unit conversions in parentheses
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. 13 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. 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 unit inside a bracketed conversion, such as the 'mi' in '0.42 sq mi' or the 'F' in a Fahrenheit reading. Turning it up did not make the text about measurements, so the link is only partly visible.
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.
-
and policymakers across the country who are pushing for effective and equitable climate policies in their states. The views, thoughts, and opinions expressed in this article do not necessarily reflect those of Climate XChange.⏎Interested in submitting an article to be posted on our
-
from the top of the Carlton Centre⏎|Municipality||City of Johannesburg|⏎|• Total||1.09 km2 (0.42 sq mi)|⏎|• Density||13,000/km2 (
-
to some of the more fascinating questions that really excite mathematicians. Mathematics is the study of abstraction. Mathematicians search for similar patterns in seemingly different settings and then strive to communicate those patterns precisely. Mathematics is a creative process, not a mechanical
-
, 1991|⏎|-||Total||64,589 km2 (124th)⏎24,938 sq mi⏎|-
-
’s power.⏎While exposure to any inappropriate conditions can affect the potency of refrigerated vaccines, a single exposure to freezing temperatures (0° C [32° F] or colder) will destroy some. For example, liquid vaccines that contain
-
they are violating the negative rights of the citizens; that is, rights to have something not done to you. For example, taking money in the form of taxes to pay for such programs as described above deprives individuals of property. Proponents of
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: what is the (x) for the formulae in the formulae. So, this is the case for the formulae. I need to solve for a formula in this case. So, for example, a formula. So, the
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
0, 2016⏎- Rogue Bras to Bogarts: April Fool’s Day in the Media - April 1, 2016⏎- Romantic Writing: The History of Valentine’s Cards - February 10,
-
of a risky asset is a random walk with drift, so its return is unpredictable."⏎4. Is the following statement true, false or uncertain? "Under full liquidity constraints the marginal propensity to consume out of current income
-
1 (2):73 – 84.⏎James Behuniak (2002). Mencius on Becoming Human. Dissertation, University of Hawaii at Manoa⏎Liang Tao & Andrew Lambert (2009
-
ocean's surface to nearly a mile down, WHOI built acoustic tracking systems that monitor the movements of whales. A key component is a transmitter secured on the backs of the research subjects. Relaying information to detection buoys, the device houses
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
- L11:1116 excites
- L12:1263 excites
- L9:3062 excites
- L10:882 excites
- L12:1546 suppresses
- L12:1544 suppresses
- L12:3032 excites
- L12:1327 suppresses
- L9:2181 excites
- L12:867 excites
- L6:373 suppresses
- L12:1369 suppresses
- L12:1170 suppresses
- L11:1536 suppresses
- L12:3069 excites
- L10:108 weakly excites
- L9:797 weakly suppresses
- L11:1205 weakly suppresses
- L11:1226 weakly suppresses
- L11:1322 weakly excites
- L11:2683 weakly suppresses
- L8:2329 weakly suppresses
- L12:1805 weakly suppresses
- L12:9 weakly suppresses
Feeds later layers
- L14:1503 strongly excites
- L17:1229 excites
- L16:2759 excites
- L15:598 suppresses
- L15:1540 excites
- L14:981 excites
- L14:2022 excites
- L14:2313 excites
- L17:326 suppresses
- L18:2572 excites
- L14:1664 excites
- L18:1879 excites
- L14:2290 excites
- L16:2917 excites
- L16:2677 excites
- L18:250 suppresses
- L16:1394 suppresses
- L14:1609 excites
- L17:2365 excites
- L15:2450 excites
- L16:2512 excites
- L14:378 excites
- L15:1482 excites
- L17:1204 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
- :⏎
- :⏎
- :⏎⏎
- :⏎⏎
- ]:⏎
- :{⏎
- :⏎
- 에는
- //:
- pois
Suppresses
- authority
- <Entry
- .datab
- nth
- odore
- 感激
- feb
- Authority
- Supplier
- ninth