Research map / Continual learning / Domain and prompt-based
PIVOT: Prompting for Video Continual Learning
Andrés Villa, Juan León Alcázar, Motasem Alfarra, Kumail Alhamoud, Julio Hurtado, Fabian Caba Heilbron, Alvaro Soto, Bernard Ghanem
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 spatial and temporal task prompts (ablation) | PIVOT w/o prompts · Kinetics, 20 tasks | Acc: 57.51% -> 55.04% | Hurt |
| add spatial and temporal task prompts (ablation) | PIVOT w/o prompts · ActivityNet-Trim, 20 tasks | Acc: 73.43% -> 73.84% | Improved |
| add spatial and temporal task prompts (ablation) | PIVOT w/o prompts · UCF101, 10 tasks | Acc: 94.80% -> 93.36% | Hurt |
| add spatial prompts with a linear classifier (ablation) | frozen CLIP · UCF101, 10 tasks | Acc: 65.73% -> 77.66% | Improved |
| add spatial prompts with a linear classifier (ablation) | frozen CLIP · ActivityNet-Trim, 20 tasks | Acc: 64.13% -> 57.89% | Hurt |
| add a replay memory buffer to spatial prompting (ablation) | CLIP + Spatial Prompting · UCF101, 10 tasks | Acc: 77.66% -> 88.42% | Improved |
| use multimodal contrastive classification (ablation) | CLIP + Memory Buffer · ActivityNet-Trim, 10 tasks | Acc: 63.33% -> 69.53% | Improved |
| add a temporal transformer encoder (ablation) | CLIP + MCL · UCF101, 10 tasks | BWF: 10.65% -> 5.95% | Improved |
| set prompt length to 3 rather than 1 (ablation) | PIVOT · ActivityNet validation | Acc: 73.2% -> 73.8% | Improved |
| use 2 prompts per task instead of 1 (ablation) | PIVOT · ActivityNet validation | Acc: 73.8% -> 72.60% | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- iCaRL — iCaRL: Incremental Classifier and Representation Learning (CVPR 2017) · 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
- vCLIMB: A Novel Video Class Incremental Learning Benchmark (CVPR 2022 · similarity 0.88)
- CRAM: Large Scale Video Continual Learning with Bootstrapped Compression (ICCV 2025 · similarity 0.88)
- CRNet: A Fast Continual Learning Framework With Random Theory (TPAMI 2023 · similarity 0.88)
- Learning To Prompt for Continual Learning (CVPR 2022 · similarity 0.88)
- Convolutional Prompting meets Language Models for Continual Learning (CVPR 2024 · similarity 0.87)
- REP: Resource-Efficient Prompting for Rehearsal-Free Continual Learning (NeurIPS 2025 · similarity 0.87)
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.
