Search papers, labs, and topics across Lattice.
4
0
7
6
This work defines the scaling inflection point as the per-session budget where marginal Elo gains match the independent-sampling reference, and proposes Elo-per-token analysis, which tracks the best solution found at each token budget and uses a Bradley-Terry model to aggregate within-task orderings into Elo ratings across tasks with different score scales.
Achieve 17x faster video object tracking without sacrificing accuracy by intelligently sampling only the relevant tiles in each frame.
Scaling prompt learning by 17x without sacrificing accuracy is now possible, unlocking efficient self-improvement for LLM agents.
Get 2x faster video generation from diffusion transformers without sacrificing quality, thanks to a clever parameter-free error compensation technique.
