"up" in "sign up"
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 "up" in the phrase "sign up", as in signing up for a newsletter or an account. Turning it up made the model write about registering for something new, which fits the same 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.
-
been viewed as a resource to be conquered.⏎Elizabeth Flock Elizabeth Flock. Become a member of the book club by joining our Facebook group , or by signing up to our newsletter. Learn more about the book club here. Below are questions to
-
DRM protected ebook on your desktop or laptop you will need to have Adobe Digital Editions installed. It is a free software. We also strongly recommend that you sign up for an AdobeID at the Adobe website. For more details please see FAQ
-
DRM protected ebook on your desktop or laptop you will need to have Adobe Digital Editions installed. It is a free software. We also strongly recommend that you sign up for an AdobeID at the Adobe website. For more details please see FAQ
-
ledge of Allegiance, the Ten Commandments, the Boy Scouts, the institution of Marriage — and how these wrong decisions can be avoided and remedied. Sign up now while there is still space in this special course!⏎Lessons cover issues such
-
DRM protected ebook on your desktop or laptop you will need to have Adobe Digital Editions installed. It is a free software. We also strongly recommend that you sign up for an AdobeID at
-
views that are particularly detrimental for women.⏎But nothing — not even a conservative culture or fear of the Taliban<|endoftext|>This preview shows pages 1–2. Sign up to view the full content.⏎This preview has intentionally blurred sections. Sign up to
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 am really not sure I am. I am so sure I am. Which one is the most accurate and accurate? I need to know. So, I think. So, I think. So, I think
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
the following decades, becoming the legal foundation for the reception in Germany of more than 4.5 million so-called “resettlers” (Aussiedler and Spätaussiedler) from Eastern Europe. In this contribution,
-
ic missions among the Guaranis - Relação abbreviada da republica, que os religiosos jesuitas estabelecerão no Paraguai, by Marquis of Pombal, Candide, by Voltaire
-
ain some of your losses. Don’t make applications too early, since volatilization losses could be high without rain to ensure incorporation of N.⏎Be wary
-
he not be buried until two days following his death.⏎To assuage this fear of being buried alive the safety coffin was invented. It is fitted with
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
- L0:917 weakly suppresses
- L1:1370 weakly excites
- L2:2451 weakly excites
- L2:292 weakly excites
- L1:2052 weakly suppresses
- L2:3027 weakly excites
- L2:2392 weakly suppresses
- L2:2772 weakly excites
- L0:1336 weakly suppresses
- L0:1948 weakly excites
- L1:1354 weakly excites
- L0:2171 weakly excites
- L2:2529 weakly excites
- L2:2288 weakly suppresses
- L1:2109 weakly suppresses
- L2:2596 weakly suppresses
- L2:767 weakly excites
- L1:1081 weakly suppresses
- L2:1556 weakly suppresses
- L0:410 weakly excites
- L1:2288 weakly suppresses
- L1:2922 weakly excites
- L0:1 weakly suppresses
- L0:1507 weakly excites
Feeds later layers
- L16:2884 excites
- L5:1125 excites
- L25:3019 weakly suppresses
- L16:2496 weakly suppresses
- L18:2759 weakly suppresses
- L4:1754 weakly excites
- L6:2224 weakly suppresses
- L24:2337 weakly excites
- L4:2899 weakly suppresses
- L25:1654 weakly suppresses
- L11:922 weakly suppresses
- L25:1745 weakly suppresses
- L13:2668 weakly excites
- L5:1048 weakly suppresses
- L9:1209 weakly excites
- L17:2766 weakly suppresses
- L5:1585 weakly excites
- L10:987 weakly suppresses
- L5:2746 weakly suppresses
- L7:2356 weakly excites
- L15:1496 weakly excites
- L12:2214 weakly suppresses
- L16:419 weakly suppresses
- L16:2792 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
- 伏
- Using
- using
- 维
- ke
- ](
- ee
- ósito
- using
- oto
Suppresses
- arness
- arez
- rength
- happiest
- 钫
- 就够了
- indsight
- igne
- vetica
- Wohnung