Share your thoughts, 1 month free Claude Pro on usSee more

Contamination Evaluation on ArxivRollBench

0.48Absolute RS_I

Llama3.1-70B

Updated 8d ago

Evaluation Results

Method	Links
Llama3.1-70B 2025.07		0.48	0.02
Phi-2 2025.07		0.62	0.36
Llama3.1-8B 2025.07		0.67	0.25
Qwen2-7B 2025.07		0.69	0.42
Qwen2.5-7B 2025.07		0.7	0.37
Llama3-8B 2025.07		0.74	0.74
Llama-3.1-Nemotron-70B 2025.07		0.77	0.14
Yi-1.5-34B 2025.07		0.81	0.55
Phi-1.5 2025.07		0.82	0.57
Phi-3.5-mini 2025.07		1.07	0.07
Llama2-7B 2025.07		1.14	0.11
Phi-1 2025.07		1.21	1.31
Phi-3-mini 2025.07		1.27	0.05
Qwen2.5-72B 2025.07		1.41	0.32