Search papers, labs, and topics across Lattice.
3
0
8
1
LLMs hit a "reasoning collapse" where accuracy plummets by over 50% on classical reasoning tasks once complexity exceeds a surprisingly low threshold, even with deterministic validation.
Stop prompt injections cold: PCFI's priority-aware runtime defense intercepts all attacks in testing with zero false positives and negligible overhead.
Perceptually "cleaner" audio, achieved through state-of-the-art denoising, can actually *harm* zero-shot ASR performance, challenging the assumption that better audio quality always translates to better recognition.
