List your company

Language Modeling on WikiText 1,000-example evaluation slice (test)

12.723PerplexityAdditive sparse
12.4756414.1453215.81517.48468May 4, 2026

Evaluation results

MethodLinks
Additive sparseBackbone=Qwen-2.5-0.5B-Instruct, Selection rule=KL2026.0512.723
Additive sparseBackbone=Qwen-2.5-0.5B-Instruct, Selection rule=TF-IDF2026.0512.773
Additive sparse +SBackbone=Qwen-2.5-0.5B-Instruct, Selection rule=KL2026.0513.139
Base QwenBackbone=Qwen-2.5-0.5B-Instruct2026.0513.146
Additive sparse +SBackbone=Qwen-2.5-0.5B-Instruct, Selection rule=TF-IDF2026.0513.456
LoRABackbone=Qwen-2.5-0.5B-Instruct2026.0515.47
Replacement sparseBackbone=Qwen-2.5-0.5B-Instruct, Selection rule=TF-IDF2026.0516.501
Replacement sparseBackbone=Qwen-2.5-0.5B-Instruct, Selection rule=KL2026.0516.577
Full finetuneBackbone=Qwen-2.5-0.5B-Instruct2026.0518.907