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
尚未偵測到更新。GitHub 發布版本後會顯示在這裡。
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Open-source System 1 model: typed decisions with calibrated probabilities from one forward pass.
翻譯準備中
取自儲存庫 README
正在載入完整 README。 以下為已儲存的摘錄。
Kahn1 is an open-source System One model: ask typed questions about a text (Choice, Score, Noul) and get typed
answers, each with its full distribution and a calibrated confidence, from a model you run yourself.
Playground. Paste any text, write a JEV schema, get each answer with its distribution.
Snake. Kahn1 picks every move: about 130 ms a move at k = 3 in this recording.
Both recorded with Kahn1 4B ( Okura66/Kahn1-Qwen3.5-4B ) served locally with vLLM, k = 3 for Snake. The
in-browser versions on kahn1.com run a 4-bit GGUF of the 3B instead.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。
Kahn1 is an open-source System One model: ask typed questions about a text (Choice, Score, Noul) and get typed
answers, each with its full distribution and a calibrated confidence, from a model you run yourself.
Playground. Paste any text, write a JEV schema, get each answer with its distribution.
Snake. Kahn1 picks every move: about 130 ms a move at k = 3 in this recording.
Both recorded with Kahn1 4B ( Okura66/Kahn1-Qwen3.5-4B ) served locally with vLLM, k = 3 for Snake. The
in-browser versions on kahn1.com run a 4-bit GGUF of the 3B instead.檢視完整 README ↗