research studies and survey follow-up
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. 24 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 on words inside research descriptions, like 'self-reported questionnaires', 'follow-up assessments' and 'participants completed a survey'. Turning it up made the model keep writing 'follow' and 'follow-up'.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
quality of life were assessed using self-reported questionnaires.⏎Study participants’ answers showed that 74.8% were
-
of more proficient students from language-minority homes during the instructional period. We administered follow-up assessments in the fall after the instructional period ended and in the spring of the following
-
Study on Memory, Health and Aging.⏎In 1995, 3,831 study participants 65 years of age or older completed a survey that included a food frequency questionnaire and cognitive assessment. Cognitive function was checked again
-
beginning of the study and over time. Those in the highest quintile scored 1.42 points higher at baseline and 1.81 points higher after 11 years on the 3MS than did those in the lowest quint
-
environmental characteristics relating to physical activity, perceived health status, social capital, fruit and vegetable consumption and levels of physical activity will be undertaken (baseline and at 2 year follow-up). Based on baseline
-
Adult Health (ELSA-Brazil) painted a picture of how some chronic diseases emerge and evolve in Brazil, and the possible interactions between them. Based on interviews and examinations of the study participants, physicians and researchers identified a link in the Brazilian sample
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: when we use the nonagonal (an = 10, non = 1000) nonagonal system, maybe not top to top, but top to bottom top, top to top top, top to top top top,
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
blood vessels are moved or pressed away from their usual position inside the skull.⏎Herniation syndrome; Transtentorial herniation; Uncal herniation; Subfalcine herniation; Tonsillar herniation; Herniation -
-
utor(s):||Arndt, T., Boesman, P., Olarte, L., Salaman, P. & Olaciregui , C.|⏎|Facilit
-
) and were monitored for improvement in their symptoms.⏎The researchers discovered that clozapine was approximately twice as likely to produce a treatment response as olanzapine. Both positive symptoms (psychosis) and negative
-
system of support (MTSS) or response to intervention framework in reading and mathematics. They provide examples of how to apply standards relevant instruction across core instruction (Tier 1), secondary intervention (Tier 2), intensive intervention (Tier 3)
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
- L21:205 suppresses
- L9:228 suppresses
- L18:839 suppresses
- L9:1035 suppresses
- L20:2011 weakly suppresses
- L6:2311 weakly excites
- L3:736 weakly excites
- L16:811 weakly suppresses
- L4:2715 weakly excites
- L15:1637 weakly suppresses
- L6:1951 weakly excites
- L3:37 weakly suppresses
- L4:2533 weakly suppresses
- L4:1389 weakly excites
- L22:954 weakly suppresses
- L5:319 weakly suppresses
- L8:1589 weakly excites
- L15:511 weakly excites
- L20:1384 weakly excites
- L21:1836 weakly excites
- L9:2593 weakly excites
- L4:2692 weakly excites
- L10:2705 weakly excites
- L20:150 weakly excites
Feeds later layers
- L26:1725 excites
- L25:2448 weakly excites
- L26:937 weakly excites
- L25:281 weakly suppresses
- L26:2815 weakly excites
- L26:939 weakly excites
- L25:865 weakly suppresses
- L26:1764 weakly excites
- L25:837 weakly excites
- L25:510 weakly excites
- L25:1749 weakly excites
- L25:1744 weakly excites
- L25:2266 weakly excites
- L26:2343 weakly suppresses
- L25:2766 weakly excites
- L26:2383 weakly suppresses
- L27:608 weakly suppresses
- L25:2652 weakly suppresses
- L27:1704 weakly suppresses
- L25:2837 weakly suppresses
- L26:2653 weakly excites
- L27:1242 weakly excites
- L25:2436 weakly suppresses
- L25:894 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
- Follow
- follow
- Follow
- follow
- -follow
- follows
- 随
- segu
- _follow
- FOLLOW
Suppresses
- кр
- coni
- top
- 顶部
- fter
- 无形
- 刻
- 高档
- elfth
- top