Search papers, labs, and topics across Lattice.
KAIST
4
0
8
5
Small LLMs can achieve up to 27.2% accuracy gains by leveraging hierarchical memory from larger teacher agents, reshaping how we think about agent training.
Naively scaling training data for small computer-use agents yields marginal improvements, but targeting training on weaknesses identified by a stronger agent yields substantial gains.
Forget task-specific code – coding agents get a 3.7% performance boost by sharing *meta-knowledge* like validation routines across diverse coding domains.
Current AI models fall short when asked to understand a situation from the combined perspectives of multiple embodied agents, as revealed by a new challenging benchmark.
