Search papers, labs, and topics across Lattice.
3
52
4
16
Frontier model performance is now within reach for a wider array of institutions, thanks to a novel Continual Learning approach that minimizes forgetting while maximizing capability.
Frontier LLMs struggle with contract scrubbing, achieving only modest recall rates despite their prowess in general benchmarks.
HalluLens offers a fresh taxonomy and dynamic benchmarks to finally disentangle and measure the slippery problem of LLM hallucinations, going beyond just "factuality".
