arXiv:2605.11212v4 Announce Type: replace Abstract: Computer-use agents (CUAs) rely on visual observations of graphical user interfaces, where each screenshot is encoded into a large number of visual...
The paper introduces Role-Conditioned Sub-Token Routing (RoleSub), a method that compresses the value representations of retained tokens in Vision‑Language‑Action models. By partitioning each token’s value into orthogonal groups and routing them based on token features, a latent role, and language context, RoleSub can also compress language values. Experiments on OpenVLA‑OFT‑7B show that, at matched visual‑KV budgets, RoleSub outperforms token‑only control in most settings and reduces total KV to 9.2–11.3% of the original while maintaining strong control performance.
The paper introduces Role-Conditioned Sub-Token Routing (RoleSub), a method that compresses the value representations of retained tokens in Vision‑Language‑Action models. By partitioning each token’s value into orthogonal groups and routing them based on token features, latent roles, and language context, RoleSub can compress both visual and language representations without discarding tokens. Experiments on OpenVLA‑OFT‑7B show that, at matched visual‑KV budgets, RoleSub outperforms token‑only control in most settings and can reduce total KV to 9.2–11.3% of the original while maintaining strong control performance.
VisionWeave introduces elastic visual representation weaving, a native capability for multimodal large language models that learns where and at what granularity to encode visual information. The method combines a gated spatial pooler for coarse representations with a granularity router that allocates content‑adaptive token usage, trained end‑to‑end on large‑scale data. Experiments on Qwen3.5‑4B and Qwen3.8‑27B show that VisionWeave can save 43.0% of tokens while preserving 98.9% of performance across eight benchmarks, and delivers significant throughput gains and latency reductions when deployed on the SGLang serving engine.
The paper "ActionCodec: What Makes for Good Action Tokenizers" addresses the lack of design guidance for action tokenization in Vision‑Language‑Action (VLA) models. By deriving information‑theoretic principles—maximizing temporal token overlap, reducing vocabulary redundancy, boosting multimodal mutual information, and ensuring token independence—the authors propose ActionCodec, a tokenizer that improves training efficiency and VLA performance. Experiments show that a SmolVLM2‑2.2B fine‑tuned with ActionCodec reaches a 95.5 % success rate on LIBERO, rising to 97.4 % with architectural tweaks, setting a new state‑of‑the‑art for VLA models without robotics pre‑training.
CatchUp is a training‑free framework that improves efficiency in large vision‑language models by depth‑aware re‑admission of visual tokens. It repeatedly evaluates query‑conditioned relevance, processes only selected tokens per decoder interval, and preserves deferred tokens for later re‑admission, using reference‑to‑current transition signals to approximate missed representation evolution. Experiments on three referring‑expression grounding datasets and six multimodal benchmarks show that CatchUp retains over 92% of full‑model performance with 88.9% fewer active tokens and reduces end‑to‑end latency by 8.0% at a 64‑token budget.
arXiv:2609.15131v1 Announce Type: cross Abstract: Multimodal large language models (MLLMs) require substantial computation to process numerous visual tokens across all transformer layers. Most method...
arXiv:2606.06158v2 Announce Type: replace Abstract: Adaptive video tokenisation seeks to dynamically allocate token budgets based on the underlying visual complexity of a sequence. Current continuous...
arXiv:2608. 03244v1 Announce Type: new Abstract: Image-goal visual navigation is a fundamental capability for embodied agents.
arXiv:2609.38984v1 Announce Type: cross Abstract: World-action models (WAMs) leverage pretrained video models to improve generalization in robot control by jointly predicting future visual states and...
The paper introduces CF‑Prune, a training‑free method that prunes future‑frame tokens in pretrained World Action Models (WAMs) to reduce inference latency. By using instruction and action attention during the initial denoising pass, the framework identifies task‑relevant tokens, while an entropy‑based phase identification keeps dense computation in precision phases to avoid performance loss. Experiments on RoboLab, RoboCasa, and real‑world manipulation tasks show consistent speedups—up to 1.33× at 50% pruning—while maintaining task success rates within 1 percentage point of the baseline.
arXiv:2609.10297v1 Announce Type: new Abstract: GUI agents accumulate high-resolution screenshots as the trajectory unfolds, increasing inference latency and memory usage. Training-free visual token...