Anthropic 用 Claude 为 123 个开源库的 239 个入口点生成证明。上线评审仍要核对证明命题、部署工件和残余风险,避免把有边界的证明误读为整库无漏洞。
Anthropic's 123-library experiment shows that LLMs can scale proof writing. A deployment assurance manifest shows what was proved, what ships, and wha
Claude 优化了八类量子电路任务。本文用成本与验证账本拆开 TQ 收益、失败预算、搜索成本、硬件映射和复现状态。
Claude improved eight quantum circuit tasks. A cost-to-verification ledger shows what TQ gains prove, where costs move, and reproduction gaps.
Claude 将 NIST 曲线 seed 追溯到时间戳驱动的 lrand48。本文用五级证据栈拆开可复算机制、历史归因与作者意图。
Claude traced NIST curve seeds to lrand48 timestamps. This evidence audit separates reproducible mechanism, historical attribution, and intent.
Claude formalized Fermat's Last Theorem in Lean in 11 days. Here is what the 13-million-line proof and its verification stack establish.
Claude 能自动改进十类对齐失败。真正值得复用的成果,是一套能发现过拟合、能力退化、作弊和外推失效的评测合同。
Automated alignment researchers need more than strong scores. Anthropic's experiment shows the evaluation contract that makes gains testable.
Claude 将临界线上 ζ 函数零点的无条件下界提高到 67.25%。本文拆解结果边界、Lean 证明和完整验证链。