AdaToken-3D uses attention-derived contribution scores to prune spatial tokens layer by layer in 3D LMMs, achieving about 60 percent FLOPs reduction with roughly unchanged benchmark scores.
LL3DA: Visual Interactive Instruction Tuning for Omni-3D Understanding, Reasoning, and Planning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.GR 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
AdaToken-3D: Dynamic Spatial Gating for Efficient 3D Large Multimodal-Models Reasoning
AdaToken-3D uses attention-derived contribution scores to prune spatial tokens layer by layer in 3D LMMs, achieving about 60 percent FLOPs reduction with roughly unchanged benchmark scores.