Mind the Heads: Topological Representation Alignment for Multimodal LLMs
HeRA aligns individual attention heads in MLLMs to preserve local neighborhood relationships across modalities, improving vision-centric task performance and reducing visual halluc…
Hugging Face · Daily Papers
·Davide Caffagni, Alberto Compagnoni
·
·▲ 1 upvotes
Este artigo está em destaque na seleção diária de papers do Hugging Face, curada pela comunidade de pesquisa em IA.
Autores: Davide Caffagni, Alberto Compagnoni, Federico Melis, Sara Sarto, Pier Luigi Dovesi, Mark Granroth-Wilding
- 1 upvotes da comunidade
- Temas: Multimodal Large Language Models, representation alignment, Transformer models, attention heads, Platonic Representation Hypothesis, Mutual K-Nearest Neighbor
Resumo
Resumo original (em inglês), extraído do paper:
HeRA aligns individual attention heads in MLLMs to preserve local neighborhood relationships across modalities, improving vision-centric task performance and reducing visual hallucinations.
