Cadenza-Labs dolphin-llama-3-8B

The 'Cadenza-Labs dolphin-llama3-8B' is a benchmark model used to evaluate the effectiveness of detection methods for identifying backdoored large language models through semantic drift analysis.

Papers using Cadenza-Labs dolphin-llama-3-8B (1)

Detecting Sleeper Agents In Large Language Models Via Semantic Drift Analysis2025