Search papers, labs, and topics across Lattice.
3
0
5
CRANE achieves a remarkable 96.9% Grounded Success in knowledge editing for reasoning MLLMs, overcoming traditional failure modes that plague existing methods.
LLM unlearning via counterfactual tuning can backfire, increasing hallucination rates in unexpected areas due to inconsistencies in the "fake" knowledge it's trained on.
LLM knowledge editing often fails because the "fix" only works when you ask about it in the same way it was trained, revealing a surprising brittleness in how models incorporate new information.
