Bidirectional Action Sequence Learning for Long-term Action Anticipation with Large Language Models
Not on the map. Why?
Abstract (arXiv)
Video-based long-term action anticipation is crucial for early risk detection in areas such as automated driving and robotics. Conventional approaches extract features from past actions using encoders and predict future events with decoders, which limits performance due to their unidirectional nature. These methods struggle to capture semantically distinct sub-actions within a scene. The proposed method, BiAnt, addresses this limitation by combining forward prediction with backward prediction using a large language model. Experimental results on Ego4D demonstrate that BiAnt improves performance in terms of edit distance compared to baseline methods.
Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 28 of its 31 references that are in the map. First posted on arXiv 1 Aug 2025.
Builds on
28 of 31 references.
References · 28
- 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
- 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,372
- 2019DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lightercited by 4,307
- 2019ALBERT: A Lite BERT for Self-supervised Learning of Language Representationscited by 2,790
- 2018SlowFast Networks for Video Recognitioncited by 1,867
- 2021Ego4D: Around the World in 3,000 Hours of Egocentric Videocited by 1,556
- 2021Multiscale Vision Transformerscited by 897
- 2023EgoSchema: A Diagnostic Benchmark for Very Long-form Video Language Understandingcited by 680
- 2023Ego-Exo4D: Understanding Skilled Human Activity from First- and Third-Person Perspectivescited by 490
- 2019MASS: Masked Sequence to Sequence Pre-training for Language Generationcited by 347
- 2020Understanding Human Hands in Contact at Internet Scalecited by 281
- 2022Learning Video Representations from Large Language Modelscited by 205
16 more
- 2023VideoLLM: Modeling Video Sequence with Large Language Modelscited by 116
- 2024EgoExoLearn: A Dataset for Bridging Asynchronous Ego- and Exo-centric View of Procedural Activities in Real Worldcited by 95
- 2022Ego-Body Pose Estimation via Ego-Head Pose Estimationcited by 94
- 2022My View is the Best View: Procedure Learning from Egocentric Videoscited by 81
- 2023AntGPT: Can Large Language Models Help Long-term Action Anticipation from Videos?cited by 81
- 2023AssemblyHands: Towards Egocentric Activity Understanding via 3D Hand Pose Estimationcited by 64
- 2022Future Transformer for Long-term Action Anticipationcited by 58
- 2023EgoObjects: A Large-Scale Egocentric Dataset for Fine-Grained Object Understandingcited by 40
- 2022Rethinking Learning Approaches for Long-Term Action Anticipationcited by 29
- 2024EgoGen: An Egocentric Synthetic Data Generatorcited by 27
- 2023Object-centric Video Representation for Long-term Action Anticipationcited by 24
- 2023Meet in the Middle: A New Pre-training Paradigmcited by 22
- 2023RefEgo: Referring Expression Comprehension Dataset from First-Person Perception of Ego4Dcited by 21
- 2022Video + CLIP Baseline for Ego4D Long-term Action Anticipationcited by 18
- 2023Palm: Predicting Actions through Language Models @ Ego4D Long-Term Action Anticipation Challenge 2023cited by 16
- 2023Technical Report for Ego4D Long Term Action Anticipation Challenge 2023cited by 7
Built on by
Citing papers per year: 2026: 2.
2 papers.