Metraipolis › archive › dc_52cd562a30dd

LSA-PAST: Layer-Specific Adaptation and Prototype Transport for Compositional Incremental Learning — arXiv:2610.27203

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.27203
License (as archived): CC0-1.0
Archived: 2026-10-21T07:06:13.295Z
SHA-256 of the archived text: 78b665b6256ffb837dcdfd8ae36ab717843bf388e39a1b60f7de86c8b98d08c9
Stable URL of this page: https://metraipolis.com/documento/dc_52cd562a30dd

How to cite

LSA-PAST: Layer-Specific Adaptation and Prototype Transport for Compositional Incremental Learning — arXiv:2610.27203 — arXiv (en), https://arxiv.org/abs/2610.27203 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_52cd562a30dd (Metraipolis archive dc_52cd562a30dd, sha256 78b665b6256ffb837dcdfd8ae36ab717843bf388e39a1b60f7de86c8b98d08c9)

BibTeX

@misc{dc52cd562a30dd,
  title = {LSA-PAST: Layer-Specific Adaptation and Prototype Transport for Compositional Incremental Learning — arXiv:2610.27203},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.27203},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_52cd562a30dd; sha256 78b665b6256ffb837dcdfd8ae36ab717843bf388e39a1b60f7de86c8b98d08c9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.27203. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Class-Incremental Learning (CIL) enables models to recognize new classes over time while retaining prior knowledge, but many real-world problems are compositional, where visual concepts arise from combinations of primitives such as states (i.e., red, shiny) and objects (i.e., shoe, car). Standard CIL methods treat each composition as an independent class, ignoring shared primitives and providing no mechanism to stabilize their representations as new compositions arrive. This causes parameter updates to entangle primitives and induce drift that degrades performance on earlier compositions. Compositional Incremental Learning (composition-IL) formalizes this setting by introducing new state-object compositions across tasks. In this paper, we propose Layer-Specific Adaptation with Prototype-Anchored Adapter-Space Transport (LSA-PAST), a parameter-efficient method tailored to composition-IL. LSA integrates Low-Rank Adaptation (LoRA) into a frozen ViT with depth-specialized adapters for states, objects, and compositions. PAST then maps past, anchored prototypes into the current adapter space through a rehearsal-free affine transform fit on current-task data, enabling old prototypes to remain compatible with new adapters. Extensive experiments demonstrate that LSA-PAST achieves state-of-the-art (SOA) accuracy-forgetting trade-off performance for composition-IL.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.27203 — 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 2026-10-21T07:05:43.386Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).