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

Secret Data Extraction Attack on FedLLM 100 Target Secrets

100Success Rate (P > .10)

GPT-2 Small (124M)

Updated 1mo ago

Evaluation Results

Method	Links
GPT-2 Small (124M) 2026.06		100	93
GPT-Neo (125M) 2026.06		100	94
Pythia (160M) 2026.06		100	96
OPT (250M) 2026.06		100	99
GPT-2 Medium (355M) 2026.06		100	100
Llama2 (13B) 2026.06		100	97
DistilGPT-2 (82M) 2026.06		99	85
Llama2 (7B) 2026.06		99	83
Llama3.2 (1B) 2026.06		97	92
GPT-2 Medium (355M) 2026.06		23	7
OPT (250M) 2026.06		21	3
Pythia (160M) 2026.06		17	0
GPT-Neo (125M) 2026.06		13	0
GPT-2 Small (124M) 2026.06		12	0
DistilGPT-2 (82M) 2026.06		9	0
Llama2 (7B) 2026.06		0	0
Llama2 (13B) 2026.06		0	0
Llama3.2 (1B) 2026.06		0	0