the 'm' in a.m. and p.m
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. 23 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 'm' in time markers like '7:30 a.m.' and '2 p.m.', and also on the 'C' in 'B.C.E.'. Turning it up made the model produce stray full stops and line breaks, so the steering output does not show a clear theme.
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.
-
SCD is essential for pediatric and adult hematologists. This common disease will be explored in depth in an Education Program Session today at 7:30 a.m. and again at 2:00 p.m. in rooms 3
-
be reckoned with.⏎Fort Walla Walla Museum's Living History Company will revisit those days at 2 p.m. Sunday.⏎Members of the Living History Company including Sam Pambrun as interpreter
-
of Oregon State University’s Outreach in Biotechnology program, will discuss the pros and cons of gene technology for agriculture.⏎The Science Pub presentation begins at 6 p.m. in the Majestic Theater, at 115 S.W. Second
-
the accomplishment of the Apollo 11 mission, Hardin Planetarium will be hosting a special, day-long “Moon Landing Celebration” beginning at 1 p.m. July 20.⏎The event will feature movie screenings, hands
-
results of the previous year’s RD&D program and obtained input regarding its intended spending for the following calendar year.⏎9:30 a.m. – 2:35 p.m.⏎VIEW THE RECORDING *⏎To
-
space flight, but built and flew the world’s first liquid propellant rockets.”⏎The event will take place from 10 a.m. to 3 p.m. on Saturday, June 11, on the Mall in front of the
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 think I, or I, or I, or I, or I, or I, or I, or I, or I, or I, or I, or I, or I, or I, or I, or I
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
is quite advanced than other shoe gadgets available in the market thanks to its usability patented algorithms for step analysis.⏎Krishna started his startup Fimo Technologies and services Pvt ltd at Kondapur in 2017. “I and my
-
did so far:⏎Originally Posted by MrFujin⏎Enter a 5-charcter string: hgout⏎String
-
our blog posts for Census week, we showcase recent research undertaken using Census data to explore the remarkable growth and stability of multi-ethnic neighbourhoods in England, by Dr Gemma Catney, Professor Richard Wright, and Professor Mark Ellis. The original
-
omas.⏎"Clearly there's more to cancer than just genetic mutations. Not every cancer is going to fit that box of having a genetic driver," says Sriram Venneti, M.D., Ph.D., assistant professor of pathology at the University
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
- L22:1733 suppresses
- L12:50 suppresses
- L4:1572 excites
- L17:70 excites
- L22:767 suppresses
- L18:1471 suppresses
- L22:470 excites
- L19:696 excites
- L6:35 excites
- L19:699 excites
- L21:189 suppresses document break token mid-sentence
- L16:38 suppresses
- L8:79 suppresses
- L20:317 excites
- L3:2024 suppresses
- L22:164 suppresses
- L5:2525 suppresses
- L19:1708 weakly excites
- L17:316 weakly suppresses
- L5:667 weakly suppresses
- L21:749 weakly suppresses
- L6:293 weakly suppresses
- L14:1498 weakly excites
- L1:2435 weakly excites
Feeds later layers
- L27:151 excites
- L24:31 suppresses
- L26:356 excites
- L26:21 suppresses
- L27:156 excites
- L24:569 excites
- L27:2519 excites
- L24:39 excites
- L26:1391 excites
- L24:429 excites
- L25:17 excites
- L24:228 excites
- L24:144 weakly suppresses
- L26:165 weakly excites
- L27:6 weakly suppresses
- L25:30 weakly excites
- L25:651 weakly excites
- L27:133 weakly suppresses
- L27:80 weakly excites
- L27:11 weakly suppresses unclear: mixed fragments in lists and pronunciations
- L26:1602 weakly excites
- L24:264 weakly excites
- L27:1324 weakly excites
- L24:257 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
- 哇
- 必ず
- phan
- rones
- ieces
- 城县
- isons
- epad
- .,
- prestashop
Suppresses
- most
- past
- verse
- imitives
- I
- unt
- Most
- <{
- itself
- tn