Search papers, labs, and topics across Lattice.
6
0
5
Overthinking in language models can be curtailed by segment-level credit assignment, leading to a 5.4% accuracy boost on competitive math benchmarks.
BINEVAL's binary question approach not only matches human evaluations but also reveals nuanced insights that traditional methods miss, paving the way for more interpretable LLM assessments.
SAFARI maintains diagnostic precision even when critical fault information lies five times beyond the model's context limits, a feat traditional methods cannot achieve.
Achieving a 4.48x speedup in throughput for multi-agent systems without sacrificing performance could revolutionize enterprise applications.
T1-Bench reveals that existing benchmarks fail to capture the nuanced interactions required for realistic agent performance, paving the way for more robust evaluations in AI systems.
HiViG outperforms existing critics by integrating historical context and visual grounding, achieving up to 9% higher success rates in complex GUI tasks.
