https://lab.parano1d.org/zh/O(1) Lab 的论文、协议工作、可复现的工程研究及其成果。zh-HansSat, 01 Aug 2026 15:58:14 GMThttps://lab.parano1d.org/zh/research/parano1d-soundness-industry-metrics/https://lab.parano1d.org/zh/research/parano1d-soundness-industry-metrics/Fri, 31 Jul 2026 12:00:00 GMT证明可靠性使用 Plonky2、RISC Zero 与 ethSTARK 发布的指标评估实际部署参数，并将基于猜想的评分与由有限定理支撑的上界明确分开。https://lab.parano1d.org/zh/research/frost-gkr-global-trace-protocol/https://lab.parano1d.org/zh/research/frost-gkr-global-trace-protocol/Fri, 31 Jul 2026 12:00:00 GMT证明系统三列承诺轨迹只用两次约束 sumcheck，即可归约整批 Poseidon2b 执行。https://lab.parano1d.org/zh/research/b64-b255-proof-classes/https://lab.parano1d.org/zh/research/b64-b255-proof-classes/Fri, 17 Jul 2026 12:00:00 GMT递归证明HistoryStep 是区块状态转换的递归证明。B64 可处理最多 64 个用户交易页面，并能在参考笔记本上满足准备时间预算；B255 将同一证明关系扩展到每区块 255 页的上限。测量没有发现第三种类别可以覆盖的有用运行区间。https://lab.parano1d.org/zh/research/atomic-history-step/https://lab.parano1d.org/zh/research/atomic-history-step/Wed, 15 Jul 2026 12:00:00 GMT共识早期设计先接受区块，再在后台生成对应的递归证明。因此，规范链尖可能领先于已被 HistoryStep 覆盖的链尖，节点还必须持久化并恢复另一条任务队列。正式架构只在区块与其 HistoryStep 同时就绪时接受该区块：一个原子对象、一个规范高度，不存在滞后的证明积压。https://lab.parano1d.org/zh/research/prove-once-search-nonces/https://lab.parano1d.org/zh/research/prove-once-search-nonces/Wed, 15 Jul 2026 12:00:00 GMT挖矿HistoryStep 在不含 nonce 的语义域中绑定区块头的全部语义字段；PoW 可以改变 nonce，而无需重建 State 转换证明。https://lab.parano1d.org/zh/research/proof-carrying-state-sync/https://lab.parano1d.org/zh/research/proof-carrying-state-sync/Sun, 12 Jul 2026 12:00:00 GMTProof-carrying State新接入节点验证区块头与最终性边界上的一份递归转换证明，安装精确 Live State，并应用不超过十八个近期区块。https://lab.parano1d.org/zh/research/utxo-incarnations/https://lab.parano1d.org/zh/research/utxo-incarnations/Fri, 10 Jul 2026 12:00:00 GMTLive State在现有数值字段中加入单调创建标识符，可使过期的槽位打开证明失效，而无需引入第二个共识根或设置槽位重用隔离期。https://lab.parano1d.org/zh/research/binary-field-kernels/https://lab.parano1d.org/zh/research/binary-field-kernels/Thu, 09 Jul 2026 12:00:00 GMT二进制域PCLMUL、VPCLMUL、AVX-512 与 PMULL 执行完全相同的 GF(2¹²⁸) 运算，不改变见证布局或证明字节。https://lab.parano1d.org/zh/research/single-global-walk/https://lab.parano1d.org/zh/research/single-global-walk/Mon, 06 Jul 2026 12:00:00 GMT递归证明完整组合基准表明，独立验证器遍历会把内存推高到 30 GB 以上；共享域消除了这一倍增因素。https://lab.parano1d.org/zh/research/recursive-fixed-point/https://lab.parano1d.org/zh/research/recursive-fixed-point/Sun, 05 Jul 2026 12:00:00 GMT递归证明采用 2^23 个位置约束域的类别实现递归闭合：每份证明都以相同的认证矩阵与公共布局验证前驱。https://lab.parano1d.org/zh/research/authstream-proof-compiler/https://lab.parano1d.org/zh/research/authstream-proof-compiler/Sun, 28 Jun 2026 12:00:00 GMT证明工程流式编译以 4.7% 的开销约束了 14–16 MiB 私密授权数据；所评估的终端证明未达到实际部署的延迟预算。https://lab.parano1d.org/zh/research/merkle-fusion-negative-result/https://lab.parano1d.org/zh/research/merkle-fusion-negative-result/Thu, 25 Jun 2026 12:00:00 GMT证明工程融合两个 Merkle 关系节省了 5.94 KB，却扩大了组合后的多线性轨迹并失去并行性；证明时间从 2.68 秒退化到 4.50 秒。https://lab.parano1d.org/zh/research/owner-batched-authorization/https://lab.parano1d.org/zh/research/owner-batched-authorization/Wed, 24 Jun 2026 12:00:00 GMT钱包授权公开交易运算属于区块关系。即使一次花费包含 1,020 个输入，钱包也只证明一个私密事实——所有权。https://lab.parano1d.org/zh/research/binding-a-mixed-opening/https://lab.parano1d.org/zh/research/binding-a-mixed-opening/Sun, 21 Jun 2026 12:00:00 GMT多项式承诺A/A′ 对抗测试把低度一致性与承诺来源区分开来；来源根与共享查询补上这条绑定链。https://lab.parano1d.org/zh/research/pcs-implementation-trail/https://lab.parano1d.org/zh/research/pcs-implementation-trail/Sun, 21 Jun 2026 12:00:00 GMT多项式承诺Ladder FRI 暴露了证明大小成本，FRI-Binius 暴露了来源绑定问题，BaseFold 则暴露了把验证器放进后继证明的代价。