spot just before a 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. 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 year or a date is written, such as 'the 60s' or 'in 1974'. When it was turned up, the model wrote long runs of digits, which fits its interest in numbers.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
hit with their song “Hey Schoolgirl”. The duo’s real names were Paul Simon and Art Garfunkel. Their music and lyrics are synonymous with the 60’s. From “The Sound of Silence” to “Bridge Over Trou
-
3/74 that ended the "economic miracle" once and for all. By the time Herbert Grünewald succeeded Kurt Hansen as Management Board Chairman following the 1974 Annual Stockholders' Meeting, the global economy was undergoing a
-
There is no doubt that England is far more ethnically and racially diverse than it was 20 years ago. The population identifying with an ethnic group other than White grew
-
segregationists, whatever one thought of Rosa Parks, her sit-down in 1955 in Montgomery, Alabama, was the ultimate moment of truth in a 350-year-long crisis and everybody knew it. What she did demanded a
-
, Canadian extreme weather expert and former Environment Canada research scientist. “Extreme events in the past few years are occurring with about the same frequency as they did during the 1945-1977 period when the Earth was generally cooling
-
to be prohibitively expensive and time-consuming due to self-erasing code and the fact that code can be several hundred thousand lines in length, according to the 2008 ACCURATE Report (A Center for Correct, Usable
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: . . . . . . . . . . . . . . .
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Bubbling fluidized bed boilers; Circulating fluidized bed boilers⏎- Front<|endoftext|>CNY - 中國元,人民幣⏎最常用 CNY 交叉匯率⏎為什麼您
-
UNSPSC Code 30260000 See more.Customers who bought this item also bought.Page 1 of 1 Start over Page 1 of 1 .Previous page.1/4 x 6 x
-
�� : sleep 不 : not 觉 : to feel, wake up 晓 : dawn⏎Meaning for Lien 2 处 : place 闻 : hear 啼 : cry 茑 : bird⏎Meaning for
-
to make it accessible to us. But when the salt is removed it also removes all of the minerals in the water, creating pure water; just H 2 0. This is called distilling. But I wondered if distilled water could sustain
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
- L19:581 strongly suppresses blank slot before a number
- L18:377 excites number right after 'about' or 'less than'
- L17:402 suppresses
- L19:26 suppresses
- L4:53 excites blank slot before a number
- L20:265 excites
- L18:122 excites
- L20:749 excites
- L19:165 excites
- L17:285 suppresses
- L20:629 excites
- L18:375 suppresses
- L15:307 suppresses
- L16:471 suppresses blank spot before a year
- L16:975 excites
- L15:12 excites
- L19:737 suppresses
- L5:104 excites
- L13:923 suppresses word right before a number or year
- L17:620 suppresses
- L16:363 suppresses
- L16:406 weakly excites
- L8:111 weakly suppresses the space before a year in dates
- L17:773 weakly excites
Feeds later layers
- L26:103 strongly excites blank space before a number
- L27:283 strongly excites the spot right before a number
- L26:15 suppresses
- L23:77 excites the space right before a number
- L25:33 excites spot where a number is due
- L27:170 excites unclear: spaces in mixed-language text
- L25:146 excites
- L25:182 excites
- L26:611 suppresses
- L26:216 excites blank slot before a year or number
- L26:121 excites
- L27:113 excites
- L26:172 excites
- L27:238 excites
- L22:372 excites empty spot before a number or "3D"
- L22:824 excites blank before an expected number
- L27:306 suppresses
- L24:197 excites
- L22:16 suppresses comma in dates and birth lines
- L27:48 suppresses
- L25:253 suppresses
- L27:33 excites
- L27:2172 excites
- L27:241 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
- ước
- ionic
- istringstream
- ropa
- allback
- ills
- ération
- 的概念
- uling
- 设计理念
Suppresses
- covery
- wave
- out
- →
- O
- ten