G²TR reduces visual tokens and prefill compute by 1.94x in separate-encoder UMMs via generation-guided importance from VAE latent consistency, balanced selection, and merging, while preserving reasoning accuracy and editing quality.
An image is worth 1/2 tokens after layer 2: Plug-and-play inference acceleration for large vision-language models
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
citation-role summary
baseline 1
citation-polarity summary
fields
cs.CV 3years
2026 3roles
baseline 1polarities
baseline 1representative citing papers
FIS-DiT achieves 2.11-2.41x speedup on video DiT models in few-step regimes with negligible quality loss by exploiting frame-wise sparsity and consistency through a training-free interleaved execution strategy.
SEATS adaptively selects and removes non-text tokens before and inside the LLM layers of omni-modal models, yielding 9.3x FLOPs reduction and 4.8x prefill speedup at 10% token retention while keeping 96.3% performance.
citing papers explorer
-
G$^2$TR: Generation-Guided Visual Token Reduction for Separate-Encoder Unified Multimodal Models
G²TR reduces visual tokens and prefill compute by 1.94x in separate-encoder UMMs via generation-guided importance from VAE latent consistency, balanced selection, and merging, while preserving reasoning accuracy and editing quality.
-
FIS-DiT: Breaking the Few-Step Video Inference Barrier via Training-Free Frame Interleaved Sparsity
FIS-DiT achieves 2.11-2.41x speedup on video DiT models in few-step regimes with negligible quality loss by exploiting frame-wise sparsity and consistency through a training-free interleaved execution strategy.
-
Stage-adaptive Token Selection for Efficient Omni-modal LLMs
SEATS adaptively selects and removes non-text tokens before and inside the LLM layers of omni-modal models, yielding 9.3x FLOPs reduction and 4.8x prefill speedup at 10% token retention while keeping 96.3% performance.