Metraipolis › archive › dc_9eacea48e59b
Decoding Geometry, Not Tokens: Efficient VLMs for Robot Affordance and Trajectory Prediction — arXiv:2610.39416
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.39416
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-01T02:43:47.660Z
- SHA-256 of the archived text
e88f989df3001f290c6b69bad7f9a336593a682fd992dbb25f3a9397a5c186a4- Stable URL of this page
https://metraipolis.com/documento/dc_9eacea48e59b
How to cite
Decoding Geometry, Not Tokens: Efficient VLMs for Robot Affordance and Trajectory Prediction — arXiv:2610.39416 — arXiv (en), https://arxiv.org/abs/2610.39416 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_9eacea48e59b (Metraipolis archive dc_9eacea48e59b, sha256 e88f989df3001f290c6b69bad7f9a336593a682fd992dbb25f3a9397a5c186a4)
BibTeX
@misc{dc9eacea48e59b,
title = {Decoding Geometry, Not Tokens: Efficient VLMs for Robot Affordance and Trajectory Prediction — arXiv:2610.39416},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.39416},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_9eacea48e59b; sha256 e88f989df3001f290c6b69bad7f9a336593a682fd992dbb25f3a9397a5c186a4},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.39416. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Predicting where to grasp and how to move is fundamentally a geometry problem. Vision-language models treat it as language generation instead. They tokenize spatial coordinates into text and decode them autoregressively. This mechanism ignores spatial structure and truncates continuous precision. Errors compound across coordinate tokens during decoding. In this paper, we show that compact off-the-shelf VLMs with 0.5-2B parameters already encode rich spatial information in their intermediate hidden states. Autoregressive decoding is the dominant bottleneck for trajectory prediction, while affordance prediction depends additionally on feature depth and decoder architecture. We expose the latent spatial information with a lightweight, geometry-compatible prediction head that attaches to a frozen backbone. It reads features from an intermediate hidden layer, conditions a learned query on the task instruction and refines it via cross-attention over the visual tokens to regress affordance boxes and trajectory waypoints directly as continuous values, without modifying the backbone. We evaluate on ShareRobot-Bench and AGD20K. Our 2B variant reduces trajectory error by over 60% and improves ShareRobot-Bench affordance mAP by 85% relative to the autoregressive baseline. It approaches RoboBrain on trajectory prediction (0.251 vs 0.191 DFD) at 3.8x higher throughput. These results demonstrate a favorable accuracy-efficiency trade-off for instruction-conditioned 2D geometric prediction. Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno). Categorie arXiv: Robotics (cs.RO) ; Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2610.39416 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-11-01T02:16:44.008Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).