Benchmark Radar
AI BENCHMARK PROFILE

SoftVTBench: A Deformation-Aware Visuo-Tactile Dataset and Benchmark for Deformable-Object Manipulation

Robotics & Autonomous SystemsRobotics & Embodied Intelligence

SoftVTBench is a visuo-tactile dataset and benchmark for deformable-object manipulation, providing expert demonstrations with synchronized sensor data and finite-element ground truth, with a closed-loop evaluation protocol.

Released
2026-08-19
Readiness
Paper only
Primary field
Robotics & Autonomous Systems

Why it matters

Most manipulation benchmarks evaluate task success alone, ignoring physical interaction quality. SoftVTBench enables evaluation of deformation-aware success, addressing how policies interact with deformable objects.

Motivation

Physical interaction quality is central to deformable-object manipulation, yet most benchmarks evaluate task success alone.

Primary resources

Benchmark Radar records only publicly supported details and links back to primary sources for verification.