method · first seen 2 Nov 2026
counterfactual identity templates
Evaluation method that instantiates identical text templates with different identity terms to isolate group-specific model behavior.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Same Sentence, Different Construct: Identity-Swapped Rationales Reveal Measurement Non-Invariance in LLM Hate Speech ClassifiersPapers · 2 Nov 2026 · Wenchang Wang, Yue Song
Latest sources
- Same Sentence, Different Construct: Identity-Swapped Rationales Reveal Measurement Non-Invariance in LLM Hate Speech ClassifiersPapers · 2 Nov 2026 · Wenchang Wang, Yue Song