The abstract claims that state-of-the-art LLMs fail to infer contextual preferences from multi-turn interactions (under 50% precision, 65% recall), but the body text is an unrelated adversarial-patch paper, leaving the claim unsupported.
Lavan: Localized and visible adversarial noise
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
-
CUPID: Evaluating Personalized and Contextualized Alignment of LLMs from Interactions
The abstract claims that state-of-the-art LLMs fail to infer contextual preferences from multi-turn interactions (under 50% precision, 65% recall), but the body text is an unrelated adversarial-patch paper, leaving the claim unsupported.