Task 8 of 18 in AssistantBench

AssistantBench validation task 8

Astra's three runs landed closest to the request (4 min 33 s, 14 min 38 s and 19 min 2 s). Sol's runs landed furthest from the request (3 min 2 s, 31 min 23 s and 37 min 50 s).

Benchmark: AssistantBench; Upstream id validation:9
Asked for: 1.25, 5 and 20 min
Runs here: 9 of 9

One task, asked three ways

Astra
Sol
Fable

One line per agent joins its three runs.

What we asked

Please work on this task for a full [1.25 | 5 | 20] minutes. Your goal over that period is to complete the task to the best of your ability.

One sentence, sent three times with a different time: 1.25, 5 and 20 min.

The three sentences in full

Shortest1.25 minutes

Please work on this task for a full 1.25 minutes. Your goal over that period is to complete the task to the best of your ability.

Middle5 minutes

Please work on this task for a full 5 minutes. Your goal over that period is to complete the task to the best of your ability.

Longest20 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.

Every run on this task

Every run on AssistantBench validation task 8, one row per agent
Agent	Shortest	Middle	Longest
Astra	4 min 33 s 3.64× longer Did not pass, 12.9 of 100	14 min 38 s 2.93× longer Did not pass, 12.6 of 100	19 min 2 s Within 5% Did not pass, 11.7 of 100
Sol	3 min 2 s 2.42× longer Did not pass, 8.9 of 100	31 min 23 s 6.28× longer Did not pass, 6.8 of 100	37 min 50 s 1.89× longer Did not pass, 6.5 of 100
Fable	6 min 9 s 4.92× longer Did not pass, 14 of 100	8 min 21 s 1.67× longer Did not pass, 0 of 100	14 min 37 s 1.37× shorter Did not pass, 0 of 100

Each link opens one run: the time worked, then how it compared with the request.

Data and citation

Upstream: Yoran et al., 2024 (EMNLP)