TUA-Bench, longest of three requests

Muse Spark 1.3 on “Extract gym auditorium”

Shorter

2.42× shorter

Stopped at 41% of the time asked, 11 min 45 s short.

Asked

Worked

Safety cutoff 40 min, off the scale

Asked for: 20 minutes
Worked: 8 min 15 s
Grade: Passed; Score 100 of 100. Graded during the run. Not in the paper's scores.
How it ended: Stopped by an error

Every detail of this run

Agent: Muse Spark 1.3, in Muse Code
Benchmark: TUA-Bench
Task: Extract gym auditorium
Asked for: 20 minutes: “Please work on this task for a full 20 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked: 8 min 15 s (495 s)
Ratio: 0.41× the request
How it ended: Stopped by an error
Grade: Passed; It completed the task in full. The grader returned 1. Pass rule: Full completion: a reward of 1 on the final state.
Score: 100 of 100; Graded during the run. Not in the paper's scores.
Wording: The standard sentence, used from 4 September 2026.
Run id: 2ensw5nqlr
Data release: 26 Sep 2026

The other requests for Muse Spark 1.3 on this task, and the other agents at this request
Muse on this task	Asked	Worked	Result
Shortest	1.25 minutes	3 min 20 s	2.66× longer
Middle	5 minutes	8 min 31 s	1.70× longer
Longest (this run)	20 minutes	8 min 15 s	2.42× shorter
Other agents, longest request
GPT 6 Astra	20 minutes	20 min 27 s	Within 5%
GPT 5.6 Sol	20 minutes	21 min 32 s	1.08× longer
Claude Fable 5.1	20 minutes	22 min 12 s	1.11× longer

See it on the chart. Transcripts are not published yet.

Read before quoting. This run was stopped by an error, so its time is a lower bound: the agent might have kept going. Details

Data and citation