power and energy unit symbols
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. 25 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 lights up on the part of a measurement that comes right before a unit, like the 'k' in kilowatts, the 'M' in MWh, or the 'W' in watts. When it was turned up, the model's replies drifted toward electricity and power talk, with a stray 'W' appearing, though the effect was otherwise messy.
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.
-
mann formula:⏎E = εσT4⏎where ε=emissivity, a material property; σ = 5.67×10-8 ; T = temperature in Kelvin (absolute
-
therefore expressed as a quantity of energy per unit of time. The international unit for the power is in Joules/second (J/s) which is also known as Watts (W).⏎In electricity, the power is determined by the product of the
-
NNL's solid oxide fuel cell development program. "However, this research shows that smaller solid oxide fuel cells that generate between 1 and 100 kilowatts of power are a viable
-
California — home to about half of the EVs in the United States. They find that the arrival of an EV increases household electricity consumption by 2.9 kilowatt hours per day — less than half the amount assumed by state regulators.⏎
-
s, flush residues after fossil fuels are burned and cool power plants.⏎Coal uses up to 3.2 cubic metres of water per megawatt hour (MWh) of electricity produced, gas uses up to 1.7 m3
-
out a stage in the process and making the whole system more efficient because the energy is converted only once.⏎The group estimates that the concept could reduce the cost per MWh of offshore wind energy by a factor of four, avoiding added system costs associated
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 big picture, the big picture, the big picture, the big picture... This is the big piece of the puzzle, right? The big piece of the puzzle, right? The big piece of the puzzle, right? The big piece
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
ney. He wakes up the next day, only to discover that it's not the next day at all. It's Groundhog Day all over again.⏎via Wikimedia Commons⏎For some reason he's trapped in Feb. 2, forced to
-
remains of the dismantled schooner Wawona.⏎Built by influential shipbuilder Hans Ditlev Bendixsen, the 165 ft long schooner Wawona sailed from 1897 to 1947
-
connection. Here are the steps to create an ad hoc network in Windows 7.⏎- Click on Start (Windows icon) and type wireless. Click on Manage wireless networks.⏎- Click on Add to add a network.⏎- Click on Create an
-
emissions from aircraft engines. This means it may be worthwhile for planes to fly at lower altitudes, even if they do release more greenhouse emissions that way. Robert Noland and his team calculate that if planes flew low enough to
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
- L24:1056 suppresses
- L23:895 excites
- L24:1520 weakly excites
- L8:3013 weakly excites
- L21:1978 weakly excites
- L13:1559 weakly suppresses
- L24:2592 weakly suppresses
- L20:2885 weakly suppresses
- L8:2206 weakly suppresses
- L4:1339 weakly suppresses
- L17:2350 weakly excites
- L22:1184 weakly suppresses
- L11:2779 weakly suppresses
- L7:2077 weakly suppresses
- L23:175 weakly suppresses
- L4:1652 weakly excites
- L20:31 weakly excites
- L0:1725 weakly excites
- L1:449 weakly excites
- L14:2744 weakly excites
- L23:1861 weakly excites
- L22:1701 weakly excites
- L13:2162 weakly excites
- L20:1392 weakly excites
Feeds later layers
- L27:848 strongly excites
- L26:556 strongly excites
- L27:457 strongly excites
- L27:670 excites
- L27:621 weakly excites
- L27:1625 weakly excites
- L26:973 weakly excites
- L26:1803 weakly suppresses
- L27:1121 weakly excites
- L27:2129 weakly suppresses
- L27:2707 weakly excites
- L26:3040 weakly excites
- L27:1013 weakly suppresses
- L27:2491 weakly suppresses
- L26:2677 weakly excites
- L27:1994 weakly suppresses
- L26:2967 weakly excites
- L27:2312 weakly suppresses
- L26:1167 weakly excites
- L26:1721 weakly suppresses
- L26:419 weakly excites
- L26:1960 weakly suppresses
- L26:2908 weakly suppresses
- L27:1593 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
- W
- W
- w
- w
- .w
- -w
- _w
- .W
- (W
- W
Suppresses
- stin
- essor
- utherford
- libs
- teenth
- отз
- 琯
- олж
- 第三个
- axe