TV and video media 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. 23 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them 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 on words about television, channels and video, such as 'TV', 'Channel 4' or the name of a broadcaster. When turned down, the model started writing about video and motion. Turned up, it wrote about unrelated topics.
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.
-
"Friends" (NBC), "Just Shoot Me" (NBC), "Ally McBeal" (FOX), "That '70s Show" (FOX), "The Simpsons" (FOX), and "Dawson's Creek
-
75% of e-cigarette ad exposure to youth occurred on cable networks, including AMC, Country Music Television, Comedy Central, WGN America, TV Land, and VH1. Researchers found that e-cigarette ads appeared on programs
-
marked the beginning of mainstream 3D for film and TV audiences – in fact only recently there were a spate of classic 3D movies played on Channel 4. These ‘glasses’ (the lenses in the cardboard design were a
-
exactly along the lines of history as we know it. Despite the book being banned, it has become an underground bestseller.⏎first season of Amazon’s highly rated TV series of the same name, roughly follows the outlines of Dick’s book with a
-
ysicist who writes for NPR’s 13.7 blog, described dark matter as being invisible (which makes it
-
it or discovered the alloy, and then used it to trap Warehouse agents Bering and Lattimer against the ceiling.⏎This is a good example of how the TV Universe is always in a state of flux, constantly being re-written. Until the
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: if the speed of sound can be recorded, then what would happen to the recorded motion of the moving object? How would it behave? What would it look like? How would it go? How would it move? What would it sound like
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
player must support the codec used to encode the video data.⏎||MP4 files are an open standard that everyone can use.⏎|Programs that open AVI/MP4 files<|endoftext|>Pioneer Papers connects you with professional substantive editors to build and
-
model "VisMo", based on CIDOC CRM (University of Passau)<|endoftext|>||Audio Video Interleave File⏎||MPEG-4 Video File⏎||AVI is a video wrapper format created by Microsoft, storing video data that might
-
‘Rigid Body’ and leave the type as ‘Active’. We want this to simulate animation. Now change the bounciness to 1.⏎- Playback the animation (Alt+A) and observe your bouncing ball! (Not related, but
-
, 2008, edition of The Boston Globe features an in-depth profile of Bisbee and his art. Click here to read the article and watch a
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
- L20:1199 strongly suppresses
- L20:2621 excites
- L8:874 excites
- L14:700 excites
- L9:2779 excites
- L22:1292 excites
- L19:2326 suppresses
- L18:1766 suppresses
- L10:733 suppresses
- L4:2643 suppresses
- L8:2910 suppresses
- L6:2384 excites
- L22:2171 suppresses
- L7:1332 suppresses
- L7:2775 suppresses
- L9:2321 suppresses
- L17:1228 excites
- L22:2833 excites
- L21:1083 suppresses
- L9:617 suppresses
- L7:1822 suppresses
- L3:996 weakly suppresses
- L6:1914 weakly excites
- L20:2510 weakly excites
Feeds later layers
- L26:577 suppresses
- L26:2100 suppresses
- L26:2078 suppresses
- L24:2178 suppresses
- L27:2112 suppresses
- L24:2012 weakly suppresses
- L25:1683 weakly suppresses
- L26:269 weakly suppresses
- L24:2994 weakly suppresses
- L26:2547 weakly suppresses
- L26:966 weakly suppresses
- L25:1812 weakly suppresses
- L25:2044 weakly suppresses
- L27:2424 weakly suppresses
- L24:2722 weakly suppresses
- L26:2357 weakly suppresses
- L24:2325 weakly suppresses
- L27:308 weakly suppresses
- L25:1712 weakly suppresses
- L24:1657 weakly suppresses
- L24:2015 weakly suppresses
- L24:2732 weakly suppresses
- L24:2882 weakly suppresses
- L25:859 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
- oyal
- AGMENT
- Sherlock
- Gree
- bourg
- фор
- TextAlign
- ќ
- ☒
- ISC
Suppresses
- video
- video
- 视频
- Video
- videos
- Video
- (video
- .Video
- _video
- videos