AI BENCHMARK PROFILE
HT-Bench
HT-Bench is a multi-task benchmark for tactile representation learning with egocentric vision and full-hand tactile data, including retrieval, inpainting, synthesis, and prediction tasks.
- Released
- 2026-06-17
- Readiness
- Paper only
- Primary field
- Robotics & Autonomous Systems
Why it matters
Tactile representation learning lacks universal benchmarks. HT-Bench explores egocentric vision paired with tactile data, providing a scalable evaluation direction for dexterous manipulation.
Motivation
Establishing a universal benchmark for tactile representation learning in robotic manipulation remains challenging due to the diversity of tactile sensor designs, data formats, and robot embodiments.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.