Research map / Video / Action recognition
Open-Vocabulary Long Term Action Anticipation
Syed Talal Wasim, Jinhui Yi, Hamid Suleman, Ahmad Javed, Yanan Luo, Muhammad Muzammal Naseer, Juergen Gall
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 |
|---|---|---|---|
| replace multi-head decoder with MHCAD cross-attention decoder (ablation) | LaViLa · EK100 (open-vocab, trained on Ego4D) | verb edit distance: 0.849 -> 0.728 | Improved |
| replace multi-head decoder with MHCAD (ablation) | LaViLa · Ego4D v2 test | verb/noun edit distance: 0.720/0.727 -> 0.689/0.695 | Improved |
| add causal self-attention among horizon queries (ablation) | MHCAD · EK100 (open-vocab, trained on Ego4D) | verb edit distance: 0.805 -> 0.728 | Improved |
| use causal instead of bidirectional self-attention (ablation) | MHCAD · EK100 (open-vocab, trained on Ego4D) | verb edit distance: 0.821 -> 0.728 | Improved |
| order self-attention before cross-attention (ablation) | MHCAD · EK100 (open-vocab, trained on Ego4D) | verb edit distance: 0.769 -> 0.728 | Improved |
| use MHCAD instead of FUTR decoder (ablation) | LaViLa encoder · EK100 (open-vocab, trained on Ego4D) | verb edit distance: 0.836 -> 0.728 | Improved |
| use 4D rotary embeddings instead of absolute position embedding (ablation) | MHCAD · EK100 (open-vocab, trained on Ego4D) | verb edit distance: 0.783 -> 0.728 | Improved |
| use 4D rotary embeddings instead of DPE (ablation) | MHCAD · EK100 (open-vocab, trained on Ego4D) | verb edit distance: 0.746 -> 0.728 | Improved |
| increase decoder depth from 0 to 2 layers (ablation) | MHCAD · Ego4D v2 test | verb/noun edit distance: 0.720 -> 0.689 verb, 0.727 -> 0.695 noun | Improved |
| add decoder layers beyond 2 (ablation) | MHCAD · all datasets | edit distance | Hurt |
| use adapter only instead of adapter plus prompts (ablation) | text encoder · EK100 (open-vocab, trained on Ego4D) | verb edit distance: 0.734 -> 0.728 | Improved |
| use adapter only instead of prompts only (ablation) | text encoder · EK100 (open-vocab, trained on Ego4D) | verb edit distance: 0.736 -> 0.728 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 2 on most shared results.
- AntGPT — AntGPT: Can Large Language Models Help Long-term Action Anticipation from Videos? (ICLR 2024) · better on 2 of 4 shared results
- PALM — PALM: Predicting Actions through Language Models (ECCV 2024) · better on 0 of 2 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
- AntGPT: Can Large Language Models Help Long-term Action Anticipation from Videos? (ICLR 2024 · similarity 0.90)
- Can't Make an Omelette Without Breaking Some Eggs: Plausible Action Anticipation Using Large Video-Language Models (CVPR 2024 · similarity 0.90)
- Uncertainty-aware Action Decoupling Transformer for Action Anticipation (CVPR 2024 · similarity 0.89)
- Towards Understanding Future: Consistency Guided Probabilistic Modeling for Action Anticipation (AAAI 2024 · similarity 0.87)
- ATI-VLA: Action-Centric Predictive Vision–Language–Action Models via Actionable Alignment Then Adaptive Injection (NeurIPS 2026 · similarity 0.87)
- Summarize the Past to Predict the Future: Natural Language Descriptions of Context Boost Multimodal Object Interaction Anticipation (CVPR 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.
