List your company

Interactive Agent Task Completion on ScienceWorld

90Average RewardOpenSkill
75.33679.14382.9586.757Jun 4, 2026

Evaluation results

MethodLinks
OpenSkillTarget Agent=Opus 4.62026.0690
SkillNetTarget Agent=Opus 4.62026.0688.7
Skill CreatorTarget Agent=Opus 4.62026.0688.3
Self-GenTarget Agent=Opus 4.62026.0688
CoTTarget Agent=Opus 4.62026.0688
No SkillTarget Agent=Opus 4.62026.0687.1
OpenSkillTarget Agent=GPT 5.22026.0685.3
Skill CreatorTarget Agent=GPT 5.22026.0683.1
CoTTarget Agent=GPT 5.22026.0680
No SkillTarget Agent=GPT 5.22026.0678
SkillNetTarget Agent=GPT 5.22026.0677.8
Self-GenTarget Agent=GPT 5.22026.0675.9