book numbers before a verse
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 spot right before a Bible verse number, like the gap in "Matt. 25:31" or "1 Tim. 3:8". When turned up the model drifted toward talk of gods and goddesses, which is related in theme but not clearly the same thing.
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.
-
of Creation) will come the Eighth Day, and then “the Son of Man shall come in His Glory and all the holy Angels with Him.” (Matt. 25:31)⏎St. Dionys
-
. . endless genealogies, which cause disputes rather than godly edification which is in faith (1 Tim. 1:4; cf. Titus 3:9). In the modern
-
of personhood. Therefore, the fact that we cannot see God as a physical being should not weaken our assurance that he is a real person.⏎Now read John 14:8"Philip said, 'Lord, show us the Father and
-
of modern warfare is that it provides the opportunity to those who possess modern scientific weapons—especially atomic, biological, or chemical weapons—to commit such crimes. (CCC 2314; cf. Gaudium et Spes 80
-
conate is devoted to serving in all other areas of the church.”[ii]⏎Read 1 Tim. 3:8-13.⏎What are the qualifications for deacons? How
-
ye have both learned, and received, and heard, and seen in me, do: and the God of peace will be with you” — Philippians 4:9⏎The theory or norm of what is essential in one’s life
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 believe in a god. I believe in a spirit, a god, and a goddess. I think of deities in many deities, and I have a lot of rituals, offerings, and good deo-ban deo-
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
purpose of the product, which might be used for interior decor (windows, beds and ceilings), festivities (weddings, birthdays and ceremonies), or prayers (invoking the rain, warding off the devil, and so on). As a key
-
ba also performed traditional rites and sacrifices to satisfy the demands of Yoruba Traditional Religion.⏎In 1914
-
of the setting sun at the time of summer solstice. Just north of this temple are the foundations of an earlier, but also central and primary, temple dedicated to the god Montu. Little remains of this temple, not because it was weather
-
, and that almost every king of the Middle Kingdom left some mark of his presence at Karnak. The central temple at Karnak was dedicated to the state god, Amon, and is directionally
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
- L9:2289 suppresses
- L18:3038 suppresses
- L14:985 excites
- L20:2455 excites
- L15:2805 excites
- L15:2694 excites
- L19:2546 suppresses
- L9:2060 suppresses
- L7:247 suppresses
- L6:2059 suppresses
- L20:2701 suppresses
- L4:2244 suppresses
- L3:1012 suppresses
- L16:1378 suppresses
- L20:2284 weakly excites
- L7:371 weakly suppresses
- L6:2081 weakly suppresses
- L8:2948 weakly excites
- L6:2963 weakly suppresses
- L4:1310 weakly suppresses
- L6:1923 weakly excites
- L20:2019 weakly suppresses
- L6:2545 weakly suppresses
- L11:847 weakly excites
Feeds later layers
- L26:994 suppresses
- L27:944 suppresses
- L25:2528 suppresses
- L23:2914 suppresses
- L27:943 suppresses
- L25:2148 suppresses
- L22:1923 suppresses
- L25:2222 suppresses
- L27:2262 suppresses
- L24:2871 suppresses
- L25:699 suppresses
- L23:1676 weakly excites
- L23:1705 weakly suppresses
- L25:3051 weakly suppresses
- L24:2028 weakly suppresses
- L25:1956 weakly suppresses
- L26:1177 weakly suppresses
- L26:1044 weakly suppresses
- L26:1430 weakly suppresses
- L23:2851 weakly suppresses
- L26:1689 weakly suppresses
- L23:827 weakly suppresses
- L27:847 weakly suppresses
- L22:148 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
- ldata
- 睫
- 髁
- 二手车
- sert
- fcc
- Mellon
- 冠军
- 网球
- =train
Suppresses
- 神
- gods
- god
- goddess
- God
- deity
- 神话
- 神圣
- spirits
- 信仰