'to' after 'put' (put to use)
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. 7 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 fires most on the word 'to' in phrases like 'put to use', 'put to death', and 'put to a writer', and on similar verb plus 'to' pairs. It turns off on technical economics words like 'equilibrium', 'solutions', and 'variables'. Turning it up did not bring out a clear theme in the generated text.
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.
-
reconciled will be crucial.⏎Meanwhile, the Sanders-Boxer Global Warming Pollution Reduction Act, currently in committee in the Senate, would put smart-energy solutions to work right now to solve global warming and also create new economic opportunities and hundreds of
-
and with a few modifications the arc was struck.⏎It’s not a finished project yet, and we’re looking forward to seeing how [GreatScott!] puts this to use. For now, we’re grateful for the lesson is Royer oscillators
-
istic reforms and translated the Bible into his native tongue.⏎Many Poles turned from the Roman Church, winning a measure of religious freedom. Some church buildings were put to Protestant use. Laws forbade the execution of Roman Church decrees
-
59; died at Madrid, 27 March, 1625. He was a great-grandson of the Tordesillas who was put to death by the Comuneros at Seville. He studied in Spain and Italy
-
b) Write down the consumers budget constraint and characterize the solution to her intertemporal maximization problem.⏎c) Consider the full competitive market equilibrium in this economy; normalize
-
credibility. It would not be an exaggeration to say that from ancient times until now, there has not been a period in which this question has not been put to a writer or an artist without regard to his nationality or the motives for his departure
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: to find the number of ways. In this case, the problem is to find the number of ways the students can choose their preferences, which is the number of ways to choose the students, and to find the number of ways to choose the
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
match experimental results found in the literature. These results are for steady-state solutions; a dynamic implementation requires transient modeling as well. This can be achieved by extending
-
of supply and demand. Any shift in one of these results in a price change that signals a corresponding shift in the other. Then, the prices return to an equilibrium level. A market failure results when prices cannot achieve equilibrium because of market distortions
-
models are applied to problems with the following characteristics:⏎The problem has a well-defined sole objective.⏎There are alternative courses of action to resolve the problem.⏎The decision variables are continuous and they can accept any fractional or non-negative values within the particular range
-
in Congressional Primaries, 1994-2004⏎|Pluralities Avoided by Runoffs||77|⏎|Candidates Elected After Plurality Primary Victory|⏎|Candidates Elected After Runoff Victory|⏎
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
- L6:1236 excites
- L6:2209 excites
- L3:1299 suppresses
- L3:705 weakly suppresses
- L5:1297 weakly excites
- L6:400 weakly suppresses
- L4:954 weakly suppresses
- L6:1862 weakly excites
- L3:2922 weakly excites
- L5:2096 weakly excites
- L4:1186 weakly excites
- L6:2435 weakly excites
- L6:742 weakly suppresses
- L5:91 weakly excites
- L6:7 weakly suppresses
- L4:1379 weakly suppresses
- L4:2040 weakly suppresses
- L5:1334 weakly suppresses
- L6:1147 weakly excites
- L3:105 weakly excites
- L4:879 weakly excites
- L0:2081 weakly excites
- L0:2529 weakly excites
- L2:2607 weakly excites
Feeds later layers
- L10:1311 excites
- L8:141 excites
- L12:1989 suppresses
- L8:246 suppresses
- L12:650 weakly excites
- L12:1908 weakly excites
- L12:793 weakly suppresses
- L8:767 weakly suppresses
- L11:1634 weakly excites
- L15:2531 weakly excites
- L14:583 weakly excites
- L21:3055 weakly excites
- L25:489 weakly suppresses
- L9:637 weakly excites
- L11:1959 weakly excites
- L22:1336 weakly excites
- L18:658 weakly suppresses
- L9:1844 weakly excites
- L11:86 weakly excites
- L12:2764 weakly suppresses
- L9:1290 weakly suppresses
- L10:1513 weakly suppresses
- L23:1909 weakly suppresses
- L9:2233 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
- posable
- ailing
- ibe
- arching
- ometric
- iping
- abei
- reamble
- <UFunction
- определен
Suppresses
- solutions
- Nazis
- mium
- 结论
- solution
- Auschwitz
- 结果
- :NS
- azz
- mage