Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Interactive Agent Task Completion on ScienceWorld
Loading...
90
Average Reward
OpenSkill
75.336
79.143
82.95
86.757
Jun 4, 2026
Average Reward
Updated 1mo ago
Evaluation Results
Method
Method
Links
Average Reward
OpenSkill
Target Agent=Opus 4.6
2026.06
90
SkillNet
Target Agent=Opus 4.6
2026.06
88.7
Skill Creator
Target Agent=Opus 4.6
2026.06
88.3
Self-Gen
Target Agent=Opus 4.6
2026.06
88
CoT
Target Agent=Opus 4.6
2026.06
88
No Skill
Target Agent=Opus 4.6
2026.06
87.1
OpenSkill
Target Agent=GPT 5.2
2026.06
85.3
Skill Creator
Target Agent=GPT 5.2
2026.06
83.1
CoT
Target Agent=GPT 5.2
2026.06
80
No Skill
Target Agent=GPT 5.2
2026.06
78
SkillNet
Target Agent=GPT 5.2
2026.06
77.8
Self-Gen
Target Agent=GPT 5.2
2026.06
75.9
Feedback
Search any
task
Search any
task