List your company

Legal Question Answering on MMLU-Pro Law n=1,101 (test)

75.7AccuracyEP-HUBO (Qwen-35B)
45.0252.98560.9568.915Jun 5, 2026

Evaluation results

MethodLinks
EP-HUBO (Qwen-35B)Adj.=Opus4.6, Trace generator=Qwen3.5-35B, Traces per question=202026.0675.7————
ZS (no traces/HUBO)Adj.=Opus4.6, Traces=None2026.0674.2————
EP-HUBO (OSS-20B)Adj.=Opus4.6, Trace generator=OSS-20B, Traces per question=202026.0674.1————
EP-HUBO (Qwen-35B)Adj.=Sonnet4.6, Trace generator=Qwen3.5-35B, Traces per question=202026.0667.1————
EP-HUBO (OSS-20B)Adj.=Sonnet4.6, Trace generator=OSS-20B, Traces per question=202026.0663.9————
ZS (no traces/HUBO)Adj.=Sonnet4.6, Traces=None2026.0663.1————
MV (Qwen-35B)Trace generator=Qwen3.5-35B, Traces per question=202026.0663.1————
MV (OSS-20B)Trace generator=OSS-20B, Traces per question=202026.0646.2————
HUBO prec (Opus, OSS-20B)Adj.=Opus, Trace generator=OSS-20B2026.06——555649.5
HUBO prec (Opus, Qwen-35B)Adj.=Opus, Trace generator=Qwen-35B2026.06——725556.7
HUBO prec (Sonnet, OSS-20B)Adj.=Sonnet, Trace generator=OSS-20B2026.06——918352.3
HUBO prec (Sonnet, Qwen-35B)Adj.=Sonnet, Trace generator=Qwen-35B2026.06——1298560.3
Net Opus vs MV (OSS-20B)Trace generator=OSS-20B2026.06—27.9———
Net Opus vs MV (Qwen-35B)Trace generator=Qwen3.5-35B2026.06—12.6———
Net Opus vs ZS Opus (OSS-20B)Trace generator=OSS-20B2026.06—0.1———
Net Opus vs ZS Opus (Qwen-35B)Trace generator=Qwen3.5-35B2026.06—1.5———
Net Sonnet vs MV (OSS-20B)Trace generator=OSS-20B2026.06—17.6———
Net Sonnet vs MV (Qwen-35B)Trace generator=Qwen3.5-35B2026.06—4———
Net Sonnet vs ZS Son (OSS-20B)Trace generator=OSS-20B2026.06—0.7———
Net Sonnet vs ZS Son (Qwen-35B)Trace generator=Qwen3.5-35B2026.06—4———