"to" after 'for X' infinitives

What this neuron does

Fires on the word "to" in phrases like "For it to work" or "for evolution to be true," where a "for someone or something" phrase introduces an action. Its strongest negative cases are the word fragment "bs" and "owns" in place names like suburbs. Turning it up gave text with no clear shared theme.

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

   another device.⏎For it to work, at least two devices need to be Sparsh-enabled. A user wanting

2. #2 peak

   ippedTextTolerance property allows you to⏎define a rectangular area in the center of the character, the⏎entire of which must be visible for the character to be visible.⏎The width and height of the rectangular area are determined by the⏎

3. #3 peak

   new antibiotic.⏎For treatment to be effective and practical, it must be simple to administer by mouth as a single

4. #4 peak

   and photographic lineups. Lineups involve showing the witness a group of possible suspects. Photographic lineups provide a selection of pictures of possible suspects for the witness to choose from. Showups are done by showing one suspect to the witness.⏎The

5. #5 peak

   to carbon dating, kids of faith have questions these days. They're struggling to put together what their schools teach about science and what their beliefs are.⏎For evolution to be true, the earth must be billions of years old, with life-forms

6. #6 peak

   being punishable by the assignment of monetary damages to compensate victims for their losses, and criminal fraud being punishable by jail time as well as by fines. For a fraud to be a criminal offense, it needs to have been in violation of either a federal

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

   Central Business District and surrounding areas such as Braamfontein, Hillbrow, and Yeoville for more secured houses or offices in the Northern Suburbs. By the late 1990s, the Central<|endoftext|>A shoulder

2. #2 peak

   and urban blight, many of the buildings are unoccupied as tenants have left for more secure locations in the Northern Suburbs, in particular Sandton and Rosebank. There are significant movements to revive the

3. #3 peak

   of extraordinary contrasts, of glass and steel skyscrapers and fetid shantytowns, of internationally recognized universities and widespread illiteracy, of glittering abundance and

4. #4 peak

   and⏎• increase learning opportunities outdoors.⏎Placing children’s well-being at the center of planning of the urban environment brings benefits – not only for children but for city dwellers of all ages. Child-friendly environments support inclusion, well-being, sustainability

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.