Search papers, labs, and topics across Lattice.
Affiliation:
3
0
6
12
CRISP recovers up to 28% improvement on retrieval tasks while achieving a remarkable 5.30x speedup in attention computation for long-context LLMs.
LLMs show a stark performance drop in realistic repository-level unit test generation, revealing hidden challenges in their deployment across diverse programming languages.
By explicitly modeling code structure and dependencies, Hydra lets smaller code LLMs beat much larger models on repository-level code generation tasks.
