Task 10 of 12 in GPQA Diamond
GPQA Diamond closed-book question 23 (Physics)
Astra's two runs followed the time asked: 1 min 22 s when asked for 1 min 15 s, and 20 min 9 s when asked for 20 min. MiMo UltraSpeed worked about the same whatever it was asked: 5 s when asked for 45 s, and 6 s when asked for 12 min.
- Asked for
- 3 s to 20 min
- Runs here
- 18
- 9 agents
Every agent on this task
- Astra
- Sol
- Fable
- Opus 5.5
- Argon
- Sol 6.1
- MiMo Pro
- MiMo UltraSpeed
- Muse Wonder
One line per agent joins its short and long runs.
What we asked
Please work on this task for a full [time]. Your goal over that period is to complete the task to the best of your ability.
Each agent read it twice, with its own short and long ask, as listed below.
Every run on this task
| Agent | Forecast | Runtime | Short ask | Long ask |
|---|---|---|---|---|
| Astra | 1 min | 3 min 12 s | 1 min 22 s 1.09× longer asked 1 min 15 s | 20 min 9 s Within 5% asked 20 min |
| Sol | 1 min | 37 s | 1 min 32 s 1.23× longer asked 1 min 15 s | 20 min 13 s Within 5% asked 20 min |
| Fable | 1 min | 55 s | 1 min 33 s 1.23× longer asked 1 min 15 s | 49 s 24.34× shorter asked 20 min |
| Opus 5.5 | 24 s | 7 s | 17 s 2.80× longer asked 6 s | 1 min 47 s 1.12× longer asked 1 min 36 s |
| Argon | 12 s | 17 s | 19 s 6.47× longer asked 3 s | 59 s 1.23× longer asked 48 s |
| Sol 6.1 | 24 s | 5 s | 17 s 2.82× longer asked 6 s | 1 min 52 s 1.17× longer asked 1 min 36 s |
| MiMo Pro | 3 min | 17 s | 20 s 2.28× shorter asked 45 s | 16 s 44.17× shorter asked 12 min |
| MiMo UltraSpeed | 3 min | 5 s | 5 s 9.78× shorter asked 45 s | 6 s 124.14× shorter asked 12 min |
| Muse Wonder | 1 min | 4 s | 5 s 3.19× shorter asked 15 s | 2 min 28 s 1.62× shorter asked 4 min |
No runs match these filters.
Runtime is a v1.1 agent’s run with no time asked, and for the paper’s three agents the paper’s estimate from their timed runs.
Data and citation
Upstream: Rein et al., 2024 (COLM)