'other' in 'in other words'
A neuronOne of thousands of tiny dials inside each layer. It reads the text so far and produces a number, its activation, that nudges what the model writes next. in layerOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. 22 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featuredA neuron picked for the full treatment: steering outputs and a written explanation. About 3,000 of the 86,000 neurons are featured., so it has steering outputs.
What this neuron does
This neuron lights up on the word 'other' when it comes right after 'in' and leads into a restatement, as in 'in other words'. Turning it up made the model repeat the word 'words' over and over.
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 activationThe number a neuron produces for a token. Large positive means it fired hard; negative means it pushed the other way. was highest. Amber marks positive activation; the outlined tokenA piece of text the model reads at once: often a word, sometimes part of a word or a punctuation mark. is the peak. Hover a token for its strength.
-
carriers showed significantly worse prospective memories. Far more often than non-carriers, they failed to remember to write down the desired word when they were supposed to – in other words, they forgot to do what they meant to do, when they meant to
-
code that can be hosted by the CLR, while at the same time can be accessed in a uniform manner by all languages that the .NET platform support. In other words, CLS is a subset of the full functionality of CTS.⏎-
-
- Compassion vs Empathy: What Is the Difference?⏎What Is Empathy?⏎Empathy is the ability to put yourself in someone else’s shoes. In other words, it’s the ability to feel or understand another person. Experts suggest it
-
a number of signals with rather close frequencies. As soon as one of these signals "exceeds" the others, it masks them more or less: in other words, the other sound are perceived as much weaker than they actually are, or
-
“spiritual aristocrats” Tetraphilosophy appear simple⏎with capacious dialectic models and abbreviations that are clear to any young person; in other words, we have democracy instead of aristocr
-
have various causes, ranging in severity from 'worrying' to 'generally fatal'. Finding the true cause means ruling out or confirming each possibility – in other words, diagnosis.⏎Diagnosis is usually a complex process due to the sheer number
SteeringAdding a push along a neuron while the model writes, to see how its output changes. Turning it up or down.
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.
Here is something I have been thinking about lately: the first and the second antumnumus. I mean, the first antumnumus is the antumnus antumnumus antumnumus... that antumnumum antumnum
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
submit original poems. The only constraint is that the poems must follow a format suitable for Twitter–280 characters or fewer. These bite-sized verses often prove interesting, complex,
-
online schedule and for those that don’t have the ability to move to another state, the college they wish to attend is just a few mouse clicks away.⏎Is this where grade-school level education is headed? I can’t say for certain that is
-
advance through your welding career, you will need to do various complicated welding jobs. So, learning how to weld pipes is necessary to avoid any mistakes or accidents.⏎This job can become even more challenging when it is welding pipes instead of flat surfaces.
-
.⏎This decision came down from the top brass. Don't bother questioning it.⏎I would not use it to talk about the people who administer a program.⏎Is this use of "brass" common in the UK?⏎I'm not a teacher
Connections
Show on mapWhich neuronsOne of thousands of tiny dials inside each layer. It reads the text so far and produces a number, its activation, that nudges what the model writes next. in earlier layersOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.
Fed by earlier layers
- L20:1065 excites
- L19:1587 suppresses
- L19:2150 excites
- L11:1687 suppresses
- L21:1463 suppresses
- L6:2787 weakly suppresses
- L21:2820 weakly excites
- L21:708 weakly suppresses
- L19:1926 weakly suppresses
- L18:1537 weakly suppresses
- L9:2958 weakly excites
- L17:576 weakly excites
- L13:855 weakly excites
- L3:2393 weakly excites
- L3:993 weakly suppresses
- L7:2134 weakly suppresses
- L12:148 weakly excites
- L16:973 weakly suppresses
- L16:3002 weakly excites
- L5:86 weakly suppresses
- L1:2581 weakly excites
- L14:1692 weakly suppresses
- L20:1577 weakly suppresses
- L21:998 weakly suppresses
Feeds later layers
- L27:1957 strongly excites
- L23:2317 excites
- L26:523 excites
- L23:2612 excites
- L26:556 excites
- L25:2535 excites
- L25:1331 weakly excites
- L24:268 weakly excites
- L23:1471 weakly excites
- L26:2130 weakly excites
- L25:991 weakly excites
- L23:2546 weakly excites
- L25:1822 weakly excites
- L25:2866 weakly excites
- L24:1841 weakly excites
- L24:2465 weakly excites
- L24:703 weakly excites
- L23:2872 weakly excites
- L24:1594 weakly excites
- L23:2588 weakly excites
- L27:1667 weakly excites
- L25:1571 weakly excites
- L23:409 weakly excites article 'a' after a role or title
- L24:1500 weakly excites
Activation histogramA chart of how often a neuron takes each activation value.
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 lensReading which output tokens a neuron pushes toward or away from, straight off its output weights, before the later layers have their say.
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 layersOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. instead.
Promotes
- words
- words
- Words
- Words
- word
- _words
- Word
- word
- .words
- Word
Suppresses
- 硌
- MaxY
- INESS
- asant
- directory
- ragment
- major
- ıs
- glGet
- ighth