end-of-text token after a finished passage
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. 18 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 invisible end-of-document marker that appears right after a passage stops, where the next text is unrelated. Pushing it up made the model drift into unrelated topics.
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.
-
manage operations in your warehouse.⏎- Lay out a proper strategy for handling damaged products and returns from customers.⏎- Implement pest control measures to protect your products.⏎-<|endoftext|>Nucleation properties of respirable coal dust (terminal progress report).⏎Physics Department
-
be the most effective approach. Smart work, on the other hand, can save time and increase productivity, but may lead to a lack of focus on the details<|endoftext|>Rajan Vempati of ChK Group, Inc. in Plano,
-
), Applying the ecosystem approach (Section 3), creating a management plan (Section 4) and assessing a project against the ecosystem approach (Section 5).<|endoftext|>We offer free Maharashtra
-
1, on the Mall in front of the National Air and Space Museum. The museum will also be hosting a scholarly conference on Civil War ballooning on June <|endoftext|>Most veteran peanut farmers in the Southeast will remember the devastating spread of tomato spotted wilt
-
students into groups of three.⏎Allow time for all three group members to be traced. Students may then complete⏎silhouettes at home or during free time.<|endoftext|>2010 is the International Year of Biodiversity. Why should you
-
be considered a planet, reigniting discussions.⏎In<|endoftext|>The Carnival season, or “Apokries” in Greek, is slowly coming
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 first of the years of the first of the years. I think that this is a good idea, if we can find the first of the first of the first. This is a very small step. is a small step. The first
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
920 by the Nineteenth Amendment to the federal Constitution, the...⏎Users without a subscription are not able to see the full⏎to access all content.<|endoftext|>Great Slaty becomes World's Largest Woodpecker as Ivory Billed
-
12|⏎|Journal||Classical and Quantum Gravity|⏎|State||Published - Jul 7 2003|⏎ASJC Scopus subject areas⏎- Physics and Astronomy (miscellaneous)<|endoftext|>Most computer users are familiar with the
-
silkworm-BEVS is a promising alternative strategy to produce the functional and thermostable reverse transcriptase.⏎All Science Journal Classification (ASJC) codes⏎- Insect Science<|endoftext|>An account of the religion of
-
show us God’s glory and sovereignty, and so others will believe on him.⏎Leave your comments below to share your thoughts on the subject. If you think others would appreciate reading this, please share it through the social media
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
- L17:658 strongly suppresses
- L16:235 strongly suppresses
- L17:3036 strongly excites end of document boundary
- L16:2524 strongly excites
- L17:1088 excites
- L15:1246 suppresses
- L15:2977 suppresses
- L15:2544 suppresses
- L15:2675 excites
- L15:101 excites
- L15:917 suppresses
- L15:2249 excites
- L12:2927 suppresses
- L15:1567 suppresses
- L15:774 excites
- L16:2772 excites
- L16:1411 excites
- L9:1039 excites
- L16:2510 excites
- L13:2897 excites
- L17:1523 suppresses
- L10:1298 excites
- L17:2723 excites
- L15:131 suppresses
Feeds later layers
- L19:2057 strongly suppresses
- L26:2128 strongly suppresses decimal points inside numbers
- L22:417 suppresses period closing a web blurb
- L25:808 suppresses
- L20:338 suppresses
- L24:736 suppresses
- L21:1150 excites
- L27:725 suppresses
- L19:1069 suppresses
- L21:301 excites
- L22:552 suppresses
- L20:309 suppresses
- L19:301 suppresses
- L19:1105 weakly excites
- L26:449 weakly excites
- L24:211 weakly excites
- L26:288 weakly excites
- L19:1527 weakly suppresses
- L20:460 weakly excites
- L19:2031 weakly suppresses line breaks in web footers and metadata
- L26:16 weakly suppresses
- L24:36 weakly excites
- L20:44 weakly excites
- L22:91 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
- ';⏎⏎
- ":⏎⏎
- _);⏎⏎
- _;⏎⏎
- $⏎⏎
- лиц
- /";⏎⏎
- ':⏎⏎
- uko
- ksen
Suppresses
- <|endoftext|>
- enu
- 返回搜狐
- سب
- firstname
- firstname
- </
- ']?>"
- token
- }`