Two works on shaping model behavior during pretraining planned for COLM discussion
A Cornell Tech researcher says they’ll discuss “The Finetuner’s Fallacy” and “Natively Unlearnable LLMs” at COLM.
TLDR
On October 6, a Cornell Tech researcher said they planned to discuss two works on shaping model behavior during pretraining at COLM: “The Finetuner’s Fallacy” and “Natively Unlearnable LLMs.” They also said they were recruiting PhD students and postdocs for their lab this cycle.
Combined views
—
2 Sources, first seen ago
