GGUF Version of Jacobian Lens Visualization Tool
Responsible_Fig_1271 · reddit · 2026-07-12
The author released an interactive Jacobian-Lens visualization and steering tool for GGUF / llama.cpp: jlens-gguf. He noted that while there are many jacobian lens projects for HF and PyTorch, there are almost no implementations supporting GGUF or llama.cpp, so he built this project using Fable 5 under manual supervision.
Main Features
- Provides a native GGUF server based on llama.cpp, synced with the upstream git
- Supports model observation, as well as j-space swapping / abliteration / steering
- Can also connect to a running llama-server model for observation, but cannot steer
- Compatible with both dense and MoE GGUF
Resources & Implementation Info
- Repository: igorbarshteyn/jlens-gguf
- Inspired by Anthropic's Global Workspace / Jacobian Lens paper and code
- Also provides Anthropic's original code and Matthew Berman's explainer video
Additional Notes
- The memory overhead of this lens grows at roughly 1/8 of the model's size
- For example, a model around the 160GB scale requires an additional 20GB RAM to host the lens
More from coding & agent
- Spring AI 2.1.0-RC1 adds tool-call confirmation, pausing agent loops before real-data access — therealdanvega · 2026-10-26
- Morgan Stanley and US banks block third-party AI agents over credentials — thione · 2026-10-26
- METR red-teams OpenAI's agent monitoring, finds bypasses and gaps — thione · 2026-10-26
- OpenAI: coding agents weakened safety monitoring, one lied about it — thione · 2026-10-26
- ConvFill: A 0.6B Talker Buys Time While a Large Reasoner Whispers the Real Answer — andimarafioti · 2026-10-26
- 250 Builders Hack With Codex at Berlin Community Hackathon Ahead of OpenAI DevDay — paw_lean · 2026-10-26