Skip to content
View Anhao1314's full-sized avatar
  • 19:22 (UTC -12:00)

Block or report Anhao1314

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Anhao1314/README.md

Eason

中文 · English · 视觉主页 ↗

独立研究者。研究一个问题:**复杂系统的结论,如何不冒充现状。**研究假设展开见 life-finds-a-way。

人负责定义值得解决的问题。AI 负责探索道路。真实证据决定保留什么。

🔭 观察点 2026-08-01 → 2026-11-05

声明台账时间轴:2026 年 8 月至 11 月的变更事件、观察窗与现在游标

主线

ClaimLedger — AI Agent 的决定台账:声明 + 状态 + 来源;冲突写入必须显式替代或提出争议,--at 重建任意时刻的认知。

v0.10.0 · macOS / Linux · 纯标准库 · 基准报告 · 发布

基准是合成的(40 会话 / 24 查询):样本小,勿外推;冲突类 Recall@5 0.50。不做向量库、自动记忆捕获、守护进程、云同步。

声明台账

声明 证据 · 自 状态
ClaimLedger:过期结论不再冒充现状 24 合成查询,过期命中 54%→0 · 2026-10-18~ ◐
FCode:模型只提案,事件落哈希链 实验中 · 2026-10-10 ○
lastgreen:完成需要「最后一次绿」的证据 v0.1 试点 0 胜 9 平 1 负,未证明 · 2026-10-20 ✕
go2w-locomotion:关节级 RL 可达实物 Isaac Lab 三种子达标,sim2real 未开始 · 2026-10-27 ◐
g1-jev-swarm-lab:可支撑长时程任务 仍是目标而非结果 · 2026-10-06 ○
RL-Sentinel:训练失败可回放复查 工具可用,含规则漏检负结果 · 2026-08-17 ◐
flowcredit:风险结论可校准 确定性规则承载,PD 与 EL 仍为空 · 2026-09-07 ◐
flowcredit-research:投研观点可追溯 盲测未通过,归档在案 · 2026-10-15~ ✕

检验记录

失败也是证据:没复现的、没通过的、证据不足的结果保留归档,按原样标注。

  • lastgreen v0.1 10 组配对 0 胜 9 平 1 负 → 未证明;v0.2 的 12 次消融未运行;
  • flowcredit-research 盲测未通过 → 归档;
  • go2w / g1 结果都在仿真中,尚无域随机化与实物验证——仿真结果不等于实物可靠。

现在

  • 📍 在做 ClaimLedger v2 争议闭合、双时态与依赖回溯已合入 main,未进发布;
  • 未做 go2w 的 sim2real 验证;g1 的长时程任务;
  • 习惯 小工具与创意编程保持试错的手感,不进正式仓库;
  • 📮 给 30 天后的我 只信带日期与来源的行;其余是草稿。

校验

CI 用 test_profile.py 校验本页:声明集合与状态记号覆盖两份文件,章节顺序、宪法句与口径在此断言。检查器只验证结构与在场性,不验证真伪——结论的真伪,在链接里。

个人研究系统,非生产系统。未立声明者不上墙;带 ~ 为占位日期,待接入 ClaimLedger。

状态记号:● 已复现 · ◐ 有证据·样本小 · ○ 未证明 · ✕ 已证伪·在案 · ⚡︎ 争议中 —— ● 至今无一行适用。

🌱 生命会自己寻求出路

Pinned Loading

  1. go2w-locomotion go2w-locomotion Public

    分层具身智能受控验证:四足 RL 训练 → 跨引擎迁移 → ROS2 真 DDS 闭环 → VLA 模仿学习(预注册协议 + 证据链)

    Python 8

  2. flowcredit flowcredit Public

    Risk intelligence infrastructure for AI-native businesses and agents — evidence-aware assessment, deterministic TAI/CCI, integrity veto and versioned APIs.

    JavaScript 2 1