the 'st' piece inside words

What this neuron does

This neuron fires on the 'st' piece in the middle of a word, most clearly in 'Mon-st-era' (the plant Monstera), and also in words like 'foodstuffs' and 'stomach'. Turning it up gave text with repeated words like 'pretty', and turning it down gave repeated 'EMH' and 'one's own', so the writing change does not clearly match the trigger.

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

   –⏎- Prune the plant strategically:Cutting the plant irregularly without any proper knowledge of how it grows, may stop the further growth of the Monstera. Therefore, the first thing that we should keep in mind while applying growth

2. #2 peak

   because it generally lights upon its feet without injury, the foot and toes being padded so as to break the fall. (See NINE.)⏎Tub What wouldst thou have with me?⏎Mer. Good king of cats, nothing but one of

3. #3 peak

   grow effectively in water:-⏎Ways to care for Monstera and make them grow faster:-⏎The Monsteras are widely growing plant types

4. #4 peak

   imported foodstuffs from the empire, of course, but also from Western Europe, Scandinavia

5. #5 peak

   then values in this “generalized<|endoftext|>The most popular plant, is a tropical plant that propagates through cutting and re-shaping of it. When a Monstera begins to grow, its leaves start sprouting in all ways possible. The

6. #6 peak

   the lining of the stomach, so there is a risk of gastrointestinal side effects such as indigestion and stomach aches.⏎You might also develop a gastric (stomach) ulcer and need to start taking proton pump inhibitors like omepr

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

   bike, walk or take public transportation.⏎Metro is working with regional partners to increase the region’s effectiveness in development and secure funding for the region’s network of on-street and off-street bikeways and walkways integrated with transit and supported by

2. #2 peak

   by the U.S. military, with the purpose of keeping up with the Jones-niks. Technologically.⏎So it's no surprise that surviving-a-nuclear-strike would become an objective for the boys at DARPA.⏎That's where the

3. #3 peak

   : http://www.fbi.gov/stats-services/crimestats/ by the FBI Crime Statistics Division, explore trends in crime using one of

4. #4 peak

   canal. Water levels in the Corbett Area declined in response, and the District coordinated with the Indian Trail Improvement District to optimize drainage in the area.⏎A post-storm analysis shows the District moved an estimated 105 billion gallons of water

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.