ResearchBenchmarksCalibration on MMLU-Pro (test)Follow16.2ECELMSI16.05617.0281818.972Jul 2, 2026Evaluation ResultsMethodMethodLinksECELMSIEval.=crossEval.=cross2026.0716.2LMSIEval.=inEval.=in2026.0717.1IntuitorEval.=inEval.=in2026.0717.4N-OPSDEval.=inEval.=in2026.0718Qwen3-4B baseEval.=baseEval.=base2026.0718.4IntuitorEval.=crossEval.=cross2026.0718.5N-OPSDEval.=crossEval.=cross2026.0718.6TTRLEval.=crossEval.=cross2026.0719TTRLEval.=inEval.=in2026.0719.8