Chain-of-thought prompting primarily helps on tasks involving mathematical, logical, or algorithmic reasoning, with limited benefits on other types of reasoning tasks.
A decoder-only transformer model, LifeGPT, can accurately simulate the dynamics of Conway's Game of Life on a toroidal grid without any prior knowledge of the grid size or boundary conditions.
Möbius Attention, a novel attention mechanism based on Möbius transformations, enhances the expressivity of Transformer models by enabling them to capture more intricate linguistic patterns and relationships between tokens.
The proposed LogoRA framework employs a two-branch encoder to extract both local and global representations from time series data, and utilizes various alignment strategies to learn domain-invariant features for robust time series classification across different domains.
panCAKES, a novel hierarchical compression algorithm, enables efficient, exact k-NN and ρ-NN search on compressed data by leveraging the low-dimensional structure of the data.
제안된 RD3G 알고리즘은 에이전트들이 보상과 상태 제약을 통해 결합된 문제에 대한 국소 내쉬 균형을 찾는다.
The proposed RD3G algorithm is a novel Newton-based method that efficiently solves constrained multi-agent game-control problems by partitioning the inequality constraints and removing dual variables associated with inactive constraints to reduce the scale of the linear problem in each iteration.
我們提出了第一個在亞線性時間內計算 LZ77 分解的算法,打破了近50年來一直存在的線性時間障礙。
MAGICORE, a framework for Multi Agent Iteration for Coarse-to-fine Refinement, improves Large Language Model reasoning by adaptively applying coarse-grained aggregation or fine-grained, iterative multi-agent refinement based on problem difficulty.
본 논문은 Lempel-Ziv (LZ77) 요인화를 선형 시간 이하로 계산할 수 있는 최초의 알고리즘을 제시한다.