency; 55.9k -> 174.6k tokens | Hurt |
| hybrid Clip Search instead of DVD dense top-k retrieval | AD-MIR · AdsQA | latency: 1.2 -> 0.3 s | Improved |
Closest papers by meaning
- BLM-Guard: Explainable Multimodal Ad Moderation with Chain-of-Thought and Policy-Aligned Rewards (AAAI 2026 · similarity 0.84)
- GeoReward: Mitigating Contextual Variable Overestimation in Vision-Language Models for Cross-Market Preference Prediction (ICML 2026 · similarity 0.84)
- A3: Towards Advertising Aesthetic Assessment (CVPR 2026 · similarity 0.84)
- When Thinking Drifts: Evidential Grounding for Robust Video Reasoning (NeurIPS 2025 · similarity 0.82)
- Retrieving Counterfactuals Improves Visual In-Context Learning (CVPR 2026 · similarity 0.82)
- Mimic Human Cognition, Master Multi-Image Reasoning: A Meta-Action Framework for Enhanced Visual Understanding (CVPR 2026 · similarity 0.82)
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.
