Gemma

The 'Gemma' dataset is used to evaluate the effectiveness of LLM-generated text detection methods, specifically in the context of detecting text from unseen, privately-tuned large language models.

Papers using Gemma (4)

RAS: Measuring LLM Safety Through Refusal Alignment2026PhantomHunter: Detecting Unseen Privately-Tuned LLM-Generated Text via Family-Aware Learning2025Mitigating Memorization in LLMs using Activation Steering2025AttnGCG: Enhancing Jailbreaking Attacks on LLMs with Attention Manipulation2024