Search papers, labs, and topics across Lattice.
Northeastern University
3
0
6
TRAM leverages the model's reasoning history to create a compact memory that boosts performance on complex reasoning tasks without extra training.
Seemingly harmless fine-tuning data can stealthily nudge LLMs toward unsafe behavior by subtly shifting model parameters in "danger-aligned" directions.
LLMs with induced personalities don't just *sound* different – they exhibit measurable and predictable cognitive performance changes, mirroring human psychology.
