List your company

Language Modeling on The Pile (Teacher Comparison & Efficiency)

4.05Inference SpeedupBudgeted LoRA
0.8781.70152.5253.3485May 29, 2025Jul 24, 2025Sep 19, 2025Nov 15, 2025Jan 11, 2026Mar 9, 2026May 5, 2026

Evaluation results

MethodLinks
Budgeted LoRADense budget (F)=0.02026.054.0525.1328.4528.860.3880.2—
AC-ODM(160M) (Proxy Policy)AC Parameters=17M, LLM Parameters=160M, Total Parameters=177M, Time (s)=0.65, Model Family=Pythia2025.052.08—————28,690
AC-ODM(160M) (Target Transfer)AC Parameters=17M, LLM Parameters=1B, Total Parameters=1.02B, Time (s)=2.48, Model Family=Pythia2025.052.08—————12,500
Budgeted LoRADense budget (F)=0.42026.051.7424.3426.1926.580.5953.9—
AC-ODM(410M) (Proxy Policy)AC Parameters=17M, LLM Parameters=410M, Total Parameters=427M, Time (s)=1.41, Model Family=Pythia2025.051.47—————28,690
AC-ODM(410M) (Target Transfer)AC Parameters=17M, LLM Parameters=1B, Total Parameters=1.02B, Time (s)=2.48, Model Family=Pythia2025.051.47—————12,010
AC-ODMAC Parameters=17M, LLM Parameters=1B, Total Parameters=1.02B, Time (s)=2.48, Model Family=Pythia2025.051.46—————28,356
PiKEAC Parameters=0, LLM Parameters=1B, Total Parameters=1B, Time (s)=2.53, Model Family=Pythia2025.051.3—————31,250
KD Full2026.05120.8818.2617.8610—
KD LoRA2026.05124.3226.0626.50.910—
ODMAC Parameters=0, LLM Parameters=1B, Total Parameters=1B, Time (s)=2.47, Model Family=Pythia2025.051—————41,667