Search papers, labs, and topics across Lattice.
16 papers published across 0 labs.
Latent LLM personas persist in weights post-training, readily resurfacing with specific prompts, and standard unlearning methods are insufficient to remove them without performance degradation.
Standard information retrieval methods can match LLM-mediated skill selection for agents, slashing per-task costs from $51 to $27.
Distilling real GPT-4o-labeled reports, not synthetic data, enables open-weight models to match proprietary LLMs for high-value clinical entity extraction.
Forbidden capabilities can still be learned even when suppressed in current outputs, as internal gradients enable future acquisition, but gradient sealing effectively blocks these pathways.
Latent LLM personas persist in weights post-training, readily resurfacing with specific prompts, and standard unlearning methods are insufficient to remove them without performance degradation.
Standard information retrieval methods can match LLM-mediated skill selection for agents, slashing per-task costs from $51 to $27.
Distilling real GPT-4o-labeled reports, not synthetic data, enables open-weight models to match proprietary LLMs for high-value clinical entity extraction.
Forbidden capabilities can still be learned even when suppressed in current outputs, as internal gradients enable future acquisition, but gradient sealing effectively blocks these pathways.
