opening parenthesis for a gloss or abbreviation
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. 13 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly 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 the bracket that comes right after a term to explain it, such as a translation or the full form of an abbreviation. Its negative side fires on 'which' after a comma. Turning it up did not produce any clear related text.
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.
-
portrayed as being conducted by two dare-devil aviators. Nothing could be further from the truth. In fact Alcock was trained and flew in the RNAS (Royal Naval Air Service) and Brown was trained and flew in the RFC (Royal
-
Yoga is a form of exercise that focuses on strength, flexibility and breathing to boost physical and mental wellbeing. The main components of a yoga class are asanas (postures) and movements designed to increase strength and flexibility. Breathing is actively
-
.⏎Additional diagnostic tests include:⏎Specific treatment for osteogenesis imperfecta (OI) will be determined by your child's physician based on the following:⏎To
-
that you may have gingivitis, schedule an appointment at Something To Smile About for a dental exam.⏎What is gingivitis?⏎Gingivitis is the first and only stage that periodontal disease can be reversed. This is because
-
Rates for the Global Poor,"⏎American Economic Journal: Applied Economics,⏎American Economic Association, vol. 3(2), pages 137-66, April.⏎- Angus Deaton & Olivier Dupriez, 200
-
both the beginning of the Jewish high holidays and the civil year on the Jewish calendar. It’s a good time to wish your Jewish friends leshana tova (happy New Year), to send greeting cards, and perhaps plant trees in Israel in
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 it's a problem of the non-constant and the problem of the non-constant, which is the two separate problems of the non-constant and the two separate problems that the are the non-constant, which is the two separate
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
much, start with five or 10 minutes and gradually build up, adding a minute at a time.⏎Exercise can help keep you stay fit during pregnancy, which will be helpful during labor, and may help limit gestational diabetes.⏎For more
-
later. It is possible to send and receive bitcoins using a Bitcoin wallet.⏎Even when the value of an asset drops, it’s possible to make a profit, which is counterintuitive
-
help increase flexibility, which is essential for overall health. Improved flexibility can help you perform everyday activities more easily
-
continuous running and kicking. Physical contact is against the rules, which makes it a team sport that is less injury-prone than what we know as
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
- L12:2704 suppresses
- L12:2424 suppresses
- L6:904 suppresses
- L7:1253 suppresses
- L12:1095 excites
- L10:271 excites
- L10:206 excites colon or equals sign before a definition
- L5:1035 weakly excites
- L11:1910 weakly suppresses
- L9:1065 weakly suppresses
- L9:298 weakly suppresses
- L7:2850 weakly excites
- L10:2003 weakly excites
- L11:2066 weakly suppresses
- L12:2240 weakly suppresses
- L8:485 weakly suppresses
- L6:1551 weakly suppresses
- L6:1620 weakly excites 'in other words' restatement cues
- L7:2184 weakly suppresses
- L8:1885 weakly suppresses
- L5:952 weakly suppresses
- L6:1324 weakly excites
- L6:1471 weakly excites
- L12:2606 weakly excites
Feeds later layers
- L16:637 excites
- L23:1100 suppresses
- L19:3020 suppresses
- L20:930 suppresses
- L16:918 suppresses
- L14:1592 weakly suppresses
- L14:1244 weakly suppresses
- L15:2030 weakly suppresses
- L15:2554 weakly suppresses
- L15:2092 weakly suppresses
- L15:2938 weakly suppresses comma after an if-clause
- L14:1538 weakly suppresses
- L21:2380 weakly suppresses
- L18:2246 weakly suppresses
- L15:1543 weakly suppresses
- L22:1253 weakly suppresses
- L16:2177 weakly suppresses
- L17:984 weakly suppresses
- L22:1807 weakly suppresses
- L23:1671 weakly suppresses
- L20:984 weakly suppresses
- L15:1875 weakly excites
- L17:1508 weakly suppresses
- L16:2609 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
- abol
- óng
- ện
- useMemo
- 宪
- ähr
- theid
- รณ
- @param
- ув
Suppresses
- ilee
- iped
- 跡
- ethyst
- rophic
- ersh
- التجارية
- anical
- iy
- ASIC