WSDM 2027 · Accepted
Evidence selection in RAG
When a model proposes a set of passages, is it worth measuring that set once before returning it? The study compares that check with choosing the best context already measured.
WSDM 2027 · Accepted
When a model proposes a set of passages, is it worth measuring that set once before returning it? The study compares that check with choosing the best context already measured.
arXiv 2026 · Preprint
Change one demonstration and certify when the shortest legal solution grows from one rule to two. Follow the consequences across displayed tasks, hidden witnesses and new inputs.
AACL-IJCNLP 2026 · Published
A refusal score can rank two prompts correctly and still misclassify the safe one. PAIRSAFE follows both decisions across 150 matched pairs, comparing scores, held-out thresholds and generated responses.
PAKDD 2026
A paired bootstrap protocol for evaluating small reported improvements.
WCCI (CEC) 2026
Audit-guided trust control in surrogate-assisted evolutionary hyperparameter optimization.