List your company

Search Routing on PopQA & KUQ Oracle-eligible (test)

83.65Macro F1Qwen3.5-4B
70.005273.547677.0980.6324Jul 7, 2026

Evaluation results

MethodLinks
Qwen3.5-4BPolicy=Preference Opt.2026.0783.6584.2723.1210.3492.9691.84
Gemma E2BPolicy=Preference Opt.2026.0782.3583.2920.5914.3481.0394.34
Gemma E2BPolicy=SFT2026.0782.0782.9219.2815.7477.5988.68
Qwen3.5-4BPolicy=SFT2026.0782.0782.6322.8413.3992.9687.76
Gemma E2BPolicy=Ppre2026.0770.8275.7454.255.9868.9771.7
Qwen3.5-4BPolicy=Ppre2026.0770.5371.4836.4922.7287.3265.31