During April 2025, this developer focused on enhancing documentation for attention backends in the yhyang201/sglang and Furion-cn/sglang repositories. They authored comprehensive guides detailing support matrices for features such as page size, spec decoding, MLA, sliding window, and multimodal capabilities, using Markdown and RST to ensure clarity and consistency. Their work included actionable user guides with launch commands for FlashInfer, FA3, Triton, and Torch Native, all integrated into the main documentation index to improve discoverability. By standardizing terminology and structure across repositories, they streamlined onboarding and provided a unified reference for users and contributors.