"at least" and minimum-quantity phrases
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. 20 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 the words "at least" when they set a minimum amount, like "at least one" or "at least 70%". When it was turned up, the model kept producing "at least" and "least" over and over.
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.
-
of dog food, but all of them must meet certain requirements to be considered “real” dog food. For example, the food must have a meat content of at least 70% and it must not contain any artificial colors or flavors.
-
across disparate operating systems.⏎Kubernetes clusters are actually a set of nodes categorized into master nodes and worker nodes. Every cluster needs to have at least one of each. While the master node is meant to control the state of
-
its evolution into an all-engulfing process, why it is so difficult to stop once it has begun, and what our best hopes are for ending war or at least limiting its dangers.⏎Several important developments have taken place since Dr. Dyer
-
or disorder is the genotype with two recessive alleles. So I guess in order for the normal genotype, or what is considered normal, a person has to have at least
-
using it.⏎Shared libraries can also help you get around the 32K code size limit; you can break up your code into a main portion (of at most 32K), and a number of libraries (each of at most
-
1 cm diameter at 1.3 m from the rooting point) they hosted. Overall, 57.9% of trees in the study sites carried at least one liana. The infested trees hosted an average of 1.
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 two of the two of the two of the two of the two of the two of the two of the two of the two of the two of the two of the two of the two of the two of the two of the two of
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
ibility.⏎To reduce reverberation and noise levels, acoustical ceiling panels should have a noise reduction coefficient (NRC) of 0.70 or greater. (The NRC indicates the average percentage of sound a material absorbs;
-
that meet minimum population density requirements. These urbanized areas are determined by the U.S. Census Bureau as having a population of 50,000 or more people.⏎- 2010 Census Urban Area FAQs⏎- Missouri
-
Physical Security Measures in K–12 Schools⏎Download eBook for Free⏎|PDF file||0.6 MB||⏎Use Adobe Acrobat Reader version 10 or higher for the best experience.⏎- What challenges do LEAs face related to school
-
it manifests and the consequences for people and society.<|endoftext|>Fact 1: The world’s population is rapidly ageing⏎The number of people aged 60 years or older will rise from 900 million to 2 billion between 2
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
- L19:1171 strongly excites
- L16:2990 excites
- L19:1461 suppresses
- L18:1920 suppresses
- L12:2102 suppresses
- L18:2965 excites
- L17:2409 suppresses
- L18:504 suppresses
- L18:2197 suppresses
- L19:1211 excites
- L18:1760 suppresses
- L19:2623 excites
- L19:1222 suppresses
- L15:2308 weakly excites
- L15:1910 weakly excites
- L15:1212 weakly excites
- L19:2653 weakly excites "most" meaning "very" or "the most"
- L12:294 weakly excites
- L10:2816 weakly excites
- L19:2969 weakly excites
- L11:245 weakly suppresses
- L18:569 weakly excites
- L8:1710 weakly excites
- L7:846 weakly excites
Feeds later layers
- L23:1589 strongly excites 'at' in 'at least' and 'minimum'
- L23:1543 excites 'up' in 'up to' limits
- L27:599 excites
- L26:910 excites
- L24:380 excites the word 'at' in place phrases
- L26:2537 excites
- L26:835 excites
- L24:2610 excites
- L25:2301 excites
- L24:1826 excites
- L21:2199 excites
- L27:1465 excites
- L25:501 suppresses
- L22:615 excites
- L27:863 excites
- L23:380 excites
- L27:1086 excites
- L21:1256 excites
- L27:1060 weakly excites
- L25:2378 weakly excites
- L23:2966 weakly suppresses
- L27:778 weakly excites
- L22:969 weakly excites
- L24:1585 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
- 至少
- least
- 最少
- least
- 起码
- 少なく
- atleast
- minimum
- minimum
- Least
Suppresses
- ét
- 还有什么
- elé
- éric
- 配上
- manipulate
- 配
- hormones
- 真的
- 才发现