List your company

User Profile Summarization on HPR-Bench sum

75.28AccExUserGPT
50.652857.046463.4469.8336May 9, 2026

Evaluation results

MethodLinks
UserGPTTraining Protocol=RL post-training2026.0575.2897.4719.316.9182.156.35599.89966.58826.0502
Qwen3-235B-A22B-Thinking-2507Model Type=Thinking2026.0570.1496.3818.85.9279.176.25499.80976.12967.177
UserGPTSFTTraining Protocol=SFT2026.0568.9995.8716.313.7878.546.04479.84066.12536.4236
GLM-5Model Type=Thinking2026.0567.7594.2218.045.1481.136.04329.87346.196.5677
Qwen3-235B-A22B-Instruct-2507Model Type=Non-Thinking2026.0564.2394.7116.164.2776.75.91929.8216.0866.9236
Kimi-K2.5Model Type=Thinking2026.0563.5693.3214.913.1878.675.65179.86245.78256.4214
DeepSeek-V3.2Model Type=Thinking2026.0563.0493.4116.654.4880.625.69639.84936.00525.9607
DeepSeek-R1-0528Model Type=Thinking2026.0562.5592.2111.852.3276.665.60629.84285.49985.3581
Qwen3.6-PlusModel Type=Thinking2026.0557.1390.4613.712.6277.285.44739.88395.58837.8214
Qwen3-8B ThinkingModel Type=Thinking2026.0551.9288.018.171.5377.324.70629.85154.84894.9192
Qwen3-8B Non-ThinkingModel Type=Non-Thinking2026.0551.688.218.281.7578.114.75689.77954.96164.4236