anatomy terms, especially salivary glands
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. 9 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. 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 responds to body and biology words, most clearly the pieces of 'salivary' and related gland and tissue terms. Turning it up did not produce a clear change in what the model wrote.
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.
-
/etc as a search path too.<|endoftext|>Salivary gland (Endocrine system)⏎: #1 is , #2 is , #3 is⏎The salivary glands in are that produce . They also secrete , an that breaks down
-
.⏎During pretreatment, biomass feedstock undergoes a process to mechanically or chemically fractionate the lignocellulosic complex into soluble and insoluble components. Soluble components include mixtures of
-
? Why was Jesus worthy of worship? (God, Jesus is God the Son in human flesh–fully God and fully human)⏎Take time to talk to your kids about
-
saliva proceeds to aid in digestion. The saliva that salivary glands release is quickly inactivated in the stomach by the acid that is present there.⏎The salivary
-
itudinal, circular, oblique), myenteric (Auerbach’s) plexus|⏎|Serosa||Connective tissue, mesoderm|⏎Histology may not be the easiest to digest, but we will help you
-
. Approximately 5% of saliva entering the oral cavity come from these glands.⏎Minor Salivary Glands⏎There are over 600 minor saliv
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: can we use a higher budget for the 2010s? And I don't want to see the same amount of change in the 2010s compared to 2015, like the one before?
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Digital Audio, that’s technological illustrations. You are having indications about weather from a weather sensor, that’s technological illustrations. You are taking to Driver Instructor while driving the car, that’s technological illustrations
-
) is a proto-oncogene that encodes a protein known as hepatocyte growth factor receptor (HGFR). The hepatocyte growth factor receptor protein possesses tyrosine-kinase activity. The primary single chain precursor protein is post-translation
-
common in children. A type of cancer derived from immature precursor cells or embryonic tissue.|⏎Cell Signaling in Tumorigenesis⏎There are multiple signaling pathways that contribute to cell proliferation, motility and survival. In cancer, components
-
lived in isolation from human contact from a young age and has no experience of human care, loving or social behaviour and even human language. Feral children often become mentally and physically impaired.⏎There are a few examples
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
- L6:1201 suppresses
- L6:1518 weakly suppresses
- L3:1259 weakly suppresses
- L0:384 weakly excites
- L8:2188 weakly excites
- L0:260 weakly suppresses
- L1:2280 weakly excites
- L7:1325 weakly suppresses
- L7:1095 weakly suppresses
- L3:862 weakly excites
- L7:204 weakly suppresses
- L1:2456 weakly excites
- L1:1587 weakly excites
- L5:2004 weakly suppresses
- L3:1866 weakly excites
- L6:1207 weakly excites
- L6:778 weakly suppresses
- L7:1071 weakly excites
- L6:3046 weakly suppresses
- L8:743 weakly suppresses
- L3:540 weakly excites
- L3:2556 weakly suppresses
- L0:940 weakly suppresses
- L6:2161 weakly suppresses
Feeds later layers
- L11:1555 excites
- L10:540 suppresses
- L12:662 weakly suppresses
- L16:1604 weakly suppresses
- L11:2557 weakly excites
- L12:2150 weakly excites
- L10:1864 weakly excites
- L12:1454 weakly excites
- L15:372 weakly suppresses
- L22:3016 weakly excites
- L11:172 weakly suppresses
- L10:2260 weakly excites
- L16:2587 weakly excites
- L12:2059 weakly excites
- L20:2468 weakly excites
- L14:1658 weakly excites
- L14:2329 weakly suppresses
- L12:1007 weakly excites
- L20:2226 weakly suppresses
- L18:2152 weakly excites
- L21:811 weakly excites
- L12:714 weakly excites
- L10:2903 weakly excites
- L10:345 weakly suppresses
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
- 性质
- 就需要
- 借助
- 我们就
- -app
- 中介机构
- 能力和
- [Y
- TextInputType
- 提出了
Suppresses
- dbuf
- ่าง
- mlin
- phan
- _ping
- uations
- neath
- pac
- odes
- _cover