Task 3 of 10 in Sakana ALE-Bench
AtCoder Heuristic Contest 9
Astra's two runs followed the time asked: 10 min 10 s when asked for 10 min, and 2 h 31 min when asked for 2 h 30 min. Muse Wonder's two runs landed furthest: 1 min 38 s when asked for 22 min 30 s, and 3 min 26 s when asked for 6 h.
- Asked for
- 6 min 15 s to 16 h
- 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 | 45 min | 26 min 51 s | 10 min 10 s Within 5% asked 10 min | 2 h 31 min Within 5% asked 2 h 30 min |
| Sol | 1 h 30 min | 6 min 11 s | 11 min 1.10× longer asked 10 min | 2 h 32 min Within 5% asked 2 h 30 min |
| Fable | 10 min | 23 min | 5 min 16 s 1.90× shorter asked 10 min | 1 h 22 min 1.82× shorter asked 2 h 30 min |
| Opus 5.5 | 45 min | 2 h 23 min | 4 min 10 s 2.70× shorter asked 11 min 15 s | 2 h 28 min 1.21× shorter asked 3 h |
| Argon | 45 min | 1 h 35 min | 4 min 34 s 2.46× shorter asked 11 min 15 s | 3 h 1 min Within 5% asked 3 h |
| Sol 6.1 | 25 min | 15 min 53 s | 6 min 28 s Within 5% asked 6 min 15 s | 1 h 41 min Within 5% asked 1 h 40 min |
| MiMo Pro | 4 h | 56 min 15 s | 1 h 10 min 1.16× longer asked 1 h | 3 h 26 min 4.66× shorter asked 16 h |
| MiMo UltraSpeed | 4 h | 44 min 16 s | 1 h 11 min 1.19× longer asked 1 h | 48 min 46 s 19.69× shorter asked 16 h |
| Muse Wonder | 1 h 30 min | 59 min 52 s | 1 min 38 s 13.72× shorter asked 22 min 30 s | 3 min 26 s 104.65× shorter asked 6 h |
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: Imajuku et al., 2025