unclear: word after 'just', 'full' or 'whole'

What this neuron does

This neuron is hard to pin down. It often lights up on the word that finishes a set phrase like 'just war', 'full swing' or 'full houses', and it goes quiet or negative on 'only to' as in 'only to discover'. Turning it up did not produce any clear change in what the model wrote.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

   <|endoftext|>“There shall never be any bond slaverie, villinage or Captivitie amongst us unles it be lawfull Captives taken in just warres, and such strangers as willingly selle themselves or are sold to us.

2. #2 peak

   weapons—to commit such crimes. (CCC 2314; cf. Gaudium et Spes 80)⏎Under the Catholic understanding of just war, not only must the cause of

3. #3 peak

   , Our Explosive Sun and The Northern Lights: A Guide. One of the foremost scholars on the subject, Mr. Brekke has spoken multiple times to full houses at the Smithsonian and elsewhere. In this interactive session,

4. #4 peak

   read in a sustained way, then a chunk of his schooling should be focused on helping him squeeze meaning and richness out of the experience. You may remember the whole-language ideas about exposing kids to print through fiction, poetry, newspapers, and drama?

5. #5 peak

   news HTTP/1.1 Host: www.example.ai Accept-Language: fr⏎HTTP/1.1 <|endoftext|>The move to electronic health records is in full swing for most of the medical community, but holdouts point to security issues as an

6. #6 peak

   bu Anne mau makan di KFC.⏎Harimau tidak mau berlari ke rumah.⏎Perempuan muntah pohon.⏎Students then choose words from the above lists to make a sentence

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

   the festivities. But when a blizzard shuts down the highway, Phil finds himself trapped in Punxsutawney. He wakes up the next day, only to discover that it's not the next day at all. It's Groundhog Day

2. #2 peak

   the most efficient processing of the data; there is an old joke we like to tell our students about a meteorologist who wrote a weather prediction program, only to find out that it took one week to predict tomorrow's weather! (He later

3. #3 peak

   influential funder who told me told me that parents were a sinkhole, only to tell me a bit later that parents were essential. Whiplash!⏎Funny thing

4. #4 peak

   912, stretched a canvas for it in 1913, and returned to the composition late in 1915, only to start again on a new canvas in early 1916. Black is

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.