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

Multi-turn task completion on PersuadeGym

57.9Task Completion Score

GRPO + Qwen2.5-4B

36.89242.34647.853.254Apr 8, 2026
Updated 3mo ago

Evaluation Results

MethodLinks
2026.04
57.9
2026.04
56.8
2026.04
53.2
2026.04
53.2
2026.04
53.2
2026.04
49.8
2026.04
48.4
2026.04
47.2
2026.04
44.1
2026.04
42.5
2026.04
40.9
2026.04
40.5
2026.04
37.7