HIDE Benchmark Exposes Memory Gaps in Robotic Manipulation Under Partial Observability
Yansong Shi · hf · 2026-10-02
Researchers introduce HIDE, a benchmark for manipulation memory under partial observability, and SEEK, a memory-augmentation framework.
- HIDE spans 15 tasks—repetition counting, historical-state recall, execution-progress tracking—with randomized setups and decision points where similar observations require different actions based on prior events
- SEEK combines three complementary memory mechanisms to retain historical evidence and track execution state
- Evaluations show substantial limitations in existing policies; memory augmentation improves success rates in both simulation and real-world tests, with the combined configuration achieving the highest average success rate
- Takeaway: policies need internal representations of hidden task states, and memory design must match task-specific information needs
More from Embodied
- VirtualMacOniPad virtualizes full macOS on jailbroken iPads, running Xcode and Final Cut Pro — tom_doerr · 2026-10-02
- InterEvolve Evolves Reward Programs at Test Time to Teach Humanoid Robots New Skills — UIUC-CS · 2026-10-02
- Human beats Sharpa humanoid robot in IROS reflex drop-sticks challenge, video shows — TinfoilTricorn · 2026-10-02
- Robotic gripper now runs on Opus-written code, aligning parts via built-in light imaging — ihorbeaver · 2026-10-02
- AMD's World Labs Deal Validates 4D World Models; Chinese Rival Lands $28M Factory Order — 机器之心 · 2026-10-02
- Boston Dynamics' new Atlas hand GR3 drops the pinky: 13-DoF design trade-offs explained — xiaohu · 2026-10-02