"be" after modal verbs

What this neuron does

This neuron lights up on the word "be" when it follows a helper word like "may", "will" or "would". Turning it down made the model repeat "not only not", so it seems tied to the negation words it can hold back.

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

   main point; and evidence substantiation.⏎You will need to understand the sentence’s intention and keep an eye out for brevity and succinctness.⏎You will be quizzed on all the usual suspects: commas, semicolons and col

2. #2 peak

   your children appropriate bathroom habits or want to know whether he or she may be hitting puberty too early? If you are a parent with an uncomfortable question, you might be worried about broaching sensitive topics for fear of being judged or embarrassed. Don’t

3. #3 peak

   adult teeth and can prevent complications and the need for more orthodontic therapy. Removable computer-generated appliances can treat selective cases in which orthodontics would otherwise be needed, but these appliances also may cost more. Removable appliances that use wires

4. #4 peak

   including any advertising you do for your business.⏎You may be proud to tell anyone who asks that you’ve processed $50,00

5. #5 peak

   devalue the perspectives, opinions and contributions of other team members. However, while individuals with a high subjective sense of power tend to dominate team discussions, they may be able to do so only if other team members permit it. Most likely, team

6. #6 peak

   inske2 IMAGE OF SCIENTIST)⏎Day2PMSchinske2 IMAGE OF SCIENTIST⏎In addressing this, however, faculty members may be reluctant to engage in frank discussions about race, gender, and identity with their students

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

   times in our history where the U.S. was just not the country it was supposed to be.⏎In that sense, the story of how our nation has come to be is a story of cultural achievement, and we owe that to the people who

2. #2 peak

   result, however, is<|endoftext|>The Getty Museum challenged people worldwide to recreate famous works of art using items from home.⏎Once again, Leah made my dream come to my life. Her chameleonic beauty makes it so easy to transform in

3. #3 peak

   then update the positions and velocities for the next round of calculations. This scheme of computation is known as the n-body method. Perhaps Newton himself could have put it to work if

4. #4 peak

   that has been hiding at the bottom of a lake ever since it rode to earth in a meteorite, validating the Panspermia theory of how life came to this planet?⏎Or does it lend credence to the idea that life could spring

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.