opening parenthesis in citations and asides
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. 16 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 follows a citation title, a year, or a copyright line, such as 'Copyright (c)' or 'Volume (Year)'. Its strongest negative cases are brackets that open a side note after a sentence, like '(click anywhere to close)'. Steering produced looping or off-topic text, which does not help explain it.
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.
-
ter, Taslima and Simonovic, Slobodan P., "A General Overview of Multi-objective Multiple-participant Decision Making for Flood Management" (2002). Water Resources
-
of creating a “thermostatic mechanism” that would counter-balance the dominant presence of electronic media in the lives of children.⏎How to Cite⏎Copyright (c) 2017 Robert Albrecht⏎Copyrights and publishing
-
water, were the major factors shaping phytoplankton communities.⏎- Abstract views: 1777⏎- PDF: 842⏎Copyright (c) 2009 Silvia TAVERNINI, Daniele N
-
outside the team • The tasks the team perform have consequences that affect others inside or outside the organisation Figure 6.1 Characteristics of an effective team 13 (e) " .⏎Members of the team perform their roles and responsibilities with discipline,
-
a different version under "Related research" (further below) or search for a different version of it.⏎Volume (Year): 42 (2011)⏎Issue (Month):
-
2022 (during Earth Science Week). Geologic Map Day is a special event designed to promote
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 idea of a “two mind, two mind, two mind, two mind, two mind, two mind, two mind, two mind, two mind, two mind, two, two, two, two, two, two, two
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
. However, showing you is much cooler than explaining it.⏎Thanks for Subscribing!⏎You'll need to confirm your email, so please check your inbox.⏎(click anywhere to close)<|endoftext|>According to findings of a major HIV prevention trial presented Monday
-
solving of problems using Intel Visual Fortran compiler with IMSL Libraries. The use of other software packages (Maple, MathLab, Mathematica) or compilers (?, ?++, Visual Basic) for code is
-
,⏎- R is the electrical resistance,⏎- I is the electric current,⏎- V is the voltage.⏎(Note: this is not true where the impedance is complex)⏎|This page uses Creative
-
This helped the environmentalists a lot.⏎(Note: The background section of this page uses most of the same words as Terra Fut
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
- L10:66 strongly suppresses opening parenthesis after a citation or term
- L6:468 strongly excites
- L4:64 strongly excites
- L11:50 strongly excites
- L9:982 strongly suppresses
- L0:11 strongly excites
- L1:102 excites
- L7:100 suppresses
- L14:2454 excites
- L10:142 suppresses opening parenthesis mid-sentence
- L4:45 suppresses
- L15:608 suppresses
- L8:770 excites
- L1:155 suppresses
- L9:2249 suppresses
- L3:10 excites
- L0:1622 excites
- L11:1579 suppresses
- L3:57 excites
- L5:330 excites opening parenthesis after a noun
- L6:409 excites
- L14:90 suppresses
- L3:1388 excites
- L0:86 excites
Feeds later layers
- L20:335 strongly suppresses
- L21:187 strongly suppresses
- L17:238 strongly suppresses
- L18:682 strongly suppresses
- L19:1099 strongly suppresses
- L19:889 suppresses
- L18:910 suppresses
- L17:304 suppresses opening parenthesis after a claim
- L17:1509 suppresses
- L24:336 suppresses
- L20:104 excites
- L22:475 suppresses
- L18:1501 excites
- L19:453 suppresses
- L26:88 suppresses
- L25:415 suppresses
- L19:671 suppresses
- L22:453 suppresses
- L26:208 suppresses
- L21:1585 suppresses
- L22:246 suppresses
- L21:210 excites
- L18:1420 suppresses
- L24:125 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
- (
- (
- (
- raphics
- (
- 专业
- YPE
- 波
- AGO
- frei
Suppresses
- )<=
- .*?)
- )</
- +)\
- .*)
- []):
- {})
- sts
- 妲
- казать