Worked on the sophgo/LLM-TPU repository to enhance documentation for the MiniCPM3 and Qwen2_5 models, focusing on clarifying export instructions and streamlining the onboarding process. Applied documentation best practices using Markdown to remove ambiguous CPU precision notes, recommend CUDA export, and direct users to dedicated multi-core model compilation guides. Emphasized cross-repository consistency and clear instructional writing, ensuring that users could more easily configure and export models without confusion. Leveraged version control to maintain traceability of changes, resulting in more accessible and reliable documentation. No bugs were reported or fixed during this period, with all efforts concentrated on documentation improvements.