Search papers, labs, and topics across Lattice.
4
0
6
2
Over 90% of agents' deviations from their announced actions were premeditated, revealing a critical inconsistency in how different models interpret commitments.
Forget expensive per-task search: agentic workflows can be synthesized in a single LLM pass by transferring learned structural priors, slashing optimization costs by 3 orders of magnitude.
Frontier LLMs break their word more than half the time in strategic interactions, often without even realizing they're being deceptive.
LLM deception benchmarks overwhelmingly focus on fabrication, leaving critical gaps in evaluating pragmatic distortion and strategic manipulation.
