Research map / Video / Action recognition
Progressive Cross-Modal Causal Intervention for Long-Term Action Recognition
Shaowu Xu, Xibin Jia, Chao Fan, Junyu Gao, Jing Chang, Qianmei Sun
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 |
|---|---|---|---|
| add ICH (OT-based back-door intervention for co-occurrence hallucination) to baseline (ablation) | PCMCI baseline · Breakfast | accuracy: 91.55 -> 94.93 | Improved |
| add ICI (relation-aware intervention) to baseline (ablation) | PCMCI baseline · Breakfast | accuracy: 91.55 -> 93.80 | Improved |
| add IVC (front-door adjustment) to baseline (ablation) | PCMCI baseline · Breakfast | mAP: 80.23 -> 88.31 | Improved |
| add IVC on top of ICH and ICI (ablation) | PCMCI · Breakfast | accuracy: 95.77 -> 97.46 | Improved |
| add IVC on top of ICH and ICI (ablation) | PCMCI · COIN | mAP: 82.25 -> 86.54 | Improved |
| ICH alone versus ICI alone (ablation) | PCMCI · COIN | accuracy: 92.56 vs 91.74 | Improved |
| apply IVC first instead of last (ablation) | PCMCI · Breakfast | accuracy: 97.46 -> 92.96 | Hurt |
| apply ICI before ICH (ablation) | PCMCI · Breakfast | mAP: 90.51 -> 87.43 | Hurt |
| use deconfounded text embeddings plus visual relational modeling as mediator instead of memory bank (ablation) | PCMCI IVC · COIN | accuracy: 90.60 -> 94.53 | Improved |
| add relational modeling in visual modality versus text-only mediator alignment (ablation) | PCMCI IVC · Breakfast | mAP: 88.75 -> 90.51 | Improved |
| use 8 relations in the relation-aware design (ablation) | PCMCI · Breakfast and Charades | accuracy and mAP | Improved |
| replace CLIP with stronger BLIP2 backbone (ablation) | PCMCI · COIN | accuracy: 94.53 -> 94.89 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- HierarQ — HierarQ: Task-Aware Hierarchical Q-Former for Enhanced Video Understanding (CVPR 2025) · better on 3 of 4 shared results
- MA-LMM — MA-LMM: Memory-Augmented Large Multimodal Model for Long-Term Video Understanding (CVPR 2024) · better on 3 of 3 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
- CofactVLA: Deconfounding Vision-Language-Action Models via Counterfactual Intervention (NeurIPS 2026 · similarity 0.88)
- Causal Tracing of Object Representations in Large Vision Language Models: Mechanistic Interpretability and Hallucination Mitigation (AAAI 2026 · similarity 0.87)
- Causal-HalBench: Uncovering LVLMs Object Hallucinations Through Causal Intervention (AAAI 2026 · similarity 0.85)
- Vision-and-Language Navigation via Causal Learning (CVPR 2024 · similarity 0.85)
- Act to See, See to Act: Diffusion-Driven Perception-Action Interplay for Adaptive Policies (NeurIPS 2025 · similarity 0.85)
- VA-AR: Learning Velocity-Aware Action Representations with Mixture of Window Attention (AAAI 2025 · similarity 0.85)
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.
