Metraipolis › archive › dc_4da7132da057

Why MLLMs Struggle to Count: Overcoming Individuation and Aggregation Bottlenecks with ConvStack — arXiv:2609.37096

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.37096
License (as archived)
CC0-1.0
Archived
2026-10-07T18:40:56.495Z
SHA-256 of the archived text
910476c2b696fd0907b23b3e324a303ec6fe8f6fd490d95daef10729fa607d76
Stable URL of this page
https://metraipolis.com/documento/dc_4da7132da057

How to cite

Why MLLMs Struggle to Count: Overcoming Individuation and Aggregation Bottlenecks with ConvStack — arXiv:2609.37096 — arXiv (en), https://arxiv.org/abs/2609.37096 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4da7132da057 (Metraipolis archive dc_4da7132da057, sha256 910476c2b696fd0907b23b3e324a303ec6fe8f6fd490d95daef10729fa607d76)
BibTeX
@misc{dc4da7132da057,
  title = {Why MLLMs Struggle to Count: Overcoming Individuation and Aggregation Bottlenecks with ConvStack — arXiv:2609.37096},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.37096},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4da7132da057; sha256 910476c2b696fd0907b23b3e324a303ec6fe8f6fd490d95daef10729fa607d76},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.37096. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Multimodal Large Language Models (MLLMs) consistently struggle with fine-grained visual counting, yet the underlying causes remain poorly understood. In this work, we present a mechanistic analysis of this failure mode, identifying two critical bottlenecks inherent to the global attention pipeline of MLLMs. First, we reveal an individuation bottleneck stemming from image patchification: because Vision Transformers process patches independently, they struggle to group fragmented geometric features across boundaries into distinct object representations. Second, we identify a collapse in the subsequent counting aggregation process, where representation separation rapidly diminishes as numerosity increases due to attention compression. Identifying and formalizing these twin bottlenecks constitutes our first major contribution. To overcome them, we propose ConvStack, a lightweight architecture that operates directly in the visual token space to explicitly aggregate and inject local spatial structures via zero-initialized residual connections. By explicitly addressing the individuation bottleneck, ConvStack provides unambiguous geometric evidence for downstream aggregation. Remarkably, by fine-tuning exclusively on counting tasks, the model achieves substantial improvements in dense object counting and broader spatial understanding benchmarks, without compromising on general visual capabilities.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-30 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.37096 — 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 2026-10-07T18:04:24.820Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).