Paper Detail
Pranav Mantini, Shishir K. Shah
We address the challenge of knowledge composition in Vision-Language Models (VLMs), where accumulating expertise across multiple domains or tasks typically leads to catastrophic forgetting. We introduce GeoStack (Geometric Stacking), a modular framework that allows independently trained domain experts to be composed into a unified model. By imposing geometric and structural constraints on the adapter manifold, GeoStack ensures the foundational knowledge of the base model is preserved. Furthermore, we mathematically demonstrate a weight-folding property that achieves constant-time inference complexity (O(1)), regardless of the number of integrated experts. Experimental results across multi-domain adaptation and class-incremental learning show that GeoStack provides an efficient mechanism for long-term knowledge composition while significantly mitigating catastrophic forgetting. Code is available at https://github.com/QuantitativeImagingLaboratory/GeoStack.
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@misc{mantini2026geostack,
title = {GeoStack: A Framework for Quasi-Abelian Knowledge Composition in VLMs},
author = {Pranav Mantini and Shishir K. Shah},
year = {2026},
abstract = {We address the challenge of knowledge composition in Vision-Language Models (VLMs), where accumulating expertise across multiple domains or tasks typically leads to catastrophic forgetting. We introduce GeoStack (Geometric Stacking), a modular framework that allows independently trained domain experts to be composed into a unified model. By imposing geometric and structural constraints on the adapter manifold, GeoStack ensures the foundational knowledge of the base model is preserved. Furthermor},
url = {https://huggingface.co/papers/2605.06477},
keywords = {Vision-Language Models, catastrophic forgetting, domain experts, adapter manifold, weight-folding property, multi-domain adaptation, class-incremental learning, code available, huggingface daily},
eprint = {2605.06477},
archiveprefix = {arXiv},
}
{}