comma in dates and birth lines
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. 22 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 comma that separates the day from the year in dates, such as "September 24, 1896" or "March 26, 1874", mostly in biographies and reference entries. Its negative side is very weak. Steering results were scrambled or off topic, so they do not show what it does.
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.
-
– in contrast to chemical pesticides – have no adverse effects on animals or humans.<|endoftext|>F. Scott Fitzgerald⏎Francis Scott Key Fitzgerald (September 24, 1896 – December 21,1940)
-
of the third week?⏎3. Solve for the following.⏎a. In a geometric sequence, and . Find .⏎b. In a geometric sequence, and . Find .⏎In Example 1b and Review Questions 5 and 6, students
-
panels can be installed on rooftops or on the ground.⏎- Sunpower solar panels<|endoftext|>By Mortimer J. Adler⏎Aristotle (384 - 322 B.C.) taught common sense to Alexander the nice and,
-
Pound, 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
-
Russia|⏎|-||Declared1||November 18, 1918|⏎|-||Recognized||January 26, 1921|⏎|-||Soviet occupation||August 5
-
non-alignment and of socialism grounded in Islam.⏎Mohammad Hatta was born in Bukittinggi, West Sumatra, Indonesia, on August 12, 1902. Although his father died while he was an infant,
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:ThetransformationalimpactofDigitalSolutionsintotheBusinessModelofTraditionalBusinessModelofDigital-Platform-Applications.在Iamamostofamaintainingthetraditionalbusinessmodelofdigital-priv-priv-priv-priv
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
and alternating current in the series RLC circuit at an instant are given by⏎v=Vm sin ωt and i = I m sin( ω t + φ)⏎where ϕ is the phase angle between υ and i.
-
.<|endoftext|>Start a 10-Day Free Trial to Unlock the Full Review⏎Why Lesson Planet?⏎Find quality lesson planning resources, fast!⏎Share & remix collections to
-
the same amount of time running at a pace of 5 miles per hour. A study in the May 1996 issue of the "Journal of the American Medical Association" also concluded that running on a treadmill is superior to using the
-
significant.⏎“Even total mortality went in the right direction,” said Gaziano, whose new data comes from the Physicians Health Study II published Wednesday in the Journal of the American
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
- L19:963 strongly suppresses
- L18:246 strongly suppresses
- L17:15 strongly suppresses
- L20:59 strongly excites
- L19:763 strongly suppresses
- L14:140 strongly excites
- L19:1133 strongly suppresses last word before end of document
- L19:430 strongly suppresses
- L19:1174 strongly suppresses
- L19:1172 strongly suppresses
- L20:433 strongly excites
- L19:1005 strongly suppresses
- L20:193 strongly suppresses
- L14:309 strongly suppresses
- L21:164 strongly suppresses
- L16:723 strongly excites
- L18:2497 strongly suppresses hyphen inside compound words
- L6:42 strongly excites
- L4:2129 strongly excites
- L21:214 strongly excites
- L20:489 strongly suppresses
- L19:674 strongly suppresses
- L19:2177 strongly excites
- L19:981 strongly suppresses
Feeds later layers
- L27:33 strongly suppresses
- L24:39 strongly suppresses
- L26:21 strongly excites
- L25:109 strongly excites
- L25:161 strongly excites
- L24:51 strongly suppresses
- L25:30 strongly suppresses
- L27:263 strongly excites
- L27:71 strongly excites
- L27:112 strongly suppresses
- L23:2741 strongly excites
- L23:80 strongly suppresses blank slot before a year or number
- L27:97 strongly suppresses
- L27:117 strongly suppresses
- L27:825 excites
- L25:311 excites
- L23:60 suppresses
- L27:85 suppresses
- L27:11 excites unclear: mixed fragments in lists and pronunciations
- L27:126 suppresses
- L26:189 suppresses
- L24:609 excites
- L24:33 excites
- L25:112 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
- .⏎
- ,
- ),
- ,
- ).
- )
- 。
- .
- )(
- ):⏎
Suppresses
- 的作用
- 的工作
- *self
- 的对象
- sar
- 'S
- 的颜色
- Cx
- "She
- 幸运