A pretrained LLM is adapted via LoRA fine-tuning into a content-adaptive compressor that maps long texts to compact variable-length Z-token sequences while preserving reconstruction quality and downstream performance.
Gomez, Lukasz Kaiser, and Illia Polosukhin
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
verdicts
UNVERDICTED 3representative citing papers
Gaze regularization aligns VLA attention with human visual patterns via KL divergence on patch distributions, yielding 4-12% gains on manipulation benchmarks.
Co-Me distills a confidence predictor to selectively merge low-confidence tokens in visual geometric transformers, delivering up to 21.5x speedup on VGGT and 20.4x on Pi3 while preserving spatial coverage and performance.
citing papers explorer
-
Large Language Model as Token Compressor and Decompressor
A pretrained LLM is adapted via LoRA fine-tuning into a content-adaptive compressor that maps long texts to compact variable-length Z-token sequences while preserving reconstruction quality and downstream performance.
-
Gaze-Regularized Vision-Language-Action Models for Robotic Manipulation
Gaze regularization aligns VLA attention with human visual patterns via KL divergence on patch distributions, yielding 4-12% gains on manipulation benchmarks.
-
Co-Me: Confidence-Guided Token Merging for Visual Geometric Transformers
Co-Me distills a confidence predictor to selectively merge low-confidence tokens in visual geometric transformers, delivering up to 21.5x speedup on VGGT and 20.4x on Pi3 while preserving spatial coverage and performance.