'the' right before a year or number

What this neuron does

This neuron fires on the word 'the' when a year or a number like '1950s' or '20th century' comes next. It seems to expect a date. Turning it up did not produce date-related text in the samples, so the link is mostly from the examples.

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

   The 1992 fieldwork consisted of⏎surface collection and the excavation of five

2. #2 peak

   The 2025 targets are unlikely to be met, and health service providers will

3. #3 peak

   the 20th century there was a wave that brought in immigrants from Asia and Africa

4. #4 peak

   the 1950s and 1960s, and the Free

5. #5 peak

   the 16-county District averaged 3.89 inches – the second-largest

6. #6 peak

   the 1890s the Night Riders had done their part, Jim Crow laws

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

   , or qubits, can encode both at the same time. This essentially enables quantum computers to work through multiple scenarios simultaneously, rather than exploring them one after the other. However, these mixed states don’t last long, so the information processing must

2. #2 peak

   the partial results one below the other, starting with the product of the last digit of the second number. Each partial

3. #3 peak

   directions; the cold will be severe and the winds will be fierce. The sun will be of no use. Three of these winters will come, one after the other, with no summer in between. But before that, there will have

4. #4 peak

   from one to the other. One qualification: OED refers only to a Roman vaulter but the

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.