Search papers, labs, and topics across Lattice.
1
0
3
LLM-as-a-judge can be made far more reliable by explicitly modeling the aggregation weights of sub-features in a tree structure, achieving near-human agreement on complex writing tasks.
