pieces of split scientific or place names

What this neuron does

This neuron fires on the middle chunks of split names, especially Latin species names like 'Dios-py-ros' or 'Ps-amm-omys' and place names like 'To-hoku'. Turning it up made the model produce unrelated text, so the species link is only visible in the examples.

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

   2) (Learn how and when to remove this template message)|⏎Sendust is a magnetic metal powder that was invented by Hakaru Masumoto at Tohoku Imperial University in Sendai, Japan, about 1936 as

2. #2 peak

   the fall. Japanese maple can be grown in Southern California, but trees must be in a protected location and watered regularly.⏎Japanese Persimmon (Diospyros kaki): This is a fruit-bearing tree that grows to 30

3. #3 peak

   North American diving duck related to the goldeneye, with a large puffy head. The male has white plumage with a black back.⏎- ‘Neophytes at duck identification often confuse a bufflehead with a male hooded merg

4. #4 peak

   x-ray images of small animal spines identifying the effects of disc degeneration. The age-related disc and spine degeneration that occurs in the sand rat (Psammomys obesus) has previously been documented radiologically; selected representative radiographs

5. #5 peak

   const void * memchr ( const void * ptr, int value, size_t num );⏎void * memchr ( void * ptr, int value, size_t num );⏎Locate character in block of memory⏎Searches within the first num bytes

6. #6 peak

   or if it were truly indicative of Homer’s personal psychosomatic beliefs, it’s difficult to argue that his approach to storytelling<|endoftext|>Oak spider mites (Oligonychus bicolor) may be hard to see with the naked eye but

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

   by the routine. The template can consist of:⏎- An introductory paragraph highlighting the problem and promising a solution;⏎- Body of text with several tips in a logical order;⏎- The final paragraph where the author encourages the reader to follow the advice.⏎

2. #2 peak

   damage (hepatitis, cirrhosis, tumors);⏎- bowel diseases in violation of protein digestion and absorption of amino acids (inflammations, post-operative conditions, paras<|endoftext|>Alice’s Adventures in Wonderland⏎The classic tale in the realm of

3. #3 peak

   is harmful to the eyes since it increases the risk of cataracts and ARMD;⏎- Wear sunglasses with UV protection - ultraviolet radiation is harmful to the eyes;⏎- Monitor eye health regularly<|endoftext|>Introduction to Machine Learning with R: Rigorous

4. #4 peak

   first and second control signal generating means for generating a first error signal (e.g. e3) proportional to the algebraic difference between said first and second control signals for input into said power management control circuit;⏎means connected to the output of said

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.