unclear: energy, sunlight and rate 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. 9 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. 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 responds to words about energy, sunshine and amounts of time or rate, such as 'sunlight', 'hours of sunshine' and 'energy consumption'. It is strongly negative on words like 'time' and 'at' in rate problems. The link is loose, and turning it up did not give a clear theme.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
the use of enzyme replacement therapy—however, this happened at an earlier age in male patients.⏎Having a
-
view PDF:⏎Energy community Nos Bambins explained⏎Consumers in an energy community buy energy at a lower rate from one or more producers through collective self-consumption. For this purpose, producers and consumers must be jointly represented by a legal
-
winter/early spring, for flowers.⏎Origin: Native to South Africa.⏎Cultivation: Grow Clivia miniata in bright light, but not in direct sun. Use well-draining soil and keep it on the dry side, watering the
-
To give you an idea of the difference between the weather in the UK and southern Europe. In the UK we have on average about 1000+ hours of sunshine per year, where as in southern Europe it's more like 2
-
aide.⏎*U.S. Department of Health and Human Services, December 1, 2015.⏎3. Doesn't Medicare pay for long-term care?⏎Many people mistakenly believe that Medicare
-
for particular product categories. ENERGY STAR has expanded to cover new homes, most of the buildings sector, residential heating and cooling equipment, major appliances, office equipment, lighting, consumer electronics, and more product areas.⏎If all consumers, businesses, and
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 get 100000000000000000000000000000000000000000000
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
?<|endoftext|>A particle is moving along the parabola 4y = (x + 2)^2 in such a way that its x-coordinate is increasing at a constant rate of 2 units per second. How fast is the particle's
-
energy can be of different forms: kinetic, magnetic, electric … etc⏎In any domain, the power is therefore expressed as a quantity of energy per unit of time. The international unit for the power is in Joules/second (J/s
-
take Bond A, issued by Company Alpha, as an example. It has the following data:⏎- Face value: $1,000⏎- Annual coupon rate: 5%⏎- Coupon Frequency: Annual⏎- Years to maturity:
-
.⏎How to calculate the bond price? The bond price formula⏎Bond price is calculated as the present value of the cash flow generated by the bond, namely the coupon payment throughout the life of the bond and the principal payment, or the balloon payment
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
- L7:2819 suppresses
- L7:1649 excites
- L8:874 excites
- L7:500 suppresses
- L6:1137 suppresses
- L8:1835 suppresses
- L7:2156 suppresses
- L8:3 excites
- L8:377 suppresses
- L7:1549 suppresses
- L6:2465 weakly excites
- L7:951 weakly excites "in" right before a year
- L6:2580 weakly suppresses
- L8:1666 weakly excites
- L6:2461 weakly suppresses
- L7:869 weakly excites
- L8:2910 weakly suppresses
- L8:1880 weakly excites
- L7:1610 weakly suppresses
- L8:12 weakly excites
- L8:2294 weakly suppresses
- L8:2471 weakly excites
- L7:1332 weakly suppresses
- L5:1267 weakly suppresses
Feeds later layers
- L20:1199 suppresses
- L11:1679 suppresses
- L10:733 suppresses
- L19:1348 suppresses
- L10:986 suppresses
- L10:915 excites
- L22:2171 suppresses
- L17:1228 excites
- L20:1387 suppresses
- L11:834 excites
- L12:2224 excites
- L22:1063 suppresses
- L10:1321 suppresses
- L17:2346 suppresses
- L16:2226 suppresses
- L10:637 suppresses
- L10:1707 excites
- L14:700 suppresses
- L11:292 suppresses
- L12:2782 suppresses
- L10:1098 suppresses
- L10:1350 suppresses
- L10:1925 excites
- L17:1416 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
- .IsSuccess
- ModelError
- edef
- JSGlobal
- 钤
- UObject
- eties
- 自卑
- 屐
- InitializeComponent
Suppresses
- continuously
- stream
- gesch
- streams
- steadily
- running
- 潺
- continual
- continually
- streams