陈默涵实验室 https://1cc1.com/ 国产算力与 AI 推理的独立实测：只写有原始数据、能复现、经得起复核的结论。 zh-CN Sat, 17 Oct 2026 15:36:36 GMT <![CDATA[LLM 推理性能的评测口径：为什么 token/s 会骗人]]> https://1cc1.com/reports/inference-benchmark-metrics/ https://1cc1.com/reports/inference-benchmark-metrics/ Sat, 17 Oct 2026 00:00:00 GMT <![CDATA[可复现的推理评测：一份复现包应该包含什么]]> https://1cc1.com/reports/reproducible-benchmark-kit/ https://1cc1.com/reports/reproducible-benchmark-kit/ Fri, 16 Oct 2026 00:00:00 GMT <![CDATA[为什么我坚持公开测试数据与脚本]]> https://1cc1.com/reports/why-open-data/ https://1cc1.com/reports/why-open-data/ Thu, 15 Oct 2026 00:00:00 GMT