adjectives right before a noun

What this neuron does

This neuron fires on a single descriptive word, like 'lengthy', 'ordinary' or 'strange', that comes right before the thing it describes. Turning it up made the model's writing drift into random topics, so it does not seem to change the style much.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

1. #1 peak

   lengthy work experience in space. As part of three space missions, he has spent

2. #2 peak

   insulting when we imagine countries outside the West have something unique in their history where they crave

3. #3 peak

   refine, and establish clinical uses for liquid biopsy tests.⏎Different liquid biopsy<|endoftext|>Biology

4. #4 peak

   undeniable that teaching mathematics in the bilingual classroom, when the language of learning and teaching (

5. #5 peak

   ordinary panel. This manufacturing method significantly reduces the weight of the concrete panel, without any

6. #6 peak

   ordinary)⏎→ The two countries have resumed normal diplomatic relations.⏎→ Traffic was normal for

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

1. #1 peak

   patient’s ‘kidney history’ and leads into a description of the natural history and

2. #2 peak

   Server, loading data onto the LASR Analytic Server, and manipulating data on the

3. #3 peak

   Math and computer science teacher Myra Deister of Sunny Hills High in Fullerton didn

4. #4 peak

   server. Examples stock feeds, system status, weather conditions or other monitoring applications. This

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.