neuron-explorer
Steering-first neuron explorer for Qwen3.5-4B: push every MLP neuron, keep every rollout, browse the results
翻譯準備中
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證具版本控制的向量索引,支援 MVCC、快照隔離與可分支的代理記憶體。
AI 翻譯
Versioned vector index with MVCC, snapshot isolation, and branchable agent memory
取自儲存庫 README
正在載入完整 README。 以下為已儲存的摘錄。
Most vector libraries optimize for a static corpus. Real systems do not stay still: agents write memories continuously, documents get corrected, users invoke the right to erasure, and teams need to reproduce what a retriever saw last Tuesday.
ChronoVec is an approximate-nearest-neighbour index built for that case. Every record carries a version interval, so queries can read the index as of any past moment, deletions physically reclaim space on a bounded budget, and continuous writes never force a rebuild or block a reader.檢視完整 README ↗目前無法載入完整 README。請透過原文連結閱讀。 以下為已儲存的摘錄。
Most vector libraries optimize for a static corpus. Real systems do not stay still: agents write memories continuously, documents get corrected, users invoke the right to erasure, and teams need to reproduce what a retriever saw last Tuesday.
ChronoVec is an approximate-nearest-neighbour index built for that case. Every record carries a version interval, so queries can read the index as of any past moment, deletions physically reclaim space on a bounded budget, and continuous writes never force a rebuild or block a reader.檢視完整 README ↗