Search papers, labs, and topics across Lattice.
Hanyang University
1
0
4
1
Forget full-history prompting: this work shows you can slash token costs by 98% while boosting tool-calling accuracy by explicitly modeling and refining latent user preferences.
