Search papers, labs, and topics across Lattice.
University of Oxford
1
0
3
Training on mixed complexity datasets can yield up to 5x sample efficiency in low data regimes, challenging conventional wisdom about data quantity in LLM fine-tuning.
