unclear: words before a source or reference
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. 12 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 is hard to pin down. It fires on words that often come just before a source, a reference or a continuation such as 'from', 'Source' or 'index'. Turning it up gave off-topic text, so the steering does not help.
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.
-
six sections has an introduction by the editor including the latest research, references, and updates in the field. This book is enhanced with numerous color illustrations, a comprehensive index, and the many computer program examples encourage reader
-
factory automation equipment; their site has a great section on roll-to-roll plastic electronics assembly technologies. It is fascinating to look through the process – it gets my imagination running on what the future of light engines will look like.⏎Here is their PDF brochure
-
Please remember to answer the questions in full sentences.⏎- Where did the children used to play?⏎- Describe the garden. What did it look like? Use words from the text to help you.⏎- Were the children
-
she was a “bride,” so that one can easily tell at a glance if a man be or be not mortgaged as to his affections.⏎Source: "German Betrothal Customs," (Harvard) Herald, 6
-
Smith first introduces herself by saying, “Hi Stephanie, my name is Dr. Smith, but you can call me Julie. Nice to meet you.” Stephanie looks up and nods but does not say anything. Dr. Smith introduces herself to Stephanie’s
-
00 wounded.⏎- Richter, G (1870). Manual of Coorg. Stolz. p. 337. Retrieved 14 August 2014.⏎- Archer, Jeremy. "Sir
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 how of a single person, the reason, and the relationship between the human and the animal. I have been thinking about people, but only the human in one person, but the human in one person, the relationship between the human and
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
curator and historian, is in charge of the whole collection.⏎A unique piece of art⏎The entire exhibit is composed of different pieces. While each piece aims to depict a different story, together they tell a more beautiful tale. It is unlike any
-
This allows children to draw and write to show their mathematical thinking.⏎Contents 4-8 Counting and Cardinality Task Cards
-
Kaplan. The expose was written to show the appalling condition in various state schools through the country. The majority of the institutions
-
abuses across the world for their simplicity, universality, and ordinariness.⏎What type of person was Robert Frost?⏎Robert Frost was an American poet who depicted realistic New England life through language and situations familiar to the common man. He won
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
- L6:3018 suppresses
- L6:2735 weakly excites
- L10:524 weakly excites
- L11:2350 weakly excites
- L11:1998 weakly suppresses
- L4:766 weakly excites
- L10:741 weakly excites
- L4:1175 weakly suppresses
- L11:2159 weakly suppresses
- L6:1551 weakly suppresses
- L11:914 weakly suppresses
- L10:1068 weakly excites
- L3:1037 weakly suppresses
- L9:2665 weakly excites
- L5:697 weakly excites
- L9:2671 weakly excites
- L3:2586 weakly suppresses
- L6:2139 weakly excites
- L7:170 weakly suppresses
- L11:754 weakly excites
- L9:1185 weakly excites
- L11:2985 weakly excites
- L6:816 weakly suppresses "D" in vitamin D
- L10:858 weakly suppresses digit 1 in COVID-19 and 1941
Feeds later layers
- L19:2359 weakly suppresses
- L18:1172 weakly suppresses
- L14:2460 weakly suppresses
- L16:266 weakly excites
- L22:1036 weakly suppresses
- L20:1615 weakly suppresses
- L21:52 weakly excites
- L15:2073 weakly suppresses
- L13:2238 weakly suppresses
- L21:2546 weakly suppresses
- L23:2743 weakly suppresses
- L15:1138 weakly suppresses
- L26:1660 weakly suppresses
- L15:2394 weakly suppresses
- L18:979 weakly suppresses
- L24:1984 weakly suppresses
- L23:941 weakly suppresses
- L18:2269 weakly suppresses
- L14:2597 weakly suppresses
- L17:3038 weakly suppresses
- L21:31 weakly excites
- L13:1515 weakly excites "your" before doctor or health provider
- L20:2298 weakly suppresses
- L20:1131 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
- liner
- ANCES
- lickr
- Recipes
- slack
- preced
- edList
- éis
- bin
- pas
Suppresses
- phis
- escription
- organism
- нюю
- 茧
- organisms
- Planet
- DETAIL
- (detail
- Private