'of' after a noun phrase

What this neuron does

This neuron fires on the word 'of' when it follows a noun and introduces what that noun belongs to or is about, as in 'sequences of' or 'condition of'. When turned up, the model got stuck repeating 'a certain type of' phrases, which fits the same 'of' habit.

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

   Res. (2006) 16: gr.5565706)⏎The human genome contains about 8% of sequences of retroviral origin, remnants of different exogenous retrovirus infections of the

2. #2 peak

   physically different green boy, instead of a black or a yellow one, was a way the Authors found to give a better approach to the moral feeling regarding having friends of different out look. Finally, in Story Three, the description of a poorly dressed

3. #3 peak

   , not because the problem is solved⏎- There is no immediate and no ultimate test of a solution to a wicked problem – any solution will generate waves of consequences of an extended period of time

4. #4 peak

   those dealing in, acquiring, buying or borrowing items of cultural property have a duty to:⏎- keep abreast of legal issues;⏎-

5. #5 peak

   …⏎(8 days ago) Primary peritoneal carcinoma has a prognosis that overlaps with, or is the same as, tubal/ovarian carcinomas of similar histologic grade. This is reflected in the latest American Joint Commission on Cancer

6. #6 peak

   x; END;⏎Thus, lazy evaluation allows us to define cyclic graphs of pointers with warrantedly valid pointers. In contrast, C allows cyclic graphs of pointers

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

   Fall 2019⏎Instructor: Norman Smith⏎This course examines a selection of topics in “modern” Chinese history. We will explore, in a roughly chronological fashion, how Chinese lives intersected, and were intersected by,

2. #2 peak

   Paul, maintaining that he was an apostate from the law. As to the prophetical writings, they endeavor to expound them in a somewhat singular manner: they practice circumcision, persevere in the observance of those

3. #3 peak

   augment classroom instruction. While it has forums and⏎wikis to create communities around courses, it can also be used with⏎assignments and quizzes to teach in a more traditional format.⏎How widespread is Moodle used in education? Currently,⏎there are

4. #4 peak

   7 of the two Chinese characters forming it, has been the object of various interpretations, mainly by Oriental scholars.⏎This script does not aim to solve the problem in an irrefutable way, but simply tries to clear up some historical-lingu

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.