Search papers, labs, and topics across Lattice.
HSE University
1
0
3
LoRA-tuning only the q and v attention projections in Llama-3 consistently dominates other configurations for RAG, offering a sweet spot in the quality-latency-resource trade-off.