'static' and 'static'-like 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 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 reacts most to the word "static", especially in its programming sense (a static inner class), and also to words that sound or look similar, like "eggs" or "passive". Its strongest negative cases are "static" meaning radio noise. Turning it up or down gave unrelated text, so the link to the trigger is only loose.
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.
-
behavior. Here's a class called CarParts that declares a static inner class called Wheel. If we consider the real-world, we can find many
-
5 and others.⏎> Let us start from some anecdotal evidence: isinstance() is one of the most called functions in large scale Python code-bases (by static call count). In particular, when analyzing some multi-million line production code base,
-
:37⏎Exercise7. Read the following sentences, underline the main verb and state the verb: Active or Passive.⏎1.Railroads were developed in the 19 century and became the
-
streaked with yellow. Male painted turtles have longer front toenails than the females, apparently to "tickle" the female's head during courtship and stimulate egg production.⏎Only a few sightings of the rubber boa have been made in the park
-
home floral arrangements. "New weed infestations can occur when people transport the pretty flowers," he says.⏎Yellow toadflax, sometimes called butter and eggs, was actually introduced as an ornamental plant. "It's a member of
-
participlePast perfect (had been) Will + be + past participle Can + be + past participle Will + be Can + be⏎Active voicePassive voice Helen cleans the house The house is cleaned by Helen Some people collect pictures Pictures
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 user, you, is the one who said, "2011". So, the first part of this is the... for this 2011 The you are the one who... so, the first
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
cars, the BMW i3 and Tesla Model X, are ditching terrestrial AM radio because electromagnetic noise from the electric motor interferes with the broadcast reception, causing static, as the blog Music 3.0 recently pointed out.⏎Electronics have
-
thousands of times a second, basically chopping up energy. This process causes electromagnetic interference that gets picked up by the radio.⏎AM radio has always been more susceptible to static than its partner on the dial. "AM" stands for amplitude modulation, which
-
21st, 2019⏎Overview of chronic pelvic pain⏎Pelvic pain is a serious form of non-communicable disease, which directly interferes with healthy living. Acute pelvic pain conditions can
-
less severe limitations in functioning.Fact 3: The most common health conditions in older age are noncommunicable diseases⏎Older people in low- and middle-income countries carry a greater disease
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
- L4:2345 excites
- L7:76 excites
- L7:122 weakly excites
- L6:572 weakly suppresses
- L6:1852 weakly suppresses
- L7:545 weakly excites
- L0:1301 weakly suppresses
- L6:379 weakly suppresses
- L6:8 weakly suppresses
- L4:1080 weakly suppresses
- L0:1224 weakly suppresses
- L6:16 weakly excites
- L4:339 weakly excites
- L2:2471 weakly suppresses
- L0:1 weakly suppresses
- L0:178 weakly suppresses
- L7:2187 weakly excites
- L1:1864 weakly suppresses
- L0:2768 weakly suppresses
- L6:2888 weakly suppresses
- L4:1 weakly excites
- L3:1339 weakly excites
- L3:2019 weakly excites
- L6:1961 weakly suppresses
Feeds later layers
- L10:1686 excites
- L9:3054 excites
- L12:2634 excites
- L23:2616 weakly suppresses
- L14:1421 weakly suppresses
- L10:231 weakly excites
- L25:2652 weakly suppresses
- L9:2987 weakly suppresses
- L14:1668 weakly excites
- L14:970 weakly suppresses
- L16:972 weakly suppresses
- L9:2942 weakly suppresses
- L11:1589 weakly suppresses
- L18:2025 weakly excites
- L9:692 weakly excites
- L11:3016 weakly excites slash in fractions
- L24:2137 weakly suppresses
- L11:2442 weakly excites
- L19:2822 weakly suppresses
- L15:2376 weakly excites
- L14:967 weakly excites
- L14:2144 weakly excites
- L11:2009 weakly suppresses
- L19:3064 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
- ire
- IR
- ayscale
- polled
- IQ
- ir
- Toll
- ULT
- 在全球
- èle
Suppresses
- expense
- 工程机械
- 你以为
- 达不到
- 玑
- 需要用到
- .StoredProcedure
- buster
- 需要用
- 罟