'for' after check or look
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. 6 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 fires on the word 'for' right after verbs like 'check', 'checked' or 'look', as in 'check for infection'. When it was turned up, the text drifted toward the idea of checking things, though only loosely.
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.
-
is torn.⏎If the results of your exam suggest you have a torn meniscus, you may need more tests, including:⏎- Knee X-rays to check for bone injuries, including fracture, that can cause symptoms similar to a torn menisc
-
NF), a protein that protects neurons and is important for learning and memory. In a study<|endoftext|>During a baby’s prenatal care, the heart is monitored closely for any problems or irregularities. Pediatric Cardiologist John E. Pliska, MD
-
in production shops where the “screws” (externally threaded parts) or the “nuts” (internally threaded parts) are to be quickly checked for acceptance. But when you consider the metrology (the science of measurement) of
-
to see his or her healthcare provider. The provider will do a throat swab to check for infection. The provider will also ask about the person's vaccine history. A booster
-
(Engine Off)⏎- Look for puddles on the ground.⏎- Look for dripping fluids on underside of engine and transmission.⏎- Inspect hoses for condition and leaks.⏎- Indicate where dipstick is located.⏎- See that
-
child's symptoms and not simply to detect the presence of an H. pylori infection. An endoscopy can help look for other problems in addition to evaluating for H. pylori.⏎The best way to get rid of H. pylori
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 time, and the second time, and the third time. So, I will put them in order. Let me think. - I have to use the first to solve the problem. So, they are ... (answer) This
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
, and compensate for fluctuations in blood pressure?⏎A complex of structures located NEXT TO THE GLOMERULUS in the kidney that regulates the function of each nephron:⏎Macula densa, Juxtaglomerular cells, Mesang
-
lectronics has not been given much consideration previously, though much effort has been devoted to the development of general-purpose nanoelectronic systems, i.e., nanocomputers. This paper demonstrates via
-
radii, outer layer pigmented.⏎Tetrasporangia zonately divided, in raised nemathecia, usually on ultimate⏎branches; spermatangia borne in chains in slightly swollen nemathecia at⏎base of branch
-
or cup-shaped; perianth segments 6, of 2 similar whorls; nectaries 6, on perianth parts; stamens 6, included, inserted at perianth base, anthers attached
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
- L4:1382 excites
- L4:1655 excites
- L5:86 weakly suppresses
- L5:1203 weakly suppresses
- L4:265 weakly suppresses
- L0:1470 weakly excites
- L5:1090 weakly excites
- L1:1281 weakly excites
- L5:3034 weakly suppresses
- L3:2799 weakly suppresses
- L5:1147 weakly suppresses
- L5:2674 weakly excites
- L3:2116 weakly excites
- L4:1436 weakly suppresses
- L3:2876 weakly suppresses
- L2:1883 weakly suppresses
- L2:2742 weakly suppresses
- L3:2358 weakly suppresses
- L0:1859 weakly excites
- L2:2092 weakly suppresses
- L1:764 weakly suppresses
- L3:1906 weakly suppresses
- L5:1650 weakly suppresses
- L3:2183 weakly suppresses
Feeds later layers
- L22:899 excites
- L9:406 excites
- L8:2217 excites
- L11:2716 excites
- L18:573 weakly excites
- L10:1451 weakly excites
- L22:1933 weakly excites
- L8:1492 weakly excites
- L7:945 weakly excites
- L10:628 weakly excites
- L12:1607 weakly excites
- L11:1678 weakly excites
- L22:125 weakly suppresses
- L15:1273 weakly excites
- L17:2561 weakly suppresses
- L16:2253 weakly suppresses
- L9:933 weakly suppresses
- L10:263 weakly excites
- L7:394 weakly suppresses
- L12:1688 weakly excites
- L10:2812 weakly excites
- L9:2642 weakly excites
- L21:1592 weakly excites
- L9:2365 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
- 检查
- Checked
- .detect
- .check
- prepared
- Kling
- Checking
- ByPrimaryKey
- 是否存在
- -check
Suppresses
- ymes
- antas
- yme
- 彘
- iser
- lined
- hear
- liest
- 酞
- ▾