unclear: quantity and choice 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. 14 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 seems to respond to words that set up a choice or a range, such as 'both', 'neither', 'of' in a list, or numbers like 'Years 10 and 11'. The link is weak. Turning it down made the model repeat 'the following' over and over.
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.
-
plays, poetry and non-fiction texts.⏎All our teachers are English specialists.⏎For more information contact the head of department, Miss S Beadle.⏎In Years 10 and 11, all pupils study for the AQA GCSE English
-
, more than 200 social scientists (including myself) who study games in the fields of psychology and communication among others, signed a statement urging the American Psychological Association to repeal
-
Katherine who was neither divorced, beheaded, or died. She was by any standards a remarkable woman: beautiful enough to marry the King of England, despite having neither royal nor court background; shrewd enough to remain his Queen, despite court
-
ulsions. While the motion of single particles in Newtonian fluids is well understood, suspensions of large numbers of particles that interact through the fluid remain challenging, both analytically and numerically. A complicating factor is that in many application
-
opportunity to consume fruit and vegetables in abundance as well as quality protein, whole grains and low-fat dairy products or affordable plant milks. Regular, physical activity (both weight-bearing for musculoskeletal health benefits and intensive for cardiovascular health benefits) would
-
, 8 and 10 students take a writing test which consists of an essay and multiple choice questions. A reading and mathematics test is given in grades 3 through 10. In grades 5, 8
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 the following questions: 1. I have the following questions: 1: I have the following: 1: The following: 1: The following: 1: The following: 1: The following: 1: The following: 1:
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
general informational purposes only and is not a substitute for professional medical advice.⏎If you think you may have a medical emergency, call your doctor or (in the United States) 911 immediately. Always seek the advice of your doctor before starting
-
to those found in human blood. The most significant elements are:⏎Sea vegetables are also a rich source of fiber and protein, with some varieties containing up to 28% protein. They are also an excellent supply of vitamins A,
-
, how do you know if your baby is ready? Babies show certain signs of development or growth, which indicate they're ready to eat solid foods. One such sign is their ability to chew and swallow. There are some important signs of readiness,
-
and, Massachusetts, can be seen scurrying about the freshly dug school garden planting Swiss chard, spinach, peas, and pumpkin seeds. The first three items will be harvested by mid-June to make way for a second crop.⏎
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
- L12:42 suppresses
- L13:2381 suppresses
- L13:1675 excites
- L13:1063 suppresses
- L13:1731 excites
- L13:2006 suppresses
- L7:385 suppresses
- L11:172 suppresses
- L12:1189 excites
- L12:1873 excites
- L1:951 suppresses
- L13:1895 suppresses
- L13:2782 weakly excites
- L12:763 weakly suppresses
- L13:562 weakly excites
- L9:112 weakly suppresses
- L12:21 weakly suppresses
- L13:199 weakly suppresses
- L12:1814 weakly suppresses
- L13:328 weakly excites
- L12:1894 weakly suppresses
- L13:749 weakly excites
- L12:1087 weakly suppresses
- L12:463 weakly excites
Feeds later layers
- L16:2950 strongly suppresses
- L18:1746 strongly suppresses
- L16:2677 suppresses
- L15:709 suppresses
- L19:2423 suppresses
- L16:1127 excites opening bracket after an imperial measurement
- L18:486 excites
- L19:254 excites
- L17:1462 suppresses
- L17:2130 suppresses
- L15:247 excites
- L16:1039 suppresses
- L17:234 excites
- L18:2649 suppresses
- L16:473 suppresses
- L18:2872 suppresses
- L16:2568 suppresses
- L18:1978 suppresses
- L15:1994 suppresses
- L19:2131 suppresses
- L15:1679 suppresses
- L18:2867 suppresses
- L16:2801 suppresses
- L16:2748 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
- arella
- |_|
- (Level
- ory
- 唧
- IRON
- yours
- getProperty
- 意大
- ures
Suppresses
- 两个
- 两种
- 这三个
- 三个
- 三种
- 这两个
- arı
- both
- 哪些
- 这两种