opening parenthesis in formulas and citations
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 an opening bracket '(' that starts a formula term or a copyright notice, such as 'Copyright (c)' or 'n = N / (1 + ...)'. The activations are fairly weak. Turning it up did not produce a clear theme in the generated text.
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.
-
the behavior of a population is known at all. * How to Use Slovin's Formula⏎Slovin's formula is written as:⏎n = N / (1 + Ne^2)⏎n = Number of samples⏎N = Total population
-
of the two overloaded versions provided in C++.⏎/* memchr example */⏎int main ()⏎char * pch;⏎char str = "Example string";⏎pch = (char
-
water, were the major factors shaping phytoplankton communities.⏎- Abstract views: 1777⏎- PDF: 842⏎Copyright (c) 2009 Silvia TAVERNINI, Daniele N
-
viable. A symbol or instrument found in the Tuturangiana Andala ritual of green bamboo, betel nut, black goat, and red flag.⏎Copyright (c) 2022 Proceedings Of International Conference On Communication Science⏎This
-
Here are examples of creating markers, setting markers, and moving point to markers:⏎;; Make a new marker that initially does not point anywhere: (setq m1 (make-marker)) ⇒ #<marker in no buffer>⏎m1to point between
-
leaves to refer to their objects⏎> through a Map. Then changing a leaf object requires only that the value⏎> associated with key represented by the leaf be (re-)inserted into the⏎> Map. The Tree itself need not change
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 have been thinking about it, and I feel it. I have been thinking about it, and I feel it. I feel, I- I- I- I- I- I- I- I- ... I- I-
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Satellite data can help scientists understand:⏎- The CO2 pump: remote sensing of phytoplankton (green plant-life) in the ocean.⏎- Sea Surface Temperature variation: which shows the temperature variation, dynamics and some chaos – as shown
-
behavior of a population is known at all. * How to Use Slovin's Formula⏎Slovin's formula is written as:⏎n = N / (1 + Ne^2)⏎n = Number of samples⏎N = Total population⏎
-
ave and another convex, which it termed perspicillum: was born the first astronomical telescope refractor. The manufacturer of lenses designed it with the idea of observing distant objects on the Earth's surface,
-
risk factors. Mark Goldstein, 79,<|endoftext|>Sea surface temperature⏎Sea surface temperature (SST) is the water temperature at the surface. In practical terms
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
- L8:3 excites
- L6:144 excites
- L9:796 suppresses
- L8:7 excites
- L8:12 excites
- L3:464 weakly excites
- L8:13 weakly suppresses
- L8:1765 weakly excites
- L2:28 weakly excites
- L8:2143 weakly excites
- L8:16 weakly suppresses
- L3:843 weakly suppresses
- L5:1993 weakly excites
- L9:1532 weakly suppresses
- L7:288 weakly excites
- L3:1693 weakly excites
- L6:43 weakly suppresses
- L7:2665 weakly suppresses
- L8:10 weakly excites
- L5:2107 weakly suppresses
- L7:76 weakly excites
- L4:2793 weakly suppresses
- L8:2776 weakly excites
- L4:779 weakly suppresses
Feeds later layers
- L16:995 strongly excites
- L14:1318 excites "both" and "either" before alternatives
- L11:1272 excites unclear: sentence ends in factual prose
- L14:1274 excites
- L11:2310 excites opening parenthesis in formulas and math
- L19:1497 weakly excites
- L15:2042 weakly excites
- L16:2034 weakly suppresses
- L13:2444 weakly excites
- L15:801 weakly excites 'of' inside organisation names
- L13:2590 weakly excites
- L13:1022 weakly excites
- L18:217 weakly excites
- L15:1794 weakly excites
- L13:2903 weakly excites
- L18:1459 weakly suppresses
- L23:2411 weakly suppresses
- L26:1018 weakly excites
- L16:1513 weakly excites opening parenthesis before a numbered list
- L19:2390 weakly suppresses
- L22:246 weakly suppresses
- L17:3071 weakly suppresses
- L13:1698 weakly excites
- L15:24 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
- rias
- otos
- \",\
- obar
- pto
- airo
- 专线
- rios
- éri
- etAddress
Suppresses
- onnen
- idelity
- わけではない
- ╝
- almart
- borg
- ises
- ству
- .instance
- esting