period in abbreviations and initials
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. 3 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 full stops that come after a short abbreviation, such as 'c.' for 'circa' in dates or after a person's initial in a citation. It goes negative on the period after initials in author names. Turning it up or down did not clearly change the 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.
-
topic.⏎Outer Kingdoms of the First Millennium BC⏎MITANNI – c. 1150-950 BC⏎Mitanni was a multi
-
outside vascular bundles; seed solitary.⏎Distribution and occurrence: World: c. 65 genera, c. 900 species, cosmopolitan.
-
3. Demonstration on pest management technologies like pheromone traps, fruit bagging, and etc.<|endoftext|>Little Wound (c. 1835–Winter 1899;⏎His father Chief
-
a province within the Middle Assyrian Empire.⏎ Some historians place the Kassites and Hurrians within the bounds of Mitanni.⏎The KASSITES – c. 1184-763⏎When Matanni rose in northern
-
in range margins and community composition in response to global change. In the European Alps, plant species are shifting their range to a higher elevation with a speed of c. 3 – 5 m per year, and non-native species are moving even
-
a version counting back to 3076 BCE is still in use in parts of India today.⏎Astronomy flourished under the Gupta Empire (c. 320-550 CE) during which time Ujjain
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 need to be honested in the last. I have no idea what it is. So I have no idea what it is. But I have no idea what it is. I am not sure of what I am. I am not
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
systems (H. Qammari, A. Venkatesan). A vacation on Mars - an artist's journey in a computer graphics world (C.A. Pickover). Part II. Attractors. Automatic generation of strange attractors
-
your brain's aging process. But what is up for debate is whether or not alcohol<|endoftext|>Using groundwater for hydro electric power generation⏎Bardsley, W.E. (1995). Using groundwater for hydro electric power generation. Journal of
-
, 1967)⏎- Mitsukurina owstoni Jordan, 1898 (Goblin shark)⏎- Jordan, D.S. (1898). "Description of a species of fish (Mits
-
Federal Death Penalty" Washington Law Review (2010).<|endoftext|>Last updated 2/27/15⏎Isothermal Community College⏎P.O. Box 804, Spindale NC 28160
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
- L1:1795 excites
- L2:2614 weakly excites
- L1:331 weakly suppresses
- L2:2837 weakly suppresses
- L0:1336 weakly excites
- L0:1978 weakly suppresses
- L2:1321 weakly suppresses
- L0:2984 weakly suppresses
- L1:353 weakly suppresses
- L0:91 weakly suppresses
- L0:314 weakly excites
- L2:2100 weakly suppresses
- L0:1042 weakly suppresses "This" starting a labelled sentence
- L1:1676 weakly suppresses
- L0:12 weakly suppresses
- L1:128 weakly excites
- L0:553 weakly excites second half of a split name
- L2:666 weakly excites
- L0:76 weakly suppresses
- L1:223 weakly excites unclear: 'selected' and 'shape' in mixed text
- L0:2974 weakly excites
- L1:63 weakly suppresses
- L2:460 weakly suppresses
- L1:3049 weakly suppresses
Feeds later layers
- L4:204 weakly suppresses
- L6:2493 weakly excites
- L4:907 weakly excites
- L24:2695 weakly suppresses
- L4:2942 weakly excites 'yet' and 'haven't' after have
- L12:878 weakly excites
- L20:2181 weakly excites
- L4:744 weakly excites
- L5:2284 weakly excites
- L4:1006 weakly excites
- L6:2812 weakly excites
- L5:593 weakly excites
- L7:1522 weakly suppresses
- L15:1953 weakly excites unclear: document ends and number gaps
- L9:1943 weakly suppresses
- L14:2420 weakly suppresses
- L26:1155 weakly suppresses
- L6:2554 weakly excites
- L4:2496 weakly excites
- L19:446 weakly excites
- L9:2665 weakly excites
- L6:439 weakly excites
- L6:357 weakly excites
- L27:953 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
- udson
- ars
- .Diagnostics
- /rs
- sburg
- 血压
- setTitle
- .setUp
- veau
- .setHeight
Suppresses
- `'
- Ė
- ngo
- licas
- PageIndex
- board
- .%
- 弋
- pretending
- vox