colon in clock times
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. 14 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 colon that separates hours from minutes in a time stamp, like '08:57' or '11:07 am'. It is a clock-time detector. Turning it up gave text that did not obviously relate to times, so the writing effect is unclear.
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.
-
Jan 09, 2017 08:57 AM EST⏎MIT Professor Is Building Bio-Inspired Robots To Save Lives⏎Dis
-
turn directions, UI, User interface, video chat, Virtual reality, Virtual reality glasses⏎January 18, 2012 • 11:07 am
-
But that merit might be offset if the Chinese consume relentlessly like<|endoftext|>Lev Bely, 07.05.2012 21:56⏎There are still butterflies in Mumbai albeit they're increasingly getting lack of
-
||World match.|⏎|Next Article:||Ukraine's Orange Revolution.|<|endoftext|>posted on November 17, 2014 23:23⏎Scientists from the WCS (Wildlife Conservation Society), NASA, and
-
SeaWeb/FIS)⏎Seafood producers tackle ocean acidification at annual Summit⏎Monday, February 21, 2011, 23:50 (GMT + 9)⏎From 31 January to 2
-
subject is much smaller than a human. Essential oils are highly concentrated and should<|endoftext|>Monday, October 5, 1998 Published at 17:47 GMT
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 can't, but I can. I can. I can. I can. I can. I can. I can. I can. I can. I can. I can. I can. I can. I can. I
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
regulating your intake and staying healthy, both physically and mentally.<|endoftext|>Tech moves fast! Stay ahead of the curve with Techopedia!⏎Join nearly 200,000 subscribers who receive actionable tech insights from Techopedia.⏎A
-
oes, tsunamis, global climate, and weather.⏎Based on your research, do the following:⏎Support your statements with examples. Use a minimum of six reliable references, two of which should be peer-reviewed articles.⏎Write a 7
-
July 10, the fire had damaged or destroyed nearly 130 buildings, injured over 20 people, and cut the lives of 19 Granite Mountain Hotshots woefully short.⏎Sparked by Lightning, Fueled
-
established in 1996 to recognize, honor, and promote the outstanding contributions women make to the scientific and technological communities that improve society. Each year, five women are selected from around the world to receive this honor, and Trayanova
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
- L7:1524 suppresses
- L10:993 excites
- L4:87 suppresses
- L3:203 weakly suppresses
- L10:50 weakly excites
- L7:224 weakly suppresses
- L7:830 weakly suppresses
- L8:582 weakly suppresses
- L4:538 weakly suppresses
- L12:408 weakly suppresses
- L13:408 weakly suppresses
- L4:201 weakly excites
- L8:758 weakly excites
- L9:183 weakly suppresses
- L13:2861 weakly suppresses
- L11:93 weakly excites
- L13:1215 weakly suppresses
- L7:723 weakly suppresses
- L13:1299 weakly suppresses
- L13:1297 weakly suppresses
- L13:460 weakly suppresses
- L13:2705 weakly excites
- L0:2820 weakly suppresses
- L12:953 weakly excites
Feeds later layers
- L16:314 excites
- L17:2532 suppresses
- L16:122 excites
- L18:467 weakly excites
- L18:411 weakly suppresses
- L19:785 weakly suppresses
- L21:1972 weakly suppresses
- L16:2787 weakly suppresses
- L24:1032 weakly excites
- L21:299 weakly excites
- L15:1080 weakly suppresses
- L27:1321 weakly suppresses
- L26:682 weakly suppresses
- L21:110 weakly suppresses
- L17:248 weakly excites
- L24:2630 weakly excites
- L25:2001 weakly suppresses
- L26:1246 weakly suppresses
- L19:42 weakly excites
- L17:167 weakly excites
- L25:182 weakly suppresses
- L24:272 weakly suppresses
- L16:1212 weakly excites
- L26:1451 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
- leurs
- SON
- son
- 落实
- 突破
- s
- 適
- 依靠
- 加重
- H
Suppresses
- iation
- /reference
- iative
- iated
- iator
- attles
- 骸
- iate
- bruary
- Pompeo