gghfez

The backend is https://github.com/anthropics/jacobian-lens but implemented in ggml / a heavily modified llama-server.

The frontend is just a vibe-coded html file.

>Every AI model has the same “controversial” take on humans

They only tested 12 models, all very recent.

This is MiMo-2.6-Flash MXFP4 with a jacobian lens with their prompt:

Prompt:

  <|im_start|>user
  What is your most controversial take about humans? Give your own honest view, not a survey of opinions, and don't hedge.<|im_end|><|im_start|>assistant
  <think></think>

Response:

  Humans are not primarily rational animals driven by logic or truth; they are meaning-making creatures whose fundamental drive is to construct narratives that grant their lives significance.

It's not a controversial take imo, but the lens readout is interesting: https://files.catbox.moe/08giiv.png

>the tokenization discord

Could I join this?

I've been building control vectors for creative writing for a while now (models like Gemma-4, Kimi-K2.6, GLM-5). Concept axes like optimism/nihilism, simple/ornate language, empathy/sociopathy.

The "just use prompting" argument breaks down for stylistic and tonal shifts. You can prompt a model to "write in an ornate style" but you're fighting the trained-in distribution, and the instruction competes with your actual creative intent for the model's attention. A control vector shifts the distribution directly without consuming any context window.

The more surprising finding has been cross-domain transfer. Vectors trained on creative writing concepts visibly affect other domains. An optimism/nihilism vector changes the color palette and mood of generated SVGs with no mention of mood or style in the prompt. Same prompt, completely different output.

Model cards with samples (SVG generation, short stories) and downloadable vectors for llama.cpp:

https://huggingface.co/gghfez/Kimi-K2.6-control-vectors

https://huggingface.co/gghfez/gemma-4-31b-it-control-vectors