unclear: word before a noun in lists

What this neuron does

It fires on a scattered mix of words, often the word just before a noun in a list or a formal phrase, such as 'both' or 'two'. The pattern is loose and hard to pin down. Turning it up made the model drift into questions or repeat itself.

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

   ulopathy, and the acute coagulopathy of trauma-shock (ACoTS) [2,3,5,6].⏎The etiology, diagnosis and treatment of coagul

2. #2 peak

   and underscored the critical need for effective cooperation on border biosecurity measures.⏎Biosecurity is critical to protecting public health, ecosystems and a country’s economy. As countries become increasingly interconnected through global trade, travel and movement of

3. #3 peak

   and group practice. Instructors also administer quizzes and full-length practice sections of the SAT. This allows instructors to identify areas in which students need improvement and provides both instructors and students with . If students do not improve their scores it is because either did

4. #4 peak

   illustration of how one subject is put for another is found in comparing the two seemingly contradictory biblical accounts of the death of King Saul. Remember that in the Old

5. #5 peak

   production assistant. Each of these positions are defined along with their respective duties, rules, and resources that figure in obtaining these jobs. Each chapter offers exercises, links

6. #6 peak

   required materials, discussion of course policies, an overview of class assignments, and grading standards. Updated:

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

   label at least 85% of that wine must come from within that AVA boundary. The AVA designation on a wine label usually appears just above the grape varietal description. The AVA designation does not limit the types of grapes

2. #2 peak

   and the goals; and evaluations of the process itself and of the implementation and results of the action plan, with an eye toward improving MAPP’s effectiveness in the community as it’s

3. #3 peak

   two days after the hot weather. Night-time cooling can help reduce the negative effects of heat load on production, but it does not reduce the welfare impact on the cows during the day.⏎In the summer of 2021 the Waik

4. #4 peak

   five hours of instruction per week. Secondary school students who qualify for homebound instruction are eligible to receive a maximum of ten hours of instruction per week depending on their academic load. If appropriate to the needs of the child, the number of instructional hours

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.