Hang Zhao argues robot benchmarks ignoring inference time overstate performance
Overview
At CoRL 2026 in Austin, Tsinghua assistant professor and Galaxea CTO Hang Zhao gave an Early Career Keynote, one of three awarded this year, arguing that benchmarks which ignore inference time overstate robot performance.
In RoboCasa experiments, direct and agentic control fell to 15% and 20% success under a real clock, according to the report, while layered designs with fast low-level control held up. The report frames these results as support for a clock-based framework for organizing robot intelligence.
Written by AI from the articles below · updated Nov 12, 4:17 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- LeiphoneNewsXing Zhao presents a clock-based framework for robot intelligence at CoRL 2026
AIAt CoRL 2026 in Austin, Tsinghua assistant professor and Galaxea CTO Hang Zhao gave an Early Career Keynote, one of three awarded this year. He argued that benchmarks ignoring inference time overstate robot performance. In RoboCasa experiments, direct and agentic control fell to 15% and 20% success under a real clock, while layered designs with fast low-level control held up.
Heat Trend
- Current comparable heat
- 9
- Peak comparable heat
- 9Nov 12, 5:00 AM ET
- Change in past 24 hours
- –
The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.