birth dates, the day before the year
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. 10 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 spot right before a birth year, just after a day and month like 'born December 7,'. It is one of the clearest patterns here. Turning it up made the writing drift into unrelated topics.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. Think it is wrong? Flag it (opens an email to the maker with the neuron already filled in).
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.
-
full Titus Vespasianus Augustus, original name Titus Flavius Vespasianus, (born Dec. 30, 39 ce—died Sept.
-
, who helped to promote and publish his work.⏎What makes Robert Frost unique?⏎Robert Frost, in full Robert Lee Frost, (born March 26, 1874, San Francisco, California, U.S.—died January
-
during the final manned flight of the Skylab space station in 1973 and 1974.⏎He was born January 23, 1930, in Okemah, Oklahoma. He received a Bachelor
-
Ticknor⏎Curtis, George Ticknor, American writer on legal topics, was born at Watertown, Mass., Nov. 28, 1812; and died at New York,
-
|⏎|Born||December 7, 1928⏎|Era||20th / 21
-
was born in Montpelier, Vermont, on December 26, 1837. A hero of the Civil War and Spanish-American War,
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: 1. A company has a report that is not in the system. 2. A company has a report that is not in the system. 2. A company has a report that is not in the system. 3.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
. AROMATIC COMPOUNDS ARE THOSE ORGANIC COMPOUNDS WHICH HAVE BENZENE RING (CYCLOHEXA-1,3,5-TRIENE). AROMATIC COMPOUNDS
-
and one family name, but can have as many additional given names and middle names as they like. A middle name is usually a recognised surname and not a patronymic.⏎What are the () called?⏎() – This are simply termed as brackets
-
: C F O S U.⏎No direct anagrams for focus found in this word list.⏎List shorter words within focus, sorted by length⏎All words formed from focus by changing one letter⏎Browse words starting with focus by next letter<|endoftext|>I
-
led the way again in Swiss Ball manufacture by eliminating all phthalates from its products.” says Bradley Wilson, Managing Director of AOK Health Pty Ltd.⏎Phthalates are a group of chemicals used to make plastics more flexible and harder to break
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
- L6:1894 excites
- L9:3038 excites
- L6:214 suppresses
- L8:590 excites
- L8:1945 suppresses
- L9:345 excites
- L9:2891 suppresses
- L4:2290 suppresses
- L8:523 suppresses
- L6:2498 suppresses
- L6:2198 excites
- L9:652 suppresses
- L8:2252 excites
- L6:1047 excites
- L6:1783 excites
- L7:577 suppresses
- L9:2498 suppresses
- L8:2919 excites
- L7:1121 excites
- L9:271 excites
- L9:2676 suppresses
- L8:2121 excites
- L9:1725 suppresses
- L7:2340 weakly suppresses
Feeds later layers
- L16:2377 excites glosses, translations and unit conversions
- L17:958 excites
- L12:597 excites
- L20:2867 excites
- L11:980 weakly excites
- L21:2797 weakly excites
- L14:2686 weakly excites
- L20:2591 weakly excites
- L11:1741 weakly excites
- L11:1703 weakly excites
- L12:1279 weakly excites
- L21:2125 weakly suppresses
- L11:847 weakly excites
- L19:2661 weakly excites
- L20:1256 weakly excites
- L14:2058 weakly suppresses
- L11:2670 weakly excites
- L11:2754 weakly excites
- L19:1876 weakly excites
- L20:1802 weakly excites
- L19:1795 weakly excites
- L21:2875 weakly excites
- L11:1844 weakly excites
- L25:1648 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
- ños
- ancestry
- 世家
- 姓
- thổ
- ño
- 遗传
- 天赋
- 天生
- born
Suppresses
- urg
- $action
- ]*
- UFC
- whel
- Massage
- eu
- action
- cn
- 愦