Task 14 of 14 in DeepSWE
Add deterministic map conflict detection to Y.Map writes
Sol 6.1's two runs followed the time asked: 12 min 52 s when asked for 12 min 30 s, and 3 h 20 min when asked for 3 h 20 min. Muse Wonder worked about the same whatever it was asked: 7 min 19 s when asked for 33 min 45 s, and 5 min 1 s when asked for 9 h.
- Asked for
- 5 min 30 s to 16 h
- Runs here
- 12
- 6 agents
Every agent on this task
- 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 |
|---|---|---|---|---|
| Opus 5.5 | 55 min | 1 h 9 min | 9 min 30 s 1.45× shorter asked 13 min 45 s | 1 h 17 min 2.88× shorter asked 3 h 40 min |
| Argon | 22 min | 1 h 5 min | 42 min 31 s 7.73× longer asked 5 min 30 s | 1 h 17 min 1.14× shorter asked 1 h 28 min |
| Sol 6.1 | 50 min | 10 min 57 s | 12 min 52 s Within 5% asked 12 min 30 s | 3 h 20 min Within 5% asked 3 h 20 min |
| MiMo Pro | 2 h 30 min | 25 min 49 s | 25 min 16 s 1.48× shorter asked 37 min 30 s | 32 min 19 s 18.56× shorter asked 10 h |
| MiMo UltraSpeed | 4 h | 9 min 49 s | 11 min 55 s 5.04× shorter asked 1 h | 36 min 24 s 26.37× shorter asked 16 h |
| Muse Wonder | 2 h 15 min | 17 min 59 s | 7 min 19 s 4.61× shorter asked 33 min 45 s | 5 min 1 s 107.57× shorter asked 9 h |
No runs match these filters.
Runtime is each agent’s run with no time asked.
Data and citation
Upstream: Huang et al., 2026