Looped Transformers provide advantages in parameter efficiency, computational capabilities, and generalization for reasoning tasks. However, their expressive power regarding function approximation remains underexplored. In this paper, we establish the approximation rate of Looped Transformers by defining the modulus of continuity for sequence-to-sequence functions. This reveals a limitation specific to the looped architecture. That is, the analysis prompts the incorporation of scaling parameters for each loop, conditioned on timestep encoding. Experiments validate the theoretical results, showing that increasing the number of loops enhances performance, with further gains achieved through the timestep encoding.

We present a low-cost, soft tactile sensor using common, easily sourced materials that can be integrated with existing robotic gripper systems without requiring complex fabrication techniques or expensive components. Our approach includes two designs: a flexible linear sensor constructed from a rubber tube and a planar sensor made with a rubber membrane stretched over an enclosure. Both sensors contain an embedded speaker and microphones that leverage active acoustic sensing to map the unique acoustic resonant response of the cavity’s structure to deformations that occur when the robotic gripper is grasping an object. Experimental results demonstrate that, using a support vector machine, the linear sensor achieves contact point estimation with an RMSE of 6 mm, while the planar sensor achieves an RMSE of 0.57–0.62 mm. Additionally, the planar sensor classifies six objects with an accuracy of 97.7%. These results demonstrate the potential for active acoustics to be an accessible method for enabling tactile sensing capabilities for robotic systems.

Kevin Xu, Yeganeh Kordi, Tanay Nayak, Adi Asija, Yizhong Wang, Kate Sanders, Adam Byerly, Jingyu Zhang, Benjamin Van Durme, Daniel Khashabi. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

A technique for synthesizing iterators from declarative abstraction functions written in a relational logic specification language is described. The logic includes a transitive closure operator that makes it convenient for expressing reachability queries on linked data structures. Some optimizations, including tuple elimination, iterator flattening, and traversal state reduction, are used to improve performance of the generated iterators.