XGUARD proposes a five-level severity taxonomy and Attack Severity Curve for evaluating LLM outputs on extremist content, tested on six open-source models with SFT and ICE defenses.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
XGUARD: A Graded Benchmark for Evaluating Safety Failures of Large Language Models on Extremist Content
XGUARD proposes a five-level severity taxonomy and Attack Severity Curve for evaluating LLM outputs on extremist content, tested on six open-source models with SFT and ICE defenses.