word fragments like ev and v

What this neuron does

This neuron fires on a short fragment of a longer word that is split in two, such as the "ev" in "evoked" or the "v" in "nightvision". It is strongly negative on a capital V inside a name, such as "Vespasianus" or "Vjosa". Turning it down or up made the model repeat phrases like "relative relative" or "the condition was", so the steering shows no clear meaning.

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

   radical idea of powerful, independent women living in exotic lands evoked ambivalent emotions in the Greeks: awe, fear, respect, and desire

2. #2 peak

   Hau.⏎While slow-speed light is now confined to the realm of laboratory research, Hau believes future practical applications include improved communications technology, television displays and night-vision devices.⏎Tags: Lene Vestergaard Hau, physicist,

3. #3 peak

   , is the 19th-century sheet music for Handel's Acis and Galatea. The Greek mythological origin of the story is evoked on the front cover, which features a couple in classical dress in a Mediterranean

4. #4 peak

   Dussehra-Vijaya Dashami-Dasara Festival Story, Why Do We Celebrate D

5. #5 peak

   on things that have meaning to us now, or that will have some value for us in the future.⏎- The Positive Principle – which coins that appreciative inquiry evokes positive emotions like hope, inquisitiveness, and motivation, all of

6. #6 peak

   channels in muscle cells that control the flow of calcium ions, creating muscle contractions. Normally, electrical stimulation (“excitation”) of isolated muscle fibers under experimental conditions evokes a muscle contraction, a phenomenon known as “excitation-contraction coupling

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

   20px Jews of Judea⏎|Titus Flavius Vespasianus||Simon Bar Giora|| John of Gischala

2. #2 peak

   idea what they are destroying. There is nothing in Europe remotely like this river system.”⏎From the mountains of Greece where it is known as the Aoös, the Vjosa flows for 270km to the Adriatic Sea, reaching

3. #3 peak

   Saint-Galmier (Loire).⏎Seven years later, he found a partnership with Bessac and the painter Joséphus Veyrat. Then, he set up his own business in Grigny,

4. #4 peak

   is free to browse the archives for truly amazing photos of celestial artwork. I’ve picked 10 of my favorite to showcase here.⏎Captured by Stephane Vetter, this image “was taken two months ago over Jökulsárl

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.