Search papers, labs, and topics across Lattice.
1
0
3
Predict how well your LLM will transfer to a new domain *before* fine-tuning, by using sparse autoencoders to spot tell-tale signs of domain shift in the model's representations.
