Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

WorkBench

Benchmarks

Task NameDataset NameSOTA ResultTrend
Misalignment DetectionWorkBench (n=326)
Average IR52
30
Misalignment DetectionWorkbench n=62 (general)
Error Rate (ER)0.3
30
Multi-agent system task solvingworkbench
Accuracy82.4
21
Job and Skill Intelligence TasksWorkBench (test)
Job2Skill MAP17.9
9
RankingWorkBench
Latency (ms)13.2
9
Agent Tool-UseWorkBench
Analytics Success Rate92.59
4
General Purpose Tool UseWorkBench
Analytics Success Rate86.11
4
Multi-tool CoordinationWorkbench
Beta_Ec x T0.267
1
Showing 8 of 8 rows