A new study published on arXiv investigates the option-set dependence of hosted decision models, specifically testing the Jev model. The research found that adding irrelevant alternatives to classification tasks significantly shifted the probabilities assigned to the correct and incorrect options. This dependence was observed across various datasets, including Banking77, CLINC150, and AG News, with shifts being directional and increasing with the number of options presented. The study suggests that the probabilities from Jev should be interpreted relative to the specific option list used in a given query. AI
IMPACT Highlights potential biases in hosted decision models, suggesting careful interpretation of probability outputs based on the option set.
RANK_REASON The cluster contains a research paper detailing experimental findings on a specific AI model's behavior. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →