List your company
ResearchFormerly WizWand

Distribution Alignment on Aggregate of Six Datasets

0.082Average MAEKL-Optimized Fine-Tuning
0.071240.143870.21650.28913Apr 7, 2026
Updated 6mo ago

Evaluation results

MethodLinks
KL-Optimized Fine-TuningModel=Qwen2.5-1.5B-Instruct2026.040.082
KL-Optimized Fine-TuningModel=Qwen2.5-7B-Instruct2026.040.085
IFTModel=Qwen2.5-1.5B-Instruct2026.040.102
KL-Optimized Fine-TuningModel=Llama-3.2-1B-Instruct2026.040.103
KL-Optimized Fine-TuningModel=Llama-3.1-8B-Instruct2026.040.116
IFTModel=Llama-3.1-8B-Instruct2026.040.131
IFTModel=Qwen2.5-7B-Instruct2026.040.135
IFTModel=Llama-3.2-1B-Instruct2026.040.142
Zero-shotModel=Llama-3.1-8B-Instruct2026.040.222
PE-ImplicitModel=Qwen2.5-7B-Instruct2026.040.228
Zero-shotModel=Qwen2.5-1.5B-Instruct2026.040.238
PE-ExplicitModel=Qwen2.5-1.5B-Instruct2026.040.247
PE-ImplicitModel=Llama-3.1-8B-Instruct2026.040.248
PE-ExplicitModel=Llama-3.1-8B-Instruct2026.040.25
Zero-shotModel=Qwen2.5-7B-Instruct2026.040.267
PE-ImplicitModel=Qwen2.5-1.5B-Instruct2026.040.267
Zero-shotModel=Llama-3.2-1B-Instruct2026.040.28
DPOModel=Llama-3.2-1B-Instruct2026.040.307
PE-ExplicitModel=Qwen2.5-7B-Instruct2026.040.311
DPOModel=Qwen2.5-1.5B-Instruct2026.040.317
PE-ExplicitModel=Llama-3.2-1B-Instruct2026.040.319
DPOModel=Llama-3.1-8B-Instruct2026.040.324
DPOModel=Qwen2.5-7B-Instruct2026.040.333
PE-ImplicitModel=Llama-3.2-1B-Instruct2026.040.351