blank before a size number
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. 21 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 empty spot right before a measurement, such as the gap in "grows to [blank] feet tall" or "about [blank] inches long". Turning it up gave text about sizes and heights.
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.
-
Then you would notice that he was slightly different, based on the brow ridge and facial features.⏎It is worth remembering that the average height of Homo erectus was 6 ft., they
-
inea pigs are considered large weighing approximately 700 and 1200 grams or 1.5 to 2.5 pounds and measuring around 20 to
-
Maturity: 45–60 days⏎Native: Widespread, probably Asia⏎Hardiness: Frost-tolerant annual⏎Plant Dimensions: 16"–22" tall⏎Variety Information: Large, dark
-
Glow," a carnelian cherry dogwood (Cornus mas) cultivar, tolerates a range of moist soils, including well-drained clay. Standing 15 to 25 feet high, with a 12- to
-
), formerly classed with lemurs on account of its woolly aspect, monkey-like dexterity, and strong prehensile tail. Its body is about 18 inches long, and its tail 12 inches; and it is
-
, the ears are rounded and small, not reaching the muzzle, and it has small feet with a keeled calcar. Body measurements are: total length, 87 to 103 mm; tail, 37 to
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 have a 48000$ on the first year of my $48000$ $38$ $200$ $200$ $200$ $38$ $2
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
everything they<|endoftext|>Variety of electromagnetic radiation with frequencies between a few and a few hundred quadrillion oscillations per second, corresponding to wave-lengths between a few hundred and a few billionths of metres. Known in everyday life as that part of
-
regular microLEDs.⏎In addition to the advantages at the chip (LED) level, the nano-membranes (which are thin, at 100nm) can be broken using mechanical force and so the microLEDs can be
-
the HiRISE camera onboard Mars Reconnaissance Orbiter -- the first image ever of a spacecraft descending to the surface of another planet. Taken from 750
-
EM waves with frequencies below and above that of humans a better definition must be made. We therefore define the visual wavelength range as wavelengths of EM between 300nm and 800nm.
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
- L10:2845 suppresses
- L18:1546 suppresses
- L19:2129 excites
- L19:2176 suppresses
- L6:1030 suppresses
- L16:1154 suppresses
- L10:1911 excites
- L10:294 suppresses
- L20:1411 excites
- L15:861 excites
- L6:3064 suppresses
- L3:1251 weakly excites
- L20:876 weakly suppresses
- L18:336 weakly excites
- L10:1496 weakly excites
- L19:1158 weakly suppresses
- L20:1510 weakly excites
- L20:2838 weakly suppresses
- L5:1247 weakly suppresses
- L16:1882 weakly suppresses
- L20:1128 weakly excites
- L9:2048 weakly suppresses
- L12:2002 weakly excites
- L20:640 weakly excites
Feeds later layers
- L24:601 excites
- L26:1508 excites
- L25:918 excites
- L23:1742 excites
- L22:557 excites
- L27:1404 excites
- L24:2403 excites
- L27:2106 excites
- L23:2862 excites
- L23:2296 excites
- L24:2592 excites
- L22:608 excites
- L23:2206 excites
- L26:1469 weakly excites
- L22:2783 weakly excites
- L22:2854 weakly excites
- L22:2474 weakly excites
- L24:1809 weakly excites
- L24:3016 weakly suppresses
- L24:1238 weakly excites blank space before a number
- L22:1020 weakly excites
- L27:1830 weakly excites
- L22:1027 weakly excites
- L24:1097 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
- size
- larger
- smaller
- -size
- taller
- サイズ
- size
- sizes
- 大小
- -sized
Suppresses
- eways
- acebook
- hesion
- ę
- pery
- adow
- redi
- 去了
- xDE
- adox