blank spot before a number

What this neuron does

This neuron fires on the empty space just before a number, such as in "fallen from 50,000" or "more than 100,000". It is about the model expecting a quantity to come next. When it was turned up, the model started writing out many numbers and counts, like "2000 people" and "10000".

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

   make the return dry season journey in June, to take advantage of the permanent water of the Tarangire River.⏎Numbers of wildebeest have fallen from 50,000 to 6,000 in the past

2. #2 peak

   facial recognition technology. San Francisco and Virginia have imposed limits on facial recognition, but the EU’s prohibition would apply to 27 nations with a population of over 447 million people.⏎How EU AI Act will affect people?⏎By requiring

3. #3 peak

   site overlooking the confluence of the rivers Irwell and Medlock.⏎This was a timber fort with earthen ramparts for an auxiliary cohort of infantry of around 500 men.⏎The fort went through several phase of rebuilding and around

4. #4 peak

   a third factor is that new businesses are hiring fewer employees than in the past. In the 1990s, new establishments opened with an average of 7.5 jobs. By 2010, the number of jobs

5. #5 peak

   . At about the same time, Jewish refugees from Bohemia and Germany founded a community in Pressburg. In the fourteenth century the Pressburg community numbered close to 800 people and seems to have been

6. #6 peak

   , China with its population of 1.3 billion people gets 1,900 electoral-college votes, but even tiny Andorra with less than 500,000 people gets three electoral-college votes. Anyone

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

   Commission, Washington, DC.⏎Given that many primaries were won by only a plurality of the vote, how do these primary winners then fare in general elections? Overall, plurality

2. #2 peak

   the protests altogether. But obviously, these demonstrations caused a decrease in social distancing among protesters.⏎"Our findings suggest that any direct decrease in social distancing among the subset of the population participating in the protests is more than offset by increasing social distancing behavior among others

3. #3 peak

   take 0 steps.⏎If we consider all permutations of the first 4 natural numbers as input the expectation value of the number of shuffles, averaged over all 4! input sequences is 27.5.⏎Consider as input sequences the

4. #4 peak

   76. They would find, he suggested, that utter disaster for the American cause was averted because of a valiant rear-guard action fought by a regiment of the Maryland line. Let them, therefore, go to the sources and find out

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.