energy and resource nouns
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. 11 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 fires on words for things we spend or use up, like electricity, energy, money and hours. Turning it up pushed the model toward unrelated everyday and problem-style text, so the steering does not clearly show this theme.
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.
-
00 tons of CO2 per year. At least 70-80% of cooking time is saved and therefore the use of valuable resources such as electricity, gas, wood, paraffin, money and time is drastically reduced.⏎
-
RO), which pushes brine water through a membrane that retains the salt and other impurities.⏎Thermal desalination uses heat as well as electricity in distillation processes with saline feedwater heated to vaporise, so fresh water
-
Greenhouse Gas Emissions⏎The impact of animal agriculture on greenhouse gas emissions goes a lot further than just cows producing methane gas. Meat production requires vast amounts of energy. Not only do you have to grow the crops to feed the animals, but
-
Pioneer Institute published an in-depth financial analysis of Common Core and determined that it would cost the state of Montana about $40 million to implement. Where will that money come from, and why wasn’t the legislature the body to decide if this was
-
olshoi” is Russian for “big” or “grand,” and the name is apt: This is a large-scale computational project, consuming six million CPU hours and producing a hundred terabytes of data. And yet, when you ponder the
-
of the very few slaves who managed to gain literacy, which made him became the “monopoly source of slave narrative”.⏎Don’t waste your time!⏎Order your assignment!⏎That Douglass possessed this monopoly astonished the public and raised his credibility quickly
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 very clean, professional, and well-structured document, with all the notes and explanations. The text is in English, with 100 words, and the notes are in a standard format. The main topics are covered in a
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Get a clear picture of the problem (and the solutions) from ?The Coming Global Superstorm,? now only $9.95 for a hardcover signed by Whitley Strieber, click hereand scroll down.⏎NOTE: This
-
their athletic skills. They work cooperatively playing soccer, badmitton, frisbee, bean bag bowling and other games.⏎17 Views 20 Downloads<|endoftext|>In Key Stage 1 the children will:⏎The five enquiry types⏎The
-
beautiful. The rarer and more bizarre anomalies and mutations are described in this fascinating book which opens a window on this developing and complex field of inquiry.⏎Profusely
-
18 x 638 mm; watermark: Atlas holding the celestial spheres; slightly frayed and darkened edges, else in fine condition.⏎Incl. BTW € 151,25⏎Excl.
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
- L4:2007 excites
- L10:884 suppresses
- L2:1899 weakly excites
- L4:690 weakly excites
- L1:2820 weakly suppresses
- L7:1436 weakly excites
- L6:808 weakly excites
- L2:352 weakly excites
- L10:1667 weakly suppresses
- L9:2351 weakly suppresses
- L4:929 weakly suppresses
- L3:3056 weakly excites
- L5:2245 weakly excites
- L7:817 weakly excites
- L10:831 weakly suppresses
- L8:1542 weakly suppresses
- L8:2663 weakly suppresses
- L1:2991 weakly suppresses
- L8:1218 weakly excites
- L9:1324 weakly suppresses
- L5:367 weakly suppresses
- L6:1283 weakly excites
- L10:391 weakly suppresses
- L7:1059 weakly excites
Feeds later layers
- L14:2981 weakly suppresses
- L15:1721 weakly suppresses
- L13:2900 weakly suppresses
- L13:2586 weakly excites
- L13:2329 weakly suppresses
- L17:2038 weakly suppresses
- L12:381 weakly excites
- L16:2967 weakly suppresses
- L19:2481 weakly suppresses
- L14:2084 weakly excites
- L22:2564 weakly excites perpendicular and geometry terms
- L15:761 weakly excites
- L17:1078 weakly suppresses
- L18:2353 weakly suppresses
- L16:1947 weakly suppresses
- L24:1503 weakly suppresses
- L25:2651 weakly excites the word 'Mount' in place names
- L17:1899 weakly suppresses
- L16:1605 weakly excites
- L21:2458 weakly suppresses
- L12:2247 weakly suppresses
- L24:1124 weakly excites
- L21:2604 weakly suppresses
- L14:2576 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
- onto
- 第一条
- toggleClass
- ighter
- Concurrency
- .hashCode
- erg
- .onPause
- 停滞
- 伫
Suppresses
- assel
- ackage
- entially
- 完整
- ован
- matlab
- initely
- ibly
- andal
- offer