'that' after a finding or surprise verb

What this neuron does

This neuron fires on the word 'that' when it introduces a clause after a verb of finding, learning or evidence, as in 'you may be shocked to learn that' or 'it's not surprising that'. Its opposite side is on words of belief, such as 'believed', 'believe' and 'sure of'. Turning it up did not give a clear theme in the sample text.

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

   ZZs like Aspilche. We asked her about how she’s managing after the first year of knowing she has Alpha 1.⏎How did you discover that you had it?⏎For about two years prior to this confirmation, I was diagnosed

2. #2 peak

   build-up. The third and most important strategy is HEPA air filtration.⏎HEPA Air Purification⏎A HEPA air<|endoftext|>You may be shocked to learn that in England alone, people take 3 billion car trips for less

3. #3 peak

   Ste. Genevieve Plant has joined it. It’s not surprising that the list hasn’t changed much – major industrial plants like these aren’t built every

4. #4 peak

   with a broader body of work indicating a profound impact of climate change on precisely these sorts of events.”⏎The new normal⏎“This analysis confirms what we already know – that we are suffering an extreme weather event caused by climate change,” says Gareth Redmond

5. #5 peak

   much and in what ways we trust AI in the workplace, how we view the risks and benefits, and what is expected for AI to be trusted. We find that only one in two employees are willing to trust AI at work. Their attitude depends

6. #6 peak

   their fat cells detected more insulin resistance in the women who ate the high-GI diet. It’s hardly surprising, then, that several large population studies have shown that people who eat diets filled with low-fibre

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

   It is generally believed that profit earning is the primary objective of business. However, as per modern thinking

2. #2 peak

   anyone how long the interruption would last. But with her intimate knowledge of volcanic ash, and a good idea of what it does to jet engines, she was sure of one thing: “I wouldn’t want to get on a plane right now.”⏎

3. #3 peak

   indicating that the head bearing Pharaoh Khafre's likeness could have been carved out of a larger lion's head. These points led some researchers to believe that the Sphinx could have been built much earlier by a more ancient civilization, thus

4. #4 peak

   intuitive. I certainly can’t claim to be a perfect listener. All I can say is that I’m learning and getting better.⏎We all know, of course, that listening and hearing are required in order to understand the heart of what 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.