Social IQa
text
+
+
+
+
About
Social-IQA is the first large-scale benchmark for commonsense reasoning about social situations, featuring 38,000 multiple-choice questions that test emotional and social intelligence in everyday scenarios. This comprehensive evaluation challenges AI models to understand social dynamics, predict human behavior, and demonstrate sophisticated reasoning about interpersonal relationships and social contexts.
+
+
+
+
Evaluation Stats
Total Models9
Organizations2
Verified Results0
Self-Reported9
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers
Score Distribution
9 models
Top Score
78.0%
Average Score
58.9%
High Performers (80%+)
0Top Organizations
#1Microsoft
3 models
75.1%
#2Google
6 models
50.8%
+
+
+
+
Leaderboard
9 models ranked by performance on Social IQa
License | Links | ||||
---|---|---|---|---|---|
Aug 23, 2024 | MIT | 78.0% | |||
Aug 23, 2024 | MIT | 74.7% | |||
Feb 1, 2025 | MIT | 72.5% | |||
Jun 27, 2024 | Gemma | 53.7% | |||
Jun 27, 2024 | Gemma | 53.4% | |||
Jun 26, 2025 | Proprietary | 50.0% | |||
May 20, 2025 | Gemma | 50.0% | |||
May 20, 2025 | Gemma | 48.8% | |||
Jun 26, 2025 | Proprietary | 48.8% |