List your company

Question Answering with Abstention on Unified Abstention Benchmark Suite

91.5MMLU AccuracyTRACE INVERSION
35.75650.22864.779.172Apr 2, 2026

Evaluation results

MethodLinks
TRACE INVERSIONBase LLM=DEEPSEEK-R1-DISTILL-QWEN-32B2026.0491.588.381.28761.673.161.265.371.371.260.267.673.3
TRACE INVERSIONBase LLM=GPT-OSS-120B2026.0488.585.177.883.858.581.46471.271.180.469.573.776.2
COOPERATEBase LLM=DEEPSEEK-R1-DISTILL-QWEN-32B, Category=collaboration2026.0484.971.572.276.270.771.829.857.348.851.14247.360.4
COMPETEBase LLM=GPT-OSS-120B, Category=collaboration2026.0478.884.761.47561.669.15963.258.861.148.756.264.8
COMPETEBase LLM=DEEPSEEK-R1-DISTILL-QWEN-32B, Category=collaboration2026.0478.464.755.666.232.948.830.137.358.360.533.850.951.5
PROBSBase LLM=DEEPSEEK-R1-DISTILL-QWEN-32B, Category=calibration2026.047773.96070.365.362.241.256.250.364.448.754.560.3
PROBSBase LLM=GPT-OSS-120B, Category=calibration2026.0476.779.957.771.461.763.241.755.561.462.440.854.960.6
ASKCALIBase LLM=DEEPSEEK-R1-DISTILL-QWEN-32B, Category=calibration2026.0476.578.460.171.755.745.431.744.351.167.240.452.956.3
REFLECTBase LLM=GPT-OSS-120B, Category=prompting2026.0475.973.953.367.743.862.330.545.561.564.441.355.756.3
COOPERATEBase LLM=GPT-OSS-120B, Category=collaboration2026.0474.981.75269.570.481.231.260.960.75938.852.861.1
REFLECTBase LLM=DEEPSEEK-R1-DISTILL-QWEN-32B, Category=prompting2026.0474.874.463.97163.361.15158.55161.83147.959.1
PROBSBase LLM=QWEN2.5-32B, Category=calibration2026.0474.171.151.265.532.955.139.742.645.664.149.853.253.8
ASKCALIBase LLM=GPT-OSS-120B, Category=calibration2026.0472.580.450.167.755.268.241.454.958.85942.653.558.7
TRACE INVERSIONBase LLM=QWEN2.5-32B2026.0471.98578.878.678.971.271.773.96773.466.869.173.8
TRACE INVERSIONBase LLM=PHI-42026.0471.273.375.773.466.38369.472.964.97161.464.370.2
ASKCALIBase LLM=QWEN2.5-32B, Category=calibration2026.0471.168.460.166.547.351.331.943.545.164.740.15053.3
REFLECTBase LLM=QWEN2.5-32B, Category=prompting2026.0468.973.163.968.661.86130.551.141.562.141.148.256
COMPETEBase LLM=QWEN2.5-32B, Category=collaboration2026.0468.874.765.669.750.967.248.855.666.770.4496262.4
COOPERATEBase LLM=QWEN2.5-32B, Category=collaboration2026.0463.772.54259.451.661.834.549.332.260.342.44551.2
COMPETEBase LLM=PHI-4, Category=collaboration2026.0457.854.751.654.742.66953.35546.76031.24651.9
PROBSBase LLM=PHI-4, Category=calibration2026.0447.750.948.849.145.166.630.347.351.262.433.248.948.4
ASKCALIBase LLM=PHI-4, Category=calibration2026.0447.150.450.649.45561.230.74955.261.829.94949.1
COOPERATEBase LLM=PHI-4, Category=collaboration2026.0442.468.5425150.471.841.454.536.959.842.246.350.6
REFLECTBase LLM=PHI-4, Category=prompting2026.0437.954.143.845.363.377.140.560.345.551.541.14650.6