Embodied Decision Making on ALFWorld held-out (test) benchmark leaderboard · SOTA2 Research

Embodied Decision Making on ALFWorld held-out (test)

95.5ScoreSkillOpt

Evaluation Results

Method	Links
SkillOpt2026.05		95.5
LLM skill2026.05		93.3
Human skill2026.05		91.8
SkillOpt2026.05		91
Trace2Skill2026.05		87.3
GEPA2026.05		85.8
SkillOpt2026.05		85.8
SkillOpt2026.05		85.1
No skill2026.05		83.6
LLM skill2026.05		83.6
TextGrad2026.05		82.8
Trace2Skill2026.05		82.8
GEPA2026.05		82.8
SkillOpt2026.05		82.1
GEPA2026.05		81.3
SkillOpt2026.05		81.3
TextGrad2026.05		78.4
Trace2Skill2026.05		76.9
TextGrad2026.05		76.1
No skill2026.05		75.4
Human skill2026.05		74.6
GEPA2026.05		74.6
GEPA2026.05		73.9
Trace2Skill2026.05		73.1
No skill2026.05		73.1
TextGrad2026.05		70.9
Trace2Skill2026.05		70.9
SkillOpt2026.05		69.4
GEPA2026.05		68.7
No skill2026.05		68.7
LLM skill2026.05		68.7
TextGrad2026.05		67.9
LLM skill2026.05		65.7
LLM skill2026.05		65.7
Trace2Skill2026.05		64.9
GEPA2026.05		60.4
LLM skill2026.05		60.4
No skill2026.05		59.7
Trace2Skill2026.05		58.2
Human skill2026.05		56.7
Human skill2026.05		56.7
LLM skill2026.05		55.2
TextGrad2026.05		53.7
Human skill2026.05		44.8
TextGrad2026.05		42.5
No skill2026.05		34.3
No skill2026.05		30.6
Human skill2026.05		29.9
Human skill2026.05		28.4