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

SciCode

Benchmarks

Task NameDataset NameSOTA ResultTrend
Scientific Code GenerationSciCode (test)
Subset Accuracy43.8
36
Scientific CodingSciCode
Accuracy0.4275
18
Scientific ReasoningSciCode
Accuracy43
10
Coding Agent CapacityScicode
Score57.7
8
Science DiscoveryScicode
Score57.7
8
Ranking PreservationSciCode (test)
Mean Spearman Rho0.753
4
Code ReasoningSciCode
Pass@142.1
4
ReasoningSciCode
Score35.2
3
ReasoningSciCode subtask
Score (SciCode Reasoning)39.6
1
Showing 9 of 9 rows