Search papers, labs, and topics across Lattice.
2
0
5
55
ContextRL reveals that fine-grained context selection can lead to substantial performance boosts in LLMs, outperforming traditional data augmentation methods.
LLM win rates in multi-agent games can nearly double (from 25% to 50%) simply by optimizing the context provided during inference.
