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

BLEND

Benchmarks

Task NameDataset NameSOTA ResultTrend
Cultural Pluralism Alignment (Cultural Knowledge)BLEND
Entropy (0-100)83
20
Short Question AnsweringBLEnD Short Question Answer
Average Accuracy41.5
18
Cross-lingual Cultural ConsistencyBLEnD Non-Indo-European
Max Sigma0.022
15
Cross-lingual Cultural ConsistencyBLEnD Indo-European
Max Sigma0.021
15
Cross-lingual Cultural ConsistencyBLEnD Lower-Resource
Max Sigma0.02
15
Cross-lingual Cultural ConsistencyBLEnD Higher-Resource
Max Sigma0.025
15
Cross-lingual Cultural ConsistencyBLEnD All 8 Languages
Max Sigma0.017
15
Cultural ReasoningBLEnD (test)
Accuracy85.81
10
Open-ended culturalBLEnD
chrF22.1
9
Cultural Question AnsweringBLEnD MCQ subset
Accuracy (low-su)68.18
6
Multilingual Cultural KnowledgeBLEnD average
Accuracy32.8
4
Cultural alignmentBLEnD Local
Cultural Alignment (UK)82.24
2
Cultural alignmentBLEnD English
Alignment Score (UK)82.24
2
Showing 13 of 13 rows