Edit Model Behavior via J-Space
Extraaltodeus · reddit · 2026-07-12
Based on Anthropic's open-source Jacobian-Lens, the author built a tool that can edit a model's J-Space and "export" the adjusted behavior into a model that maintains the same behavior.
He says the code works and he made an experimental model with notably "explicit" behavior, emphasizing research purposes. The post includes:
- A tool description based on the method
- A model Nikusui-v1 on Hugging Face
- A GGUF quantized version
The author wants community feedback. This is about model behavior editing / mechanistic interpretability toolchain / experimental model release, not casual chat.
More from Models
- Cohere Flags Qwen Sandbox Oddity; Security Researchers Call It FUD — basedjensen · 2026-10-27
- GPT-6.1 Terra Now Testable on LMArena in Battle and Agent Modes — arena · 2026-10-27
- OpenAI Launches GPT-6.1 Terra at $1/M Input Tokens, ~9x Cheaper per Task Than Opus 5.5 — VraserX · 2026-10-27
- GPT-6.1 Terra Benchmarks: Sweeps All Six Tests, SWE-bench at 76.2% — mark_k · 2026-10-27
- GPT-6.1 Terra lands: early users praise speed, value and subagent performance — BenBajarin · 2026-10-27
- Leaked GPT-6.1 Terra benchmarks: cheap but 'boring', still far behind Astra and Claude — Angaisb_ · 2026-10-27