Research map / Continual learning / Online and replay-based
CE$^4$L: Continual Ego, Exo, and Ego-Exo Learning
Hongwei Yan, Kanglei Zhou, Yuchen Liu, Qingyu Shi, Yi Zhong, Liyuan Wang
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 task adapters with whitened-subspace router (ablation) | fine-tuning · skill assessment (RAAN) | ranking accuracy: 72.94 -> 80.38 | Improved |
| replace random router with whitened-subspace router (ablation) | VISTA task adapters · skill assessment (RAAN) | ranking accuracy: 69.41 -> 80.38 | Improved |
| add task adapters with random router (ablation) | fine-tuning · skill assessment (RAAN) | ranking accuracy: 72.94 -> 69.41 | Hurt |
| replace whitened-subspace router with oracle router (ablation) | VISTA · skill assessment (RAAN) | ranking accuracy: 80.38 -> 80.42 | No clear effect |
| add whitening to router (ablation) | cosine router · task-router accuracy | accuracy | Improved |
| use whitened-subspace residual router over kNN/cosine (ablation) | VISTA · task-router accuracy | accuracy | Improved |
| increase subspace dimension q (ablation) | VISTA · action anticipation | Top-5 recall | No clear effect |
| mix too many adapters (large top-K) (ablation) | VISTA · action anticipation | Top-5 recall | Hurt |
| add Ego+Exo co-training | joint training · action segmentation cross-view | frame-wise accuracy: Exo eval 22.66 -> 38.26 (Ego-train vs co-train) | Improved |
| add VISTA to fine-tuning | fine-tuning · action segmentation Ego+Exo, eval Exo | frame-wise accuracy: 20.32 -> 31.41 | Improved |
| use VISTA vs replay on anticipation | ER · action anticipation Ego+Exo Ego-N | Top-5 recall: 25.29 -> 25.23 | No clear effect |
| use L2P+ and S-Prompt+ routing | fine-tuning · action anticipation noun prediction | Top-5 recall: Ego-N 19.26 -> 16.19 (L2P+) | Hurt |
Closest papers by meaning
- Bootstrap Your Own Views: Masked Ego-Exo Modeling for Fine-grained View-invariant Video Representations (CVPR 2025 · similarity 0.88)
- EgoExoBench: A Benchmark for First- and Third-person View Video Understanding in MLLMs (NeurIPS 2025 · similarity 0.88)
- Natural Image Pretraining Improves Abstract Reasoning (ECCV 2026 · similarity 0.88)
- EgoExo-Con: Exploring View-Invariant Video Temporal Understanding (ECCV 2026 · similarity 0.88)
- Test-time Ego-Exo-centric Adaptation for Action Anticipation via Multi-Label Prototype Growing and Dual-Clue Consistency (CVPR 2026 · similarity 0.88)
- Online Continual Learning for Interactive Instruction Following Agents (ICLR 2024 · 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.
