Breakthrough papers improving LLMs performance

Pivotal advancements from February 16th–28th, 2025, redefining performance and efficiency in large language models

In partnership with

Highlights

• Enhanced Attention Mechanisms: Native Sparse Attention and SpargeAttn propose hardware-aligned and universal sparse frameworks that accelerate long-context inference and training without compromising performance.

• Knowledge Integration & Continual Learning: "How Do LLMs Acquire New Knowledge?" and "How Much Knowledge Can You P…