'to' after adapt/adjust verbs
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 lights up on the word 'to' right after verbs like 'adapt', 'adjust' or 'adapted', as in 'adapt to climate change'. When it was turned up, the model's text drifted toward talk of change and adjustment, but the link to the prompts was loose.
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.
-
a Russian scientist, N.V. Lazarev, who was interested in substances that helped the body adapt to physical and emotional stress. Lazarev thought that adaptogens should:⏎- Produce
-
their daughter have spent most of 2008 living in primitive facilities near the shores of Lake Tanganyika.⏎There, the family has learned to adjust to life in
-
whose sweep around the sun at that time was a scientific bonus. Daily medical experiments took the most time, as the spacemen gathered data on how their bodies adapted to their weightless environment. The astronauts took four two- man space walks, with
-
people. The video highlights the need for accurate, continuous, and accessible data and computer models from NASA satellites to track and predict the challenges farmers face as they adjust to a changing climate.⏎This video is the first of a three-video series from the
-
to Solomon Islands⏎In Solomon Islands and across the Pacific Islands region, the U.S. Agency for International Development (USAID) supports programs that help communities adapt to the negative impacts of global climate change and supports disaster relief efforts and disaster risk reduction
-
morphogens. And even cancer cells can use morphogens to grow, induce a blood supply, and metastasize.⏎Although the concept of cell organization in response to these morphogen gradients is well documented, little is known about how these subtle concentration
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 can't have me and me, I can't have you and me. So I'm thinking, I'm thinking. Then, I can't think, I can't think. This is the problem. I mean, I'm thinking
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
electrodes or plates, an electrolyte that permits ionic conduction, one or more negative electrodes or plates, separators between plates of opposite polarity, and a container for all the above.⏎Battery Life — The period during which
-
Munich,⏎will also return this year.⏎Computer simulations suggest that⏎a Morning Glory is caused by the collision of two sea breezes over Cape⏎York the preceding evening. "We want to nail down the
-
charged electrons. While the positively charged ions pass through the fuel cell, the negatively charged electrons are diverted and directed through a wire to create an electric current. The electrons complete the detour to meet up with the ions in the final segment of the
-
others that need a bit more use of a microscope such as extinction (the angle at which a mineral goes black and you rotate the slide) and birefringence (the colour of the mineral using crossed polars).⏎Minerals have different
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
- L7:1782 weakly suppresses
- L7:2364 weakly suppresses
- L8:2797 weakly suppresses
- L6:1490 weakly suppresses
- L8:181 weakly suppresses
- L9:2506 weakly excites
- L7:2957 weakly excites
- L6:2173 weakly suppresses
- L5:1224 weakly excites
- L6:1158 weakly excites
- L6:2756 weakly suppresses
- L9:2757 weakly excites
- L8:706 weakly suppresses
- L3:429 weakly suppresses
- L9:2643 weakly suppresses
- L8:137 weakly excites
- L4:1431 weakly excites
- L7:564 weakly suppresses
- L6:615 weakly excites
- L4:1013 weakly suppresses
- L8:1469 weakly suppresses
- L7:3040 weakly excites
- L9:2404 weakly suppresses
- L1:2922 weakly excites
Feeds later layers
- L19:2102 excites
- L16:2556 excites
- L17:270 weakly suppresses
- L11:1350 weakly suppresses
- L22:755 weakly excites 'all' or 'any' in 'people of all ages'
- L27:1231 weakly excites
- L20:2028 weakly excites
- L14:640 weakly excites
- L17:2165 weakly excites
- L24:2320 weakly excites
- L12:2994 weakly excites
- L14:2333 weakly excites
- L14:556 weakly excites
- L14:694 weakly suppresses
- L16:2163 weakly excites
- L14:2658 weakly suppresses
- L23:977 weakly suppresses
- L15:1270 weakly excites
- L12:1137 weakly excites
- L12:1465 weakly excites
- L12:2410 weakly excites
- L21:1633 weakly suppresses
- L17:1195 weakly excites
- L14:539 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
- 处境
- 传播
- 阴
- 难度
- esp
- 生存
- 气候
- exposures
- 现实生活
- ég
Suppresses
- 最快的
- ittal
- ware
- mach
- uffed
- ITES
- 最多的
- responsable
- terminate
- .DataTable