aiqbench analysis
This research addresses a critical gap in language-conditioned robotics by introducing CompCPZ, which prevents agents from collapsing disjunctive instructions (e.g., "A or B") into ambiguous geometric centroids. By leveraging polynomial zonotope enclosures along the parse tree, the method preserves distinct feasible modes without sacrificing sub-millisecond runtime efficiency. The approach demonstrates robust transferability to real-world quadruped trials, suggesting that future robot policies must evaluate intent preservation rather than just target reach accuracy.
CompCPZ: Preserving Multi-Modal Intent in Language-Guided Robot Manipulation
arXiv:2608.17717v1 Announce Type: new Abstract: A robot asked to "place the cup near the red plate or the blue plate" may reach the centroid between them and appear geometrically successful, while satisfying neither disjunct of the instruction. This silent semantic failure exposes a structural limitation of language-conditioned robot policies: representations that collapse a disjunctive instruction into a single connected set cannot preserve all feasible modes, and planners that commit to one action degrade under run-time mode uncertainty. We address this limitation with CompCPZ, a sound algebraic layer that language-conditioned learning systems wrap to recover multi-modal disjunctive representation, recursively composing per-primitive constrained polynomial zonotope enclosures along the language parse tree with distribution-free conformal coverage and sub-millisecond runtime. On a closed-loop ManiSkill3 tabletop-manipulation benchmark, CompCPZ outperforms convex set baselines, multi-peak decoders, and a zero-shot vision-language-action model (1,900/1,918 paired wins, p << 10^(-30)); the same compiler also transfers without retuning to planar real-robot trials on a Unitree Go2 quadruped under motion capture. These results suggest that compositional language grounding should be evaluated not only by reaching a decoded target, but by whether the represented feasibility set preserves the connected-component structure of the user's intent.
Abstract:A robot asked to "place the cup near the red plate or the blue plate" may reach the centroid between them and appear geometrically successful, while satisfying neither disjunct of the instruction. This silent semantic failure exposes a structural limitation of language-conditioned robot policies: representations that collapse a disjunctive instruction into a single connected set cannot preserve all feasible modes, and planners that commit to one action degrade under run-time mode uncertainty. We address this limitation with CompCPZ, a sound algebraic layer that language-conditioned learning systems wrap to recover multi-modal disjunctive representation, recursively composing per-primitive constrained polynomial zonotope enclosures along the language parse tree with distribution-free conformal coverage and sub-millisecond runtime. On a closed-loop ManiSkill3 tabletop-manipulation benchmark, CompCPZ outperforms convex set baselines, multi-peak decoders, and a zero-shot vision-language-action model (1,900/1,918 paired wins, p << 10^(-30)); the same compiler also transfers without retuning to planar real-robot trials on a Unitree Go2 quadruped under motion capture. These results suggest that compositional language grounding should be evaluated not only by reaching a decoded target, but by whether the represented feasibility set preserves the connected-component structure of the user's intent.
Robotics (cs.RO)
arXiv:2608.17717 [cs.RO]
(or arXiv:2608.17717v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2608.17717 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 279 words
Related coverage
Links #5: 2026/07
LessWrong · Jul 31, 2026
Korea University Team Presents LLM-Qubo Framework for Multi-Drone Assignment Optimisation
Quantum Zeitgeist · Jun 17, 2026
[AINews] not much happened today
Latent Space · Sep 10, 2026
EmoPose: Vision-Language Model Guided Emotion-Aware Gesture Generation for Humanoid Robots
arXiv · Sep 22, 2026