word before 'type 2 diabetes' or a condition
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. 7 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them 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 just before a health condition or a category name, such as 'type' in 'type 2 diabetes' or 'family' in a taxonomy sentence. Turning it up made the model's writing drift into unrelated topics like history or philosophy, so the link to the steering text is weak.
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.
-
other colonial anthozoans and soft coral by their characteristic of incorporating sand and other small pieces of material into their tissue to help make their structure (except for the family Zoanthidae). A main characteristic of the order is that their tentacles
-
of Metabolic Syndrome X? Over time, normally years, the metabolic imbalances give rise to a greater risk of cardiovascular disease, heart disease, kidney failure, type 11 diabetes and so on. In fact, the cardiovascular coronary risk from
-
interested to know what to expect in the way of co-morbidities, including hypertension, impaired glucose tolerance (which leads to metabolic syndrome), heart disease, type 2 diabetes, and fatty liver disease, as well as less threatening conditions that
-
skin.⏎People, particularly men, with a specific form of the gene are both more likely to have lower percent body fat, but also to develop heart disease and type 2 diabetes. The gene seems to be associated having higher blood glucose and cholesterol
-
if learning with others can help enhance test scores.⏎There are numerous more contentious topics that might be used for great research paper issues. In case you’ve got a particular opinion about something, or about a problem that is of interest to you, it
-
move your body enough to get your heart rate up, you can start to have benefits. The mood enhancing effects of exercise can actually be used as a treatment for mild depression. And in addition, women who exercise more are also at a lower risk
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: a thing that can be measured and has a defined number of properties. Examples: A fruit: a whole, has a volume and a mass. Examples: A non-terminating decimal. Examples: A non-integer. Examples: A
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
, Sea Level, Renewable Energy, Natural Resources, Climate, Weather, Non-Renewable Energy, Land, Conflict, Fire, Atmosphere, Food, Life &
-
are many reasons to switch to solar mobile power sources. The first reason is that solar energy is a renewable resource. This means that it will never run out and we can rely on it
-
. This information is vital for decision-making, financial analysis, and evaluating the company's ability to meet its financial obligations.⏎Assets represent what the company owns and include tangible items like cash, inventory, and property, as well as int
-
. The less people know about us the better.⏎2. The fewer people know about us the better.⏎In #1 "less" applies to an uncountable abstract, the implied knowledge.⏎In #2 "fewer" applies to the
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
- L4:904 weakly excites
- L3:253 weakly excites
- L5:461 weakly excites
- L2:1133 weakly excites
- L5:771 weakly excites
- L1:2241 weakly excites
- L3:2731 weakly excites
- L2:473 weakly excites
- L3:1590 weakly suppresses
- L6:824 weakly excites
- L5:94 weakly suppresses
- L3:859 weakly excites
- L1:3070 weakly suppresses
- L1:2035 weakly suppresses
- L4:2409 weakly suppresses
- L6:263 weakly excites
- L1:1120 weakly suppresses 'for' after preparation or purpose nouns
- L5:8 weakly excites
- L0:2786 weakly suppresses
- L6:2109 weakly excites
- L6:7 weakly suppresses
- L6:1011 weakly excites
- L6:1135 weakly excites
- L4:379 weakly suppresses
Feeds later layers
- L9:1666 suppresses
- L9:2138 suppresses
- L13:2146 suppresses
- L8:391 suppresses
- L9:2714 suppresses
- L8:1135 suppresses
- L9:710 suppresses
- L9:1021 suppresses
- L9:663 suppresses
- L13:279 excites
- L8:946 suppresses
- L11:1385 suppresses
- L12:1273 suppresses
- L12:834 suppresses
- L15:260 weakly excites
- L10:1335 weakly suppresses
- L10:2456 weakly suppresses
- L16:2793 weakly suppresses
- L10:23 weakly suppresses
- L17:553 weakly suppresses
- L16:1819 weakly suppresses
- L9:2067 weakly suppresses
- L24:2777 weakly suppresses
- L20:1544 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
- BackingField
- plier
- spec
- intColor
- espec
- ting
- ыта
- 缓冲
- Facade
- RIA
Suppresses
- /non
- inear
- inite
- rous
- .non
- ечно
- 族自治州
- wrist
- OMET
- usive