Survey maps progress reward modeling across robotic learning and benchmarks

northwestern-university · hf · 2026-07-28

• A comprehensive survey of progress reward modeling for robotic learning, aimed at providing feedback during task execution rather than only at terminal success.
• The paper argues the field lacks a shared framework: methods differ in observations, goal specifications, output signals, supervision sources, and evaluation protocols.
• It organizes the literature into three layers:
• Interface — what inputs the model receives and what progress signal it emits.
• Construction methods — how the signal is estimated or turned into reward.
• Data and benchmarks — how supervision is collected and what existing evaluations actually measure.
• The survey also summarizes current limitations and sketches future research directions.

Original post →

More from Embodied

• Japan industrial robot orders rose 40% to a record ¥312.7 billion as AI spending accelerated — 创业邦 · 2026-07-28
• Embodied manipulation gets a five-layer data pyramid for robot alignment — PekingUniversity · 2026-07-28
• Dex-Net in 60 Seconds: A Classic Look at Robotic Grasping Uncertainty — berkeley_ai · 2026-07-28
• Being-H0.8 adds tactile prediction to robot world models with 500,000 hours of video — 量子位 · 2026-07-28
• Webots, an open-source robot simulator for vehicles and mechanical systems — tom_doerr · 2026-07-28
• Tesla's New Feature: Headlights Project Navigation Path Directly onto the Road — sven_ai · 2026-07-28

Embodied channel →