"addition" after "in" ("in addition")
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 the word "addition" when it follows "In" or "in", as in "In addition, ...". When it was turned up, the writing got repetitive and drifted into odd loops, with no clear theme.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
SCDAP/RELAP5 is used for core degradation calculations and model assessment using single effect tests or integrated experiments (QUENCH, PARAMETER, ...) in addition to<|endoftext|>Nowadays, a large number of people tend to believe that hemp and
-
any installation including a single MV/LV transformer, provided that the rated power of the transformer remains below the limit fixed by the local utility supplying the installation.⏎In addition to the functional requirements the construction of both main and secondary substations shall comply with
-
create cookies for a different domain. In addition, browsers can also allow users to view and delete cookies individually.⏎Availability of more
-
of play-based pedagogies. In addition, a large majority of teaching staff is female.⏎Similar to what research has found
-
with children who have developmental delays such as Down Syndrome, Autism, ADHD and other similar conditions.⏎A number of other rehabilitation specialists may be found in some settings in addition to the three main ones listed above. Some of these are vision therapists, recreational
-
chains of linked cells. In addition, the toxin-releasing self-destruction (autolysis) typical of pneum
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: the first one! So I am going to do it, the second one! So I am going to do it, the third one! So I am going to do it, the fourth one! So I am going to do it,
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
coni Anemia pathway (FANC), as well as the p53 signaling pathway. B. In response to intrastrand crosslinks (crosslink), ATR and its interacting partner ATRIP become activated and phosphorylated leading to the induction
-
, as the PI3 kinase ATM can be triggered by phosphorylation in response to two times strand breaks in DNA (DSBs), a related PI3 kinase ATR and its own interacting partner ATRIP react to intrastrand crosslinks in DNA
-
, although hatching may take up to two to three weeks in colder climates.<|endoftext|>When It Is Good To Be Bad : Medieval Indo-Tibetan Tantric Buddhist Apologetics (PDF)⏎Antinomianism comes in
-
the point of starving or exercise even if they are underweight.⏎It’s all in the mind⏎Shrouded under the cloak of an eating disorder, anorexia nervosa has a psychological element to it. People with anorexia may
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
- L2:483 weakly suppresses
- L2:1778 weakly suppresses
- L0:1224 weakly suppresses
- L0:178 weakly suppresses
- L2:2975 weakly suppresses
- L0:1301 weakly suppresses
- L2:950 weakly suppresses
- L2:1036 weakly excites
- L1:1861 weakly suppresses
- L2:1690 weakly excites
- L2:380 weakly suppresses
- L1:2333 weakly excites
- L1:2396 weakly excites
- L2:1494 weakly suppresses
- L2:2810 weakly excites
- L2:75 weakly suppresses
- L0:770 weakly excites "of" after minority, tells or news
- L2:1891 weakly excites
- L2:3046 weakly suppresses
- L0:2328 weakly excites
- L0:1669 weakly excites
- L1:24 weakly excites
- L0:2537 weakly excites
- L2:49 weakly suppresses
Feeds later layers
- L6:144 weakly suppresses
- L13:1610 weakly suppresses
- L10:1262 weakly excites
- L4:2771 weakly excites
- L17:1553 weakly excites
- L4:1058 weakly excites
- L5:952 weakly excites
- L4:1949 weakly excites
- L12:1497 weakly excites
- L7:933 weakly suppresses
- L18:3065 weakly excites
- L17:2158 weakly excites
- L7:1387 weakly excites
- L7:2350 weakly excites
- L15:2754 weakly excites
- L19:1382 weakly excites 'addition' in 'In addition to'
- L4:2117 weakly excites
- L26:2998 weakly suppresses
- L6:1901 weakly excites
- L8:1815 weakly excites
- L21:1293 weakly suppresses
- L6:220 weakly excites
- L26:2539 weakly suppresses
- L18:927 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
- ign
- igne
- ernational
- 0
- 忽略
- %\
- IGN
- iliar
- itch
- 设计
Suppresses
- annes
- 骅
- ktion
- тся
- ialis
- äu
- Hera
- IED
- ��
- bable