opening parenthesis in 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. 12 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 opening bracket that comes right after an author name in a reference list, such as 'Dyson, A. (2008)'. It is mostly quiet elsewhere, and the negative side shows up on words like 'this' or 'other' in technical sentences. Turning it up gave text with no clear link to citations.
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.
-
cornerstone of all subject areas and the Common Core. Revising to create “showing” sentences systematically helps improve writing competence.⏎Dyson, A. (2008). Playing with textual toys. In Flood, J., Heath
-
upplement 5): S213–S256.⏎- Redberg RF, et al. (2009). ACCF/AHA 2009 Performance measures
-
times of several years. The total energy radiated in a superburst is 1042 erg. Superbursts<|endoftext|>Kang, T-J. (2018) Online learning in North Korea The Diplomat, May
-
can combine this with all the other skills that are measured so that a good picture of and for the student emerges.⏎<|endoftext|>Junaidi, Juhazren (2007) Assisting learning using computers through the application of animated ped
-
of their writing. The same principle applies to adolescents and young adults.⏎- External stimuli prompt students to construct their own sentences and stories.⏎Smith, C.H. (2010). “Diving in deeper”: Bringing basic writers’ thinking
-
⏎Lin CC, Lin XM, Zou Y, Zhang L (2005) Study on the relationship between meteorological conditions and acid rain
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 think I have been thinking for 10 minutes, and I have been thinking... and I have been thinking... and... 10 minutes, and I have been thinking... and... 10 minutes... and... 1
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
was the most important covariate determining survival of adult D. couperi, suggesting lower survival for larger individuals within each sex. We are uncertain of the mechanisms underlying this result, but possibilities may include greater resource needs for larger individuals within each sex,
-
conversion speed regulation of the AC asynchronous motor, improve the operation precision, change the power factor, and Realize overcurrent, overvoltage, overload protection and other functions. The main circuit of the frequency converter can be roughly divided into two categories
-
classified as passive or active. In principle, a passive tracer blindly follows the fluid phase in which it is injected. Interpretation of tracer-production curves must account for this. The results from the application of active tracers may give information about fluid saturation
-
colleagues demonstrated that the tool could reveal signs of pulmonary fibrosis in COVID long-haulers, thus helping to account for shortness of breath, coughing and other lung troubles. A diagnosis, he suggests, that would be impossible with standard CT
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
- L9:1059 weakly suppresses
- L6:1629 weakly excites
- L2:1203 weakly suppresses
- L11:272 weakly excites
- L10:1202 weakly excites
- L6:1282 weakly suppresses
- L11:2912 weakly suppresses comma closing an introductory clause
- L6:2654 weakly suppresses
- L5:2497 weakly excites
- L8:303 weakly suppresses
- L10:2913 weakly excites
- L3:1068 weakly excites "that" pointing back to a noun
- L10:1542 weakly suppresses
- L0:1983 weakly suppresses
- L10:2144 weakly excites
- L4:2470 weakly excites
- L1:2360 weakly excites
- L6:1974 weakly excites
- L6:454 weakly suppresses
- L9:1878 weakly suppresses
- L7:504 weakly excites
- L1:1090 weakly excites
- L5:2689 weakly suppresses
- L8:1219 weakly suppresses
Feeds later layers
- L14:1402 suppresses
- L24:150 suppresses
- L16:284 suppresses
- L17:1638 suppresses
- L16:2165 suppresses
- L17:1858 suppresses
- L16:1711 suppresses
- L13:641 suppresses
- L18:156 suppresses
- L21:124 excites
- L19:1616 weakly suppresses
- L17:1022 weakly suppresses
- L18:675 weakly suppresses
- L16:675 weakly suppresses
- L14:2349 weakly excites
- L21:1722 weakly suppresses
- L20:1372 weakly suppresses
- L15:1437 weakly suppresses
- L13:1320 weakly suppresses
- L14:2156 weakly suppresses opening parenthesis after a term
- L25:339 weakly suppresses
- L20:271 weakly excites
- L14:825 weakly suppresses
- L18:2173 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
- WithEmail
- istory
- �
- Ballet
- izard
- glColor
- 甲状
- fä
- ogn
- 遇上
Suppresses
- systemic
- empre
- hai
- riors
- REW
- arching
- finity
- ALSE
- lesi
- corporate