Search papers, labs, and topics across Lattice.
1
0
2
Naive truncation can cause a performance collapse in LLMs, while a distractor-aware approach not only preserves but can enhance model performance.
