List your company

Agent Task on SciWorld (Suc@1, Suc@2, Suc@3)

98.4Success Rate @1PivoARL
5.73629.79353.8577.907Jul 4, 2026

Evaluation results

MethodLinks
PivoARLBackbone=Qwen3-4B2026.0798.4100100
MetaRLBackbone=Qwen3-4B2026.0794.597.698.4
GSPOBackbone=Qwen3-4B2026.0789.894.596
DAPOBackbone=Qwen3-4B2026.0783.594.598.4
SKillRLBackbone=Qwen3-4B2026.0782.18989.8
SimpleMem+GRPOBackbone=Qwen3-4B2026.0781.790.292.6
Mem0+GRPOBackbone=Qwen3-4B2026.0779.691.493.2
GiGPOBackbone=Qwen3-4B2026.0778.986.789.8
GRPOBackbone=Qwen3-4B2026.0775.882.886.7
Gemini-2.5-ProBackbone=Gemini-2.5-Pro2026.0742.955.460.9
GPT-4oBackbone=GPT-4o2026.0732.842.147.6
ReActBackbone=Qwen3-4B2026.079.412.512.5
ReflexionBackbone=Qwen3-4B2026.079.314.717.9