'he' in pheasant, 'mouth' in Dartmouth
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. 1 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 responds to parts of words that are split up, like the 'he' in 'pheasant' and 'mouth' in 'Dartmouth'. Turning it up or down did not give an obvious, consistent change.
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.
-
include: Oliphant, Olifant, Olifard and others.⏎First found in Perthshire (Gaelic: Siorrachd Pheairt) former county in the present day Council Area of Perth and Kinross
-
- Great Hornbill (Buceros bicornis),⏎- Rufous-necked Hornbill (Aceros nipalensis),⏎- Grey Peacock Pheasant (Polyplectron bicalcaratum),⏎
-
pheasant) is literally 'mountain bird', and ashibiki is a pillow-word for mountain, which is itself the first half of the word for pheasant.<|endoftext|>North American freshwater fishes are going extinct at an alarming rate compared with
-
Pittsburgh Research Laboratory, Pittsburgh, PA 15236⏎Other Occupational Concerns⏎Longwall USA, International Exhibition & Conference<|endoftext|>Hypertext Markup Language. The standard language used to create Web pages. Originally based on SGML,
-
– could steer sexual behavior development in offspring. It’s even conceivable that taking antihistamines or pain relievers during pregnancy could play a role.⏎“We’re really interested in the fundamental mechanisms that drive brain development and sex-specific brain development, and
-
Old Japan (The Hyakunin-isshu), tr. by William N. Porter, , at sacred-texts.com⏎LONG is the mountain pheasant's tail⏎That curves down in its flight;⏎But longer still, it
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: I have been working on this project, and I have been thinking about it. But I have been working on it for a while, and I have been thinking about it for a while. It has been a long time. It has been
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
registered user you can vote on this brain teaser, keep track of⏎which ones you have seen, and even make your own.⏎Back to Top<|endoftext|>Dartmouth researchers have developed a protocol that permits cells harvested from melanoma tumors in mice to
-
iers’ Protect Greenland’s Landscapes⏎A story by LiveScience published by NBC News features a new study by Lee Corbett, a doctoral student in Dartmouth’s Department of Earth Sciences, that reveals how “ghost glaciers” have protected Greenland
-
⏎A University of Massachusetts Dartmouth education is a foundation from which graduates continually engage and impact their community, both locally
-
a model system," said Constance E. Brinckerhoff, PhD, professor of Medicine and of Biochemistry at the Geisel School of Medicine at Dartmouth and a member of the Norris Cotton Cancer Center (NCCC) Mechanism Research
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
- L0:1301 suppresses
- L0:1224 suppresses
- L0:178 suppresses
- L0:2308 weakly excites
- L0:1 weakly suppresses
- L0:581 weakly excites
- L0:799 weakly suppresses
- L0:2316 weakly suppresses
- L0:2621 weakly excites
- L0:1999 weakly excites
- L0:489 weakly excites
- L0:1000 weakly excites
- L0:654 weakly suppresses
- L0:15 weakly excites
- L0:33 weakly excites
- L0:3005 weakly excites
- L0:1453 weakly suppresses
- L0:1102 weakly excites
- L0:1837 weakly excites
- L0:40 weakly excites 'of' in 'sort of' hedges
- L0:2047 weakly excites
- L0:1634 weakly suppresses
- L0:2321 weakly suppresses
- L0:1761 weakly suppresses
Feeds later layers
- L3:1955 weakly excites
- L19:1312 weakly suppresses
- L14:3022 weakly suppresses
- L16:1740 weakly suppresses
- L21:992 weakly suppresses
- L14:2361 weakly excites
- L22:2083 weakly suppresses
- L7:2185 weakly suppresses
- L26:2465 weakly suppresses
- L6:2923 weakly excites
- L9:1355 weakly excites
- L15:167 weakly excites
- L26:651 weakly suppresses
- L16:1324 weakly suppresses
- L21:2113 weakly suppresses
- L18:2277 weakly suppresses
- L22:1129 weakly excites
- L12:882 weakly excites
- L6:623 weakly suppresses
- L20:2509 weakly suppresses
- L20:1746 weakly suppresses
- L18:229 weakly excites
- L14:2592 weakly excites
- L22:2390 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
- @endif
- lu
- variety
- dạng
- 种类
- ا
- 恃
- 信息发布
- ApiResponse
- mới
Suppresses
- SEL
- issen
- usalem
- geist
- adden
- gross
- oha
- uales
- unded
- iples