List your company

Causal Inference on CaLM (test)

0.99ATEGRPO
-0.035440.230780.4970.76322Feb 6, 2026

Evaluation results

MethodLinks
GRPOTraining Stage=Post-training2026.020.990.9940.90.940.930.9280.8660.935
PPOTraining Stage=Post-training2026.020.9720.9820.8060.9260.9240.940.9020.921
SFTTraining Stage=Post-training2026.020.8520.8280.470.560.6040.8580.7660.702
OpenAI o3Model Type=Pre-trained LLM2026.020.840.590.30.430.720.450.550.554
Gemini 2.5 ProModel Type=Pre-trained LLM2026.020.760.710.320.590.470.240.050.448
DeepSeek-R1-0528-671BModel Type=Pre-trained LLM2026.020.740.540.220.460.450.780.80.57
KTOTraining Stage=Post-training2026.020.7160.6740.2320.4120.4720.8120.70.574
DPOTraining Stage=Post-training2026.020.6560.5140.1980.2820.510.8060.7080.524
Cold Start BaseModel Type=Base Model2026.020.6340.550.1560.2940.4340.7880.7140.51
DeepSeek-R1-Distill-Qwen-14BModel Type=Distilled LLM2026.020.5940.3640.210.4420.2120.0140.0660.272
Llama-3.3-70BModel Type=Pre-trained LLM2026.020.5720.3720.2880.430.20.010.010.269
Qwen3-235BModel Type=Pre-trained LLM2026.020.00400.180.230000.059