Metraipolis › archive › dc_d2d0790df247

Intervenable Concept Dynamics for Video Activity Forecasting — arXiv:2611.00020

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.00020
License (as archived)
CC0-1.0
Archived
2026-11-03T18:15:48.357Z
SHA-256 of the archived text
42c08b84c5baa829dca030205f31ca63e49b4a60f05d92a8b924c410d99b878b
Stable URL of this page
https://metraipolis.com/documento/dc_d2d0790df247

How to cite

Intervenable Concept Dynamics for Video Activity Forecasting — arXiv:2611.00020 — arXiv (en), https://arxiv.org/abs/2611.00020 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d2d0790df247 (Metraipolis archive dc_d2d0790df247, sha256 42c08b84c5baa829dca030205f31ca63e49b4a60f05d92a8b924c410d99b878b)
BibTeX
@misc{dcd2d0790df247,
  title = {Intervenable Concept Dynamics for Video Activity Forecasting — arXiv:2611.00020},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.00020},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d2d0790df247; sha256 42c08b84c5baa829dca030205f31ca63e49b4a60f05d92a8b924c410d99b878b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.00020. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Concept bottleneck models route predictions through human-understandable concepts, making their intermediate states inspectable and editable. Extending this model family to forecasting requires modeling how concept activations evolve over time as objects, actions, and their interactions change, yet existing CBMs ground concepts only in observed inputs and do not forecast their future evolution. We study this problem using videos of ongoing activities and forecast future activities through a named-concept state. We propose TRACE-CBM (Temporal Relational Activity Concept Evolution), which first refines observed concept activations with sparse spatio-temporal graphs. It then recursively predicts future concept states, conditioning each transition on the prior activity distribution, and decodes both observed and predicted states with a shared activity classifier. At inference time, the model supports editing concept values, activity activations, and learned edges without parameter updates, exposing the relational spatio-temporal dependencies used for forecasting. Across multiple datasets, TRACE improves observed-activity Top-1 accuracy over concept-based methods by 1.2-8.5 percentage points and attains H3 forecast Top-1 accuracy within 1.9-2.2 points of matched black-box forecasters. Intervention experiments show concept-node edits correct 25.6-34.4% of initially incorrect H3 forecasts without flipping initially correct ones, and concept steering achieves the requested target activity in up to 90.0% of cases.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2611.00020 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-03T18:09:53.470Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).