unclear: scattered word fragments
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. 5 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 weakly and on many unrelated bits of text, such as parts of acronyms (EBUS, XLOOKUP) and odd word endings. No clear theme shows up, and turning it up or down did not produce an obvious change in meaning.
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.
-
a hurry?⏎Most common for 2D lookups are INDEX/MATCH/MATCH (see method number 4 below) or a 2D XLOOKUP (click here for more information).⏎What is a 2D lookup?⏎In
-
patient care. Advances in endoscopic techniques and in imaging help achieve these goals. Among endoscopy techniques are:⏎- Endobronchial ultrasound (EBUS)⏎- Electromagnetic navigation (EMN) with volumetric high-resolution computed
-
only partial access to thoracic lymph node stations.⏎By contrast, the convex endobronchial ultrasound (EBUS) bronchoscope provides real-time guidance for transbronchial needle aspiration (
-
Work II: Triangulation Qualitative Research: State of the Art and the Future Glossary References Author index Subject index About the Author.⏎(source: Nielsen Book Data)⏎'The fourth edition of Uwe Flick's "Introduction to Qualitative
-
materials such as MyPlate. A "portion" is the amount of a food you choose to eat at
-
unda tamamlamıştır.⏎Yüksek Lisans Öğrenimine MSÜ Barbaros DEBİM Enstitüsünde devam etmekte aynı zamanda enerji sistem ve modelleme üzerine araştırmalar
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 problem. The problem is the time. The problem is ... Time What Problem: Time Time A: A A B C D E F G H I
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
in place, regardless of its first letter).⏎When you have everything set the way you want, click the OK button. You will be returned to the document and you should see that the list is now in order.⏎Quick Note: When creating a
-
were shipped out to California by train.⏎Earlier, in Houston, he had become fascinated with steam cars. He enjoyed racing the other blades on Houston streets, and he discovered that the steamer could best any other automobile in a drag race; the
-
against foreign invasion in three great wars, thus ensuring tranquillity during much of his reign. In his final years, however, he faced internal disturbances, and he was ultimately killed in an attempted coup d’état.⏎Son of Setnakht
-
him to success for example, he would celebrate the date he received his first ever job (September 26, 1855) every year and he considered the date more important than his birthday. “All my future seemed to hinge
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
- L4:2640 suppresses
- L4:131 suppresses
- L4:1831 weakly excites
- L3:3057 weakly excites
- L4:2081 weakly suppresses
- L4:2239 weakly suppresses
- L0:1437 weakly suppresses
- L1:2683 weakly excites
- L3:1415 weakly suppresses
- L0:2173 weakly excites
- L2:483 weakly suppresses
- L0:979 weakly excites
- L3:876 weakly suppresses
- L2:2404 weakly suppresses
- L1:1733 weakly suppresses
- L4:2355 weakly excites
- L0:2368 weakly excites
- L2:1056 weakly suppresses the word "thought" before "to"
- L3:881 weakly excites
- L2:399 weakly suppresses
- L2:2207 weakly excites
- L4:2102 weakly excites
- L0:963 weakly suppresses
- L2:70 weakly excites
Feeds later layers
- L9:626 excites 'and' after action verbs like click or stand
- L7:138 excites 'to' after verbs like prove or time
- L14:896 suppresses
- L15:2765 weakly excites
- L12:446 weakly excites
- L10:289 weakly excites
- L6:1880 weakly excites "by" introducing a deadline or means
- L8:369 weakly suppresses 'of' after 'learned' or 'because'
- L6:228 weakly excites comma after a filler word
- L16:1540 weakly excites
- L6:2509 weakly suppresses
- L19:121 weakly excites
- L19:1305 weakly suppresses unclear: verbs after a noun or 'that'
- L21:2238 weakly suppresses
- L20:879 weakly suppresses
- L21:1594 weakly suppresses
- L7:874 weakly suppresses
- L10:1392 weakly suppresses
- L8:70 weakly excites 'and' between stacked superlatives
- L21:2308 weakly excites
- L24:1424 weakly suppresses
- L6:347 weakly suppresses
- L17:158 weakly excites
- L15:452 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
- eva
- しかも
- 销
- były
- tod
- hra
- PIPE
- 摆在
- Export
- Spec
Suppresses
- ension
- bies
- �
- icator
- engers
- ents
- apped
- ителя
- istent
- ickest