closing bracket after an acronym
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. 6 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 closing bracket that comes right after a short form, such as the ')' in 'cerebral palsy (CP)'. It is strongly negative on words like 'stands' or 'for' that come right after an acronym, where a meaning is about to be spelled out. Turning it down made the model talk about abbreviations and what letters stand for.
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.
-
Agency, 2020; Pulse Nigeria, 2020). The World Health Organization has labeled Nigeria a human resources for health (HRH) crisis country (Global Health Workforce Alliance, 2020).⏎The
-
set monthly from October 2016-April 2017 for 1-2 weeks. Submersible ultrasonic receivers (SURs) were dispersed throughout the backwaters and river channel to detect fish movement
-
NINDS)⏎NINDS offers research information related to neurological disorders.⏎United Cerebral Palsy⏎This organization provides information about cerebral palsy (CP) as well as new research and therapies.⏎National Dissemination Center for Children with
-
Rice (GR) based on guidelines set by a 5-interdepartmental circular is a most welcome, most awaited development.⏎Vitamin A deficiency (VAD) continues
-
adhesives, paints, and petroleum products, have been detected in about one-third of the wells sampled by the National Water-Quality Assessment (NAWQA) Program of the U.S. Geological Survey (USGS).1 Chloroform
-
or diabetes type 2.⏎Diabetes is worldwide in distribution and the incidence of both types of primary diabetes, i.e insulin dependent diabetes mellitus (IDDM) or diabetes mellitus type 1 and non insulin dependent diabetes (NIDDM
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 will be using the term 'I' for the first word in the sentence 'Institute'." However, this term is not a standard term, and I need to be more specific about the term. Please provide the term and
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
see leaping robots in the future.<|endoftext|>What is an AVA?⏎The term AVA is short for “American Viticultural Area”. AVA’s are official wine grape growing regions of
-
S.M.A.R.T. Goals?⏎S.M.A.R.T. stands for Specific, Measurable, Attainable, Relevant, and Time Bound.
-
time because their learning system is so restricted, trainee have no chance to learn things other than what the teacher have actually taught. # 3 WPA stand for Works Progress Administration.⏎He will not work for it since he is to pleased with
-
and listen to each other.⏎- Identify points where team members make decisions⏎- Note whether all members of the team are involved.<|endoftext|>The acronym SWOT stands for strengths, weaknesses, opportunities and threats. Organizations of all sizes use SWOT analysis
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
- L5:2782 strongly excites
- L4:2025 weakly excites
- L4:975 weakly excites
- L1:2898 weakly excites
- L4:2720 weakly suppresses
- L4:1220 weakly excites
- L5:197 weakly excites
- L5:712 weakly excites
- L5:1221 weakly excites
- L5:160 weakly excites
- L5:169 weakly excites
- L4:984 weakly suppresses
- L5:1466 weakly suppresses
- L3:1559 weakly suppresses
- L3:263 weakly suppresses
- L5:57 weakly suppresses
- L5:94 weakly suppresses
- L0:2519 weakly excites
- L1:2342 weakly excites
- L4:2317 weakly suppresses
- L3:875 weakly excites
- L1:2326 weakly excites
- L2:2792 weakly suppresses
- L0:107 weakly suppresses newline starting a new block
Feeds later layers
- L7:3064 suppresses
- L10:1574 suppresses
- L18:2695 suppresses
- L10:2107 suppresses
- L13:1140 suppresses
- L17:2755 suppresses
- L7:2927 weakly suppresses
- L22:1991 weakly excites
- L20:2001 weakly suppresses
- L19:1913 weakly suppresses
- L11:1893 weakly suppresses
- L7:2381 weakly excites
- L9:2946 weakly suppresses
- L20:766 weakly suppresses
- L19:1011 weakly suppresses
- L8:2452 weakly suppresses
- L19:2801 weakly suppresses
- L16:2209 weakly suppresses
- L16:2107 weakly excites
- L12:241 weakly suppresses
- L13:2379 weakly suppresses the digit zero in numbers
- L11:2670 weakly suppresses
- L21:2820 weakly suppresses
- L17:2498 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
- validates
- IVES
- ppe
- fires
- 絮
- udeau
- ersive
- 范围内
- 真人
- 吁
Suppresses
- abbreviation
- acronym
- abbreviation
- ronym
- abbrev
- \"$
- colo
- 单词
- steht
- '$