'of' in US agency names
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. 10 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 reacts to the word 'of' inside the names of American health agencies, such as 'National Institute of' or 'National Center for'. Turning it up did not bring out that topic in the model's writing.
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.
-
approach to ?xml:namespace>⏎Producers based in Europe, the⏎Producers have put forward their
-
Rare Disorders⏎Canadian AIDS Society⏎CORD—Canadian Organization for Rare Disorders⏎NINDS progressive multifocal leukoencephalopathy information page. National Institute of Neurological Disorders and Stroke website. Available at http://www.ninds.nih.gov
-
by National Institutes of Health-funded Microbicide Trials Network , a worldwide collaborative clinical trials network focused on preventing the sexual transmission of HIV.⏎The NIH’s National Institute of Allergy and Infectious Diseases issued this statement about the
-
However, there is still no consensus on what exactly happens when Sars-CoV-2 invades the brain. Walter J., director of the US National Institute of Neurological Disorders and Stroke in Bethesda, Maryland. Korochetz said in a
-
health is awesome (and I’m not just saying that). Thanks to the achievements of public health, we live longer and healthier lives. According to the National Center for Health Statistics, “life expectancy at birth among U.S. residents increased by
-
ickle Cell Disease Association of America⏎This group provides education, advocacy, and other initiatives to promote awareness of and support for sickle cell disease programs.⏎National Institute of Neurological Disorders and Stroke (NINDS)⏎NINDS offers research information
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 have the following two expressions: 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
sun-exposed crowns, such as leaf water potential and photosynthesis.⏎Both tree species showed significant drought effects<|endoftext|>Introduction to Kubernetes⏎Kubernetes, also known as K8s or Kube, is an open-source system that helps organizations manage
-
power N is cxN/(1-x) ≈ 40 ⋅ 0.97N.⏎Let us look at A0073
-
zipping the .tgz files using the terminal. Let’s get into these.⏎Note: For this particular tutorial, we will use the Ubuntu 22.04. However, the commands used here are same for all the distributions.⏎
-
, which looms on the horizon north of the major city of Yogyakarta, is on the island of Java.⏎President Susilo Bambang Yudhoyono has called the volcanic eruptions a crisis situation. He and several
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
- L8:111 suppresses the space before a year in dates
- L8:25 weakly suppresses
- L7:26 weakly suppresses
- L8:1527 weakly suppresses
- L0:2893 weakly suppresses
- L9:13 weakly excites
- L8:2430 weakly suppresses
- L0:1510 weakly excites
- L8:749 weakly suppresses
- L8:447 weakly suppresses
- L8:861 weakly suppresses
- L9:746 weakly suppresses
- L8:1876 weakly excites
- L9:330 weakly suppresses gap where a number was removed
- L8:149 weakly suppresses
- L8:282 weakly suppresses
- L8:2006 weakly excites
- L6:4 weakly suppresses
- L7:8 weakly excites
- L3:492 weakly excites
- L6:428 weakly excites
- L9:401 weakly excites
- L8:1451 weakly suppresses
- L1:1474 weakly excites
Feeds later layers
- L17:876 excites
- L11:222 weakly suppresses
- L24:1486 weakly excites
- L16:1262 weakly suppresses
- L11:2138 weakly excites
- L13:52 weakly suppresses
- L18:2695 weakly excites
- L13:787 weakly suppresses
- L17:880 weakly suppresses
- L26:2872 weakly excites
- L21:2768 weakly excites
- L14:135 weakly suppresses
- L13:3027 weakly excites
- L13:1406 weakly excites
- L16:263 weakly suppresses
- L16:907 weakly suppresses
- L17:599 weakly suppresses
- L14:39 weakly suppresses
- L24:776 weakly suppresses
- L18:2869 weakly excites
- L25:598 weakly excites
- L14:2286 weakly suppresses
- L24:1328 weakly excites
- L23:1040 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
- =").
- >".
- ***/⏎
- )").
- .getTransaction
- ".⏎
- .").
- ,",
- ")).
- ,['
Suppresses
- oldt
- ски
- ework
- ull
- ild
- ştir
- ้าง
- asset
- illis
- assin