List your company
ResearchFormerly WizWand

Language Model Accuracy Evaluation on LLM Evaluation Scenarios

85.44AccuracyICL Accuracy
-0.807221.583943.97566.3661Sep 29, 2025
Updated 7mo ago

Evaluation results

MethodLinks
ICL AccuracyModel=Qwen2.5-32B2025.0985.44
ICL AccuracyModel=Llama3-70B2025.0981.93
Accuracy with LTVModel=Yi-34B, Injection Layer=50% of model's total depth, Injection Position=last-token hidden state2025.0981.37
ICL AccuracyModel=Yi-34B2025.0981.33
Accuracy with LTVModel=Llama3-70B, Injection Layer=50% of model's total depth, Injection Position=last-token hidden state2025.0978.18
Accuracy with LTVModel=Qwen2.5-32B, Injection Layer=50% of model's total depth, Injection Position=last-token hidden state2025.0975.59
Zero-Shot AccuracyModel=Yi-34B2025.0914.82
Zero-Shot AccuracyModel=Qwen2.5-32B2025.0912.52
Zero-Shot AccuracyModel=Llama3-70B2025.092.51