unclear: 'ardu' and 'axi' word fragments
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. 2 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 is unclear. It reacts to pieces of longer words, especially 'ARDUINO', 'arduous' and 'axial' or 'axiomatic'. Turning it up or down did not give a clear effect.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
are checked to verify they meet the required specifications.⏎<|endoftext|>RASPBERRY Pi - LEARNING HOW TO MAKE YOUR ROBOT WORK⏎ANTICS - ARDUINO - ARMOUR - ARTWORK - BIOLOGY - BLACK BOX - COMPUTERS
-
slightly more sustainable in the reduction of energy and time used by humans to create robots for a better world. KISS⏎ABOUT RASPBERRY Pi and ARDUINO COMPUTERS⏎The Arduino is a wonderful platform for handling the lower level work
-
particular value in ornamental planting. By placing three or five plants in clump-fashion, the beauty of this Broom is greatly enhanced.⏎C. ALDUS INCARNATUS (syn C. incarnatus) resembles C. purpureus
-
σχῆμα τῷ λόφῳ in Einklang zu bringen, ist es unumgänglich nothwendig dass der Ausdruck Roma quadrata technisch in doppeltem Sinne gebraucht wurde
-
of those very mountains. The tortuous, switchback road with steep climbs, hairpin bends and often cut by bad weather conditions made travel so long and arduous to Palma, that Sóller looked more to Barcelona – a sea
-
length and can vary in color. Mating can take from a few hours to a few days with the male sitting atop the female.<|endoftext|>Even when ‘Abdu’l-Bahá strongly disagreed with the religious beliefs from Roman Catholics. Examples
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 need to be able to do this. I need to be able to do this. I need to be able to do this. I need to be able to do this. I need to be able to do this. I need to
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
- The foundations of analysis, which begins with the study of functions and their derivatives and integrals⏎- The fundamentals of advanced algebra, which is based on axiomatic systems involving operations of addition and multiplication in general settings⏎- Additional subjects such
-
dorsum)⏎- Lateral digital vein (near dewclaws)⏎How does infection spread from the coffin joint, and what does it cause?⏎- axially: footrot⏎- abaxially: vertical fissure aka sand crack⏎
-
- More weight on the front half⏎What is the blood supply to the claw?⏎- dorsal: dorsal metatarsal a. -> axial proper digital a. -> terminal arch⏎- plantar: terminal arch & marginal
-
fasts⏎small, inconspicuous, or lacking. Commonly yellowish, but deep red⏎Medulla appear parenchymatous around central axial cell;⏎cortical filaments with few divisions on radii, outer layer
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
- L0:1224 suppresses
- L0:178 suppresses
- L0:1301 suppresses
- L0:1 weakly suppresses
- L1:1464 weakly excites
- L1:2799 weakly excites
- L1:1717 weakly excites
- L1:146 weakly excites
- L1:6 weakly excites
- L0:397 weakly suppresses
- L1:262 weakly suppresses
- L0:1813 weakly excites
- L1:1960 weakly excites
- L1:56 weakly excites
- L0:495 weakly suppresses
- L1:30 weakly excites
- L1:1692 weakly excites
- L0:644 weakly excites
- L1:1583 weakly excites
- L1:1757 weakly suppresses
- L1:2318 weakly suppresses
- L1:3038 weakly excites
- L1:2444 weakly suppresses
- L1:491 weakly excites
Feeds later layers
- L4:2785 suppresses
- L4:1843 weakly suppresses
- L5:1392 weakly excites
- L4:2774 weakly excites
- L5:1699 weakly excites
- L9:2261 weakly excites
- L16:596 weakly excites
- L7:738 weakly suppresses
- L4:666 weakly suppresses the digit 0 in dates
- L12:2637 weakly excites
- L7:1035 weakly excites
- L23:181 weakly suppresses
- L13:449 weakly suppresses
- L21:244 weakly suppresses
- L10:785 weakly excites
- L22:1874 weakly suppresses
- L7:927 weakly suppresses
- L3:1748 weakly suppresses "one" as a pronoun after a verb
- L10:818 weakly excites
- L18:1596 weakly suppresses
- L8:1096 weakly suppresses
- L19:751 weakly suppresses the article 'a' before a noun
- L11:2529 weakly excites 'here' in 'read more here' links
- L16:1049 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
- utta
- utive
- structors
- utations
- @Before
- utable
- ☰
- 腾
- paused
- pa
Suppresses
- isel
- _planes
- jer
- zie
- ActionCreators
- laştır
- sdale
- axes
- estring
- sess