Search papers, labs, and topics across Lattice.
2
0
4
0
System prompts can be optimized to ensure equitable response quality, reducing worst-case information loss by over 13% while maintaining average performance.
General-purpose helpfulness metrics fail to reliably signal effective pedagogy in LLM tutoring, revealing a critical gap in evaluation methods.
