List your company

Scientific Simulation on SciWorld

55.7Measure ScoreStraTA
-0.35614.19728.7543.303May 7, 2026

Evaluation results

MethodLinks
StraTABackbone=Qwen2.5-7B-Instruct, Category=Training-Based2026.0555.754.478.616.310063.5
Claude-4-SonnetModel Type=Closed-Source Model, Category=Prompting-Based2026.0545.454.760.345.69057.4
Gemini-2.5-FlashModel Type=Closed-Source Model, Category=Prompting-Based2026.0540.751.353.7335049.6
ScalingInterBackbone=Qwen2.5-7B-Instruct, Category=Training-Based2026.053455.488.6073.357
GPT-5.1Model Type=Closed-Source Model, Category=Prompting-Based2026.0532.146.541.869.638.343
GRPOBackbone=Qwen2.5-7B-Instruct, Category=Training-Based2026.0531.249.832.714.753.341.8
AgentGym-RLBackbone=Qwen2.5-7B-Instruct, Category=Training-Based2026.0524.55965.9066.750.5
ReActBackbone=Qwen2.5-7B-Instruct, Category=Prompting-Based2026.0512.214.433.69.628.920
VanillaBackbone=Qwen2.5-7B-Instruct, Category=Prompting-Based2026.052.112.321.29.62513.9
PPOBackbone=Qwen2.5-7B-Instruct, Category=Training-Based2026.051.857.560.96.997.851.4