hyphen joining two word parts
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. 26 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 lights up on a hyphen that glues two pieces of a word together, such as 'Bed-Ins' or 'co-creators'. Turning it up made the model's text drift into odd, repetitive loops, so the steering results say little about its role.
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.
-
o Arama'ah: Tabulae in grammaticen linguae Chaldaeae (Paris, 1560); Aseret ha-Devarim: Decalogus, with the commentary of Abraham Ibn Ezra,
-
oko Ono's youth to the violent death of Lennon's mother during his adolescence, from Lennon's post-Ed Sullivan skepticism to John and Yoko's Bed-Ins of 1969 - and unearths the secrets of its
-
a curb would be difficult, if not impossible to do without the ability to maintain single limb support for a given amount of time. The ability to switch from two- to one-leg standing is required to perform turns, climb stairs and dress.⏎As
-
. In case you need a good example here is one from Edutopia.http://www.edutopia.org/blog/flip-stem-classroom-ainissa-ramirez and all flipped classrooms do not look alike.⏎To see how technology transforms
-
of women and men to sustain the world. But it explicitly assigns women the role as sustainers of the world, again with the implicit understanding that men are co-creators and co-sustainers with her. The central source of the ethical
-
MIT News Office. 1992-04-15. http://web.mit.edu/newsoffice/1992/citation-0415.html. Retrieved 200<|endoftext|>Our herd of awesome
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: obviously, obviously, obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously obviously
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
The US federal tax on gasoline has remained at 18.4 cents per gallon during this period of rapid growth in gasoline prices, while state
-
the Belgrade Charter and the Tbilisi Declaration can be found at gdrc.org/uem/ee/belgrade.html and gdrc.org/uem/ee/tbilisi.html.⏎KAEE is one of the country’s oldest associations
-
undefined constant user_level - assumed 'user_level' (this will throw an Error in a future version of PHP) in /home/commons/public_html/wp-content/plugins/ultimate-google-analytics/ultimate_ga.php on line 524⏎GU
-
classroom may be a flipped classroom. In case you need a good example here is one from Edutopia.http://www.edutopia.org/blog/flip-stem-classroom-ainissa-ramirez and all flipped classrooms do
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
- L20:344 strongly excites hyphen inside compound words
- L20:1886 strongly suppresses
- L19:804 excites
- L16:19 suppresses
- L23:474 excites
- L20:961 suppresses
- L19:171 suppresses
- L1:217 suppresses
- L4:490 suppresses decade digit in years like 1930s
- L22:2384 suppresses
- L0:1267 suppresses
- L16:471 suppresses blank spot before a year
- L19:1016 excites
- L25:422 excites
- L16:0 suppresses
- L24:248 excites
- L3:101 suppresses
- L24:318 suppresses
- L4:46 suppresses
- L2:964 suppresses
- L19:89 excites
- L24:1223 suppresses
- L2:422 excites
- L19:216 excites
Feeds later layers
- L27:1749 strongly excites hyphen joining word parts
- L27:261 strongly suppresses
- L27:183 suppresses
- L27:458 suppresses
- L27:2439 suppresses slash inside web addresses
- L27:184 suppresses
- L27:657 weakly suppresses
- L27:334 weakly suppresses
- L27:2657 weakly suppresses
- L27:1450 weakly suppresses
- L27:2224 weakly suppresses
- L27:1196 weakly suppresses
- L27:139 weakly suppresses
- L27:782 weakly excites
- L27:981 weakly suppresses
- L27:817 weakly suppresses
- L27:327 weakly suppresses
- L27:2373 weakly suppresses
- L27:283 weakly suppresses the spot right before a number
- L27:1147 weakly suppresses
- L27:627 weakly suppresses
- L27:740 weakly suppresses
- L27:913 weakly suppresses
- L27:858 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
- :
- NULL
- odd
- ��
- .Menu
- occ
- ulan
- 調查
- ombs
- �
Suppresses
- diff
- dam
- me
- sm
- cr
- smart
- writing
- dom
- done
- re