mirror of
https://github.com/hanzoai/papers.git
synced 2026-10-29 07:34:31 +00:00
8 lines
432 B
TeX
8 lines
432 B
TeX
\section{Discussion}
|
|
|
|
|
|
\paragraph{Limitations.} Training requires large-scale paired multimodal data. The modality extension mechanism, while flexible, requires fine-tuning for optimal performance on new modalities.
|
|
|
|
\paragraph{Future Work.} We plan to explore: (1) online modality extension without any fine-tuning, (2) integration with world models for embodied AI, and (3) scaling to additional modalities (tactile, olfactory).
|
|
|