List your company

Tool-use Agent Performance on GeoDisaster

98.75Tool Usage Success Rate (Any)MAS+SFT
-3.9522.712549.37576.0375Jun 15, 2026

Evaluation results

MethodLinks
MAS+SFTType=Multi2026.0698.7598.2698.4192.4682.4388.1694.6999.4399.5899.5899.5891.01
MAS+SFT+GRPOType=Multi2026.0698.5697.7698.7294.2490.1188.8895.8910010010010091.59
o4-miniType=Single2026.0623.2620.9323.2629.1226.6430.5944.1944.1986.9786.9786.9747.1
GPT-5.5Type=Single2026.0620.9320.9330.2355.0661.9971.4410010095.8695.8695.8682.32
GPT-5Type=Single2026.0611.9911.9913.746.2760.4865.3781.5181.5169.7171.5669.7169.27
MASType=Multi2026.066.854.1110.6220.3622.9318.2649.6690.8881.7195.4662.5730.03
GPT-4oType=Single2026.063.443.443.7842.6560.1965.4380.7680.7678.5378.5378.5371.23
Qwen2.5-7B-Inst.Type=Single2026.062.40.344.453.112.534.455.825.8242.0677.0536.526.45
Qwen3-4B-Inst.Type=Single2026.060001.161.162.332.336.984023.269.33.72
Llama-3.1-8B-Inst.Type=Single2026.060000.150.250.250.250.2500.70.350.35
Mistral-7B-Inst.Type=Single2026.060000.150.250.250.250.2500.70.350.35
Qwen2.5-VL-7BType=Single2026.060000.951.52.11.163.49012.45.813.1