opening bracket before a citation
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 an opening bracket that comes right after a quote or a claim and starts a citation, such as "(Morton, 1999)". Its negative side is on the name "Judas", which is a different pattern. Turning it up did not give a clear change in the writing.
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.
-
exists a “relation between the perceiving mind and the physical world. However, the real qualities of objects are not defined by color, warmth, and taste” (Morton, 1999 p.47). Conversely, Desc
-
et al., 2011). Paper wasps are likely to influence many other species in terrestrial ecosystems because they are voracious predators of invertebrates (Ward & Ramón-Laca, 2014). Four species
-
decline within many parts of their range (Rabinowitz & Winter, 2006).⏎Carnivore populations are critically important to maintaining healthy ecosystems (Terborgh et al., 1999; Terborgh et
-
other white men at other schools did the same. That is why we now have so many Navajo families like our own with the last name of Begay” (22). Bruchac is skilled in incorporating information into his narrative that inspires
-
poor, the present results agree with Sheriff’s criticism regarding teaching ethics, once ethics “is an inherent quality learn as part of a development process of a personality” (4). It was clear that the children knew about the prevalent types of morality in
-
hou et al. 2004; Grimm et al. 2008). Consequently, growth in urban development is a major cause of extinction (Czech & Krausman 1997), with the native species
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 lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
:1-2 shows that the place where Jesus was to be betrayed was a place that He went often. His disciples also went to the garden often. Judas, one of His disciples, was not just some mere acquaintance to Christ. He
-
the events. In Matthew 26:25, Judas asked Jesus if it was he who would betray Him. He had already sealed the
-
it to name their children because of its negative connotation. John 13:18-30 records the events leading up to the betrayal by Judas. Many of the details leave some
-
13 is not the first time one reads of Judas and his dishonesty. In John 12:1-8, we read that Judas was upset with Mary, the sister of Lazarus, because she was using costly
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
- L2:53 weakly excites
- L2:111 weakly suppresses
- L2:2693 weakly suppresses
- L2:1321 weakly suppresses
- L1:515 weakly suppresses 'about' in 'it's not about' phrasing
- L1:1420 weakly excites
- L1:1766 weakly suppresses
- L0:1204 weakly suppresses
- L2:1702 weakly suppresses
- L0:1610 weakly suppresses
- L0:2050 weakly suppresses
- L1:1788 weakly excites
- L2:1292 weakly suppresses
- L1:2638 weakly excites
- L1:2089 weakly excites
- L1:2697 weakly suppresses
- L1:1226 weakly excites
- L1:1476 weakly suppresses
- L2:2881 weakly excites
- L2:845 weakly excites
- L1:1687 weakly excites
- L2:1138 weakly suppresses
- L1:2737 weakly suppresses
- L2:1178 weakly suppresses
Feeds later layers
- L9:159 suppresses
- L6:1181 excites
- L8:2562 weakly excites
- L8:248 weakly suppresses
- L10:2398 weakly excites
- L27:1899 weakly suppresses
- L25:1133 weakly suppresses
- L21:2609 weakly suppresses
- L8:2893 weakly suppresses
- L5:209 weakly excites
- L4:276 weakly excites
- L4:80 weakly excites
- L11:1874 weakly suppresses
- L4:2860 weakly suppresses
- L4:1453 weakly excites
- L19:1118 weakly suppresses
- L9:2544 weakly excites
- L13:1344 weakly excites
- L24:2498 weakly excites unclear: word before an event or name
- L13:480 weakly suppresses
- L22:2369 weakly suppresses
- L19:3003 weakly suppresses
- L14:1571 weakly excites
- L7:1730 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
- lass
- ted
- sonian
- antz
- orte
- entes
- ter
- endez
- ).⏎
- ).
Suppresses
- 其它问题
- itlement
- Mods
- fileName
- republic
- imeline
- itution
- uracy
- idelity
- owość