List your company

Streaming Spatial Reasoning on S3-Eval Sim

80.5Overall AccuracyHuman level
20.07635.76351.4567.137Mar 25, 2026

Evaluation results

MethodLinks
Human level2026.0380.571.192.670.470.99371.371.798.692.697.173.470.966.16693.592.8
AMF-VLMAlias=Ours2026.0362.957.63555.659.470.470.867.775.672.846.36768.170.861.860.960.6
cambrian-s*†trained on our data=true, modified as streaming inference=true2026.0357.852.133.551.353.262.459.862.170.567.845.958.461.364.359.258.358.8
streaming-vlm*trained on our data=true2026.035654.93151.845.258.36660.96765.64059.160.163.255.855.255
GPT-5.22026.0344.938.429.737.434.641.835.945.561.458.644.435.644.451.553.449.652.7
cambrian-s2026.0344.829.222.848.150.137.930.358.750.452.537.35838.843.458.139.256.3
Gemini-3-pro2026.0343.127.132.547.624.258.324.932.269.15931.453.75836.737.855.537.6
Gemini-3-flash2026.0341.935.627.634.531.538.832.742.358.155.541.532.541.648.549.646.849.6
vlm-3r2026.0341.931.627.135.232.340.832.541.156.253.641.141.44349.249.144.547.3
Qwen3-VL-8B2026.0341.13625.83429.739.63241.156.754.240.232.541.548.148.845.449.1
InternVL3.5-8B2026.0337.723.325.639.927.74231.93449.945.846.728.942.913.856.443.948.9
Qwen3-VL-32B2026.0337.43427.137.831.937.938.54052.248.850.146.843.41.5342.913.240.2
Frequent2026.0335.646.126.147.646.7263339.326.927.737.137.635.946.539.626.825.8
InternVL3.5-38B2026.0335.243.427.128.427.241.521.137.437.944.25223.336.30.1652.739.449.4
llava-video2026.0331.67.0322.811.429.738.421.315.550.645.546.31429.239.252.635.547.1
Qwen3-VL-4B2026.0330.827.520.735.630.226.5274114.819.226.531.436.543.742.138.931.7
llava-ov-1.5-8B2026.0328.87.8924.616.233.838.626.15.7332.341.241.712.131.9423134.444
InternVL3.5-1B2026.0326.417.726.112.911.835.910.522.249.243.1430.9420.30.2252.431.844.3
Random2026.0325.6—21——29.1——23.722.230.7————26.226.6
llava-ov-1.5-4B2026.0322.45.5327.426.516.638.12.712.2231.636.639.87.3230.40.61292937.1