A ProgramBench task from the paper, not in v1.1
kaushiksrini/parqeye
Astra's two runs followed the time asked: 30 min 54 s when asked for 30 min, and 8 h 21 min when asked for 8 h 20 min. Sol's two runs landed furthest: 16 h 40 min when asked for 30 min, and 1 h 3 min when asked for 8 h 20 min.
- Asked for
- 30 min to 8 h 20 min
- Runs here
- 6
- See Read before quoting below
Every agent on this task
- Astra
- Sol
- Fable
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 | 4 h | 1 h 24 min | 30 min 54 s Within 5% asked 30 min | 8 h 21 min Within 5% asked 8 h 20 min |
| Sol | 3 h | 35 min 19 s | 16 h 40 min 33.33× longer asked 30 min | 1 h 3 min 7.89× shorter asked 8 h 20 min |
| Fable | 1 h 30 min | 29 min 4 s | 16 min 17 s 1.84× shorter asked 30 min | 32 min 36 s 15.34× shorter asked 8 h 20 min |
No runs match these filters.
Runtime is the paper’s estimate from each agent’s timed runs.
Read before quoting. 4 runs were stopped by AgentTime, so their time is a lower bound. Details
Data and citation
Upstream: Yang et al., 2026