'like' after 'feel'
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. 1 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 'like' when it comes right after 'feel' or 'feeling', as in 'makes you feel like'. When it was turned up, the model's writing did not visibly change in a way that matches this pattern.
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 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.
-
likely to buy the outside brands. If people feel like they have no other choice but to buy the foreign good since it is not available
-
your child’s) symptoms may be related to a food allergy or not and can help identify which foods should be avoided to prevent possible allergic reaction⏎If you feel like you or your child has experienced an allergic reaction to foods, call our office to
-
., is an ideal stop. Situated on the banks of the Mississippi River at its most serene and picturesque, this is the kind of town that makes you feel like a<|endoftext|>History of crystallography⏎From Lucretius to Kita
-
, and makes you feel like you aren’t in control.⏎Ketamine is FDA-approved for use as a
-
mainstream books, not isolated as a separate genre or segregated strand of literature.⏎- aim to keep up to date with matters surrounding diversity and inclusion.⏎If you feel like we could be doing this better than we already are then please don't hesitate to
-
and everyone else graduates in skill. "This is really painful and comes up often when I talk to young athletes," says Andersonn. "It makes them feel like failures; they feel deeply humiliated."⏎Parents can be more sensitive to this issue
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 current. I think that the current. I have been thinking about lately. So the current is, I have been thinking about lately. The current is, I have been thinking about lately. So the current, I have been thinking about
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
. But there’s still a danger in the mountains in the summer.⏎“Initially, dry lightning could be a concern, so obviously, the fire season is looking not so great right now,” he said.⏎Weather data showed the last year’s conditions
-
1830. (See etymonline.)⏎Not only did Shakespeare rhyme hour with flower, but he also sometimes spelled them the same.⏎It fears not policy that Hereticke,⏎Which works on leases of short numbred howers
-
thus starting a whole host of conspiracy theories.⏎Geologist Dr. Robert Schoch and author John Anthony West made a case that the erosion marks on the Sphinx appears not to be made by wind or sand, but by water, such as falling rain
-
hear the daily stats and buzz, but presidential elections are about the big picture — where we want to go and the best way to get there. This means looking not only at political options but also at the way we humans are set up — how
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
- L0:148 weakly excites
- L0:2643 weakly excites
- L0:1730 weakly suppresses
- L0:1279 weakly excites
- L0:2828 weakly suppresses
- L0:1574 weakly excites unclear: mostly line breaks, mixed text
- L0:157 weakly suppresses
- L0:2966 weakly excites
- L0:1301 weakly suppresses
- L0:2103 weakly excites
- L0:1336 weakly suppresses
- L0:2019 weakly suppresses
- L0:1224 weakly suppresses
- L0:762 weakly excites
- L0:2456 weakly suppresses
- L0:283 weakly excites
- L0:3042 weakly excites
- L0:1651 weakly suppresses
- L0:1948 weakly suppresses
- L0:1588 weakly excites
- L0:1130 weakly excites
- L0:1110 weakly excites
- L0:2073 weakly suppresses
- L0:69 weakly suppresses
Feeds later layers
- L2:1949 suppresses
- L11:2740 weakly suppresses
- L6:311 weakly suppresses
- L13:2608 weakly excites
- L15:1916 weakly excites
- L2:1980 weakly excites
- L13:1863 weakly excites
- L24:1459 weakly excites
- L2:2850 weakly suppresses
- L15:2289 weakly excites
- L19:1726 weakly excites
- L2:893 weakly suppresses
- L17:1501 weakly excites
- L6:2453 weakly excites
- L27:1564 weakly suppresses
- L8:1214 weakly excites
- L15:2961 weakly excites
- L19:1879 weakly suppresses
- L22:427 weakly suppresses
- L3:2877 weakly excites
- L16:1644 weakly suppresses
- L2:1197 weakly suppresses
- L6:1628 weakly suppresses
- L18:2284 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
- hek
- Iteration
- ��
- 欠
- icolor
- idency
- BC
- '
- Beg
- 弁
Suppresses
- presso
- атор
- ushima
- 溱
- Filed
- NOTICE
- تعالى
- Tamb
- مية
- lửa