12.4%
#1 of 5
full task success rate of Robot Learning Collective, the top team; its partial-credit Q score was 26%
What the chart shows
BEHAVIOR-1K is a simulation benchmark of 1,000 household activities, drawn from surveys of what people want robots to help with, mostly long mobile manipulation tasks. In the 2025 challenge the top team, Robot Learning Collective, completed 12.4% of held-out tasks in full, ahead of Comet (11.4%) and SimpleAI Robot (10.8%).
The Q score gives partial credit for subgoals met. Robot Learning Collective scored 26% and Comet 25.1%, so even the best policies achieved only about a quarter of what the tasks required.
Key numbers
| Measure | Value |
|---|---|
| Robot Learning Collective: full success | 12.4% |
| Robot Learning Collective: Q score | 25.99% |
| Comet: full success / Q score | 11.4% / 25.14% |
| SimpleAI Robot: full success / Q score | 10.8% / 15.91% |
| Embodied Intelligence (5th): full success / Q score | 5.2% / 9.47% |
How to read it
- Q score is the fraction of a task’s subgoals completed, matched to the best-fitting goal; it rewards progress without full completion.
- Scores are on the challenge’s held-out test set; the AI Index shows the top five teams.
Source: BEHAVIOR Challenge Leaderboard, 2025, via the AI Index Report 2026 (Stanford HAI), figure 2.7.2: BEHAVIOR-1K: full task success rate vs. Q score (held-out-test). Chart and text by AI Adoption.
Numbers checked against the published data on 20 October 2026.
Download the data (CSV) · Cite this chart
Numbers checked against the published data on 20 October 2026.
Download the data (CSV) · Cite this chart
Discussion