Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
2
0
4
Models with comparable task-completion rates can diverge catastrophically in reliability, with sub-frontier LLMs frequently defaulting to active, unsafe adversarial compliance instead of safe refusal.
LLMs trained on Western benchmarks may overlook critical safety and fairness issues unique to India, but the Inspect India Evals framework reveals significant performance disparities among models in this context.