'iv' inside herbivores, energy fuel words
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. 8 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 piece 'iv' in the middle of the word 'herbivores', and on a few other odd word fragments. It is pulled the other way by text about burning fat or energy for fuel. Turning it up did not produce any clear change in topic.
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.
-
which biotic and abiotic drivers are redefining the distributional dynamics of native and non-native plants under global change.⏎Current study system. Plants and herbivores! Plants may seem like static
-
., Filippi, A., & Marini, L. (2021). Contrasting response of native and non‐native plants to disturbance and herbivory in mountain environments. Journal of Biogeography, 48(7
-
-scale reshuffling of species distributions creates novel interactions among species and across trophic levels that previously did not co-occur. In particular, interactions with herbivores emerged as a fundamental driver of plant community
-
human disturbance, while the same drivers promote non-native spread. Although biotic interactions should modulate plant invasions, it is still unclear how invertebrate herbivores can affect non-native spread. Most previous research focused on lab experiments, and
-
birds, and other herbivores that like to snack on fruits, leaves, and flowers. But what W
-
cheese filling. The turnover is grilled or fried, using either flour or a corn tortilla. Served with guacamole and sour cream, this simple unleavened flatbread turnover remains a favorite snack of the Mexican and Tex-Mex cuisine
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: 100 g of food is used to produce 200 kcal of energy. How many grams of carbohydrates are consumed? The given data: - 100 g of food is used to produce 20
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
When your cells can’t get glucose (a kind of sugar) to burn for energy, they burn fat instead. This leaves behind acids called ketones in your blood and urine. A
-
collision that took place 1.3 billion years ago. When it occurred, about three times the mass of the sun was converted to energy in a fraction of a second.⏎The discovery has been accepted for publication in Physical
-
Restricting calories causes muscles to rely more on glucose from the food you’re consuming for energy than from the stored fat. They also become more efficient.⏎Many of these hormones don’t return to normal
-
calorie burn:⏎- RMR=resting metabolic rate-the amount of calories you need to exist without any activity-ie:bedrest⏎- TEA-thermic effect of activity-this, obviously, represents the caloric burn of activity, both
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
- L6:1181 suppresses
- L6:1935 weakly excites
- L6:1832 weakly suppresses
- L6:2465 weakly excites
- L3:2958 weakly excites
- L7:1914 weakly suppresses
- L5:1042 weakly suppresses
- L6:2691 weakly excites
- L2:2460 weakly suppresses
- L2:493 weakly suppresses
- L5:342 weakly suppresses line breaks inside wrapped sentences
- L7:1832 weakly suppresses
- L7:1590 weakly suppresses
- L7:76 weakly excites
- L0:1301 weakly suppresses
- L3:651 weakly suppresses zero in dates inside URLs
- L0:1677 weakly suppresses
- L4:882 weakly excites 'with' after provide or familiarize
- L6:1678 weakly excites
- L7:2463 weakly excites
- L5:1917 weakly suppresses 'to' after 'what' or 'where' (what to do)
- L6:2456 weakly excites
- L6:1921 weakly excites
- L7:303 weakly suppresses
Feeds later layers
- L21:2922 suppresses
- L20:3023 weakly excites
- L18:2854 weakly suppresses
- L10:983 weakly suppresses
- L9:2709 weakly excites
- L25:722 weakly suppresses
- L21:980 weakly suppresses
- L18:969 weakly suppresses
- L19:2679 weakly excites
- L10:865 weakly suppresses
- L20:1898 weakly suppresses
- L10:628 weakly excites
- L11:2847 weakly excites
- L10:2187 weakly excites
- L11:2448 weakly excites
- L23:1859 weakly suppresses
- L9:2612 weakly suppresses
- L15:1445 weakly suppresses
- L25:1379 weakly excites
- L19:1926 weakly suppresses
- L21:2521 weakly suppresses
- L10:861 weakly suppresses
- L11:1382 weakly suppresses
- L21:2035 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
- 湿润
- anging
- tí
- ometers
- athroom
- WARDS
- typingsJapgolly
- filtro
- 妇联
- ocrats
Suppresses
- massa
- 燃料
- 重整
- igne
- 合成
- 原料
- 能量
- 化石
- energy
- 功