"as" after "known"
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 "as" when it comes straight after "known", as in "known as quercetin" or "known as Streptococcus". It is a clear and reliable pattern. Turning it up did not bring out this theme in the model's replies, so that test tells us nothing either way.
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.
-
rich in vitamins, minerals, carbohydrates and amino acids etc. Apples are also very good source of vitamin B-Complex.⏎Apple contains a kind of antioxidant known as ‘quercetin’ which is very essential for neurological health. It reduces
-
in an enzyme known as argininosuccinate synthase which is needed for this vital process, leading
-
. The trash accumulates from all over the world and as it moves into the ocean, it becomes trapped at the center of a system of rotating ocean currents known as the
-
) is a proto-oncogene that encodes a protein known as hepatocyte growth factor receptor (HGFR). The hepatocyte growth factor receptor protein
-
If strep has been diagnosed, a lack of improvement after taking antibiotics for 48 hours⏎Strep throat is caused by infection with a bacterium known as Streptococcus pyogenes, also called group A streptococcus
-
in the slave quarters. Sometimes the white masters would come to the slave quarters to watch the celebration. Parents would give children small, homemade tokens.⏎Another celebration known as Jonkonnu, or a Christmas masquerade, took place on the plant
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: "I know, I know, I know, I know, I know, I know, I know, I know, I know, I know, I know, I know, I know, I know, I know, I know
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
're gone.⏎The title of this well-known story comes from a poem by Sara Teasdale in which nature is depicted as indifferent
-
both sides. Part A: Short Answer (1 problems, 12 points).
-
and Livia Visser-Fuchs:⏎“A ‘Most Benevolent Queen’: Elizabeth Woodville’s Reputation, her Piety and her Books”, includes a convincing refutation of common allegations
-
slave economy.⏎The name of a former kingdom at the confluence of the Niger and Benue Rivers in West Africa
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
- L0:2554 weakly excites
- L0:1089 weakly excites
- L0:1224 weakly suppresses
- L0:178 weakly suppresses
- L0:1152 weakly suppresses
- L0:218 weakly excites
- L0:1960 weakly suppresses unclear: mixed medical and harsh words
- L0:1133 weakly suppresses
- L0:1301 weakly suppresses
- L0:932 weakly excites
- L0:2853 weakly excites
- L0:2257 weakly suppresses
- L0:862 weakly suppresses
- L0:12 weakly suppresses
- L0:1001 weakly suppresses
- L0:392 weakly suppresses
- L0:183 weakly suppresses
- L0:2986 weakly suppresses
- L0:976 weakly suppresses
- L0:711 weakly excites
- L0:2684 weakly excites
- L0:2529 weakly excites
- L0:1066 weakly excites
- L0:71 weakly excites
Feeds later layers
- L7:79 excites
- L19:1934 excites
- L16:1599 excites
- L8:2006 excites
- L13:838 weakly excites
- L20:954 weakly excites
- L5:680 weakly excites
- L9:2895 weakly excites
- L16:1734 weakly excites
- L6:1639 weakly excites
- L11:1051 weakly excites
- L18:469 weakly suppresses
- L18:1678 weakly excites
- L19:1897 weakly excites
- L17:1322 weakly excites
- L16:2697 weakly excites
- L17:1841 weakly excites
- L3:402 weakly excites
- L19:2068 weakly suppresses
- L12:636 weakly excites
- L10:2627 weakly excites
- L11:1306 weakly excites
- L4:1004 weakly excites
- L21:362 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
- umu
- onder
- unteer
- bler
- ousel
- auer
- дум
- ues
- igmat
- uet
Suppresses
- prere
- void
- .fd
- iect
- Dll
- rtl
- ящих
- retr
- 瘪
- 还原