"self" before esteem or image
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 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 word "self" when it comes before "esteem", "image" or "worth", as in self-esteem. When turned up, the model kept producing the word "image".
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.
-
for the hard work you ask them to do when you train and compete, an unhealthy focus on these elements of yourself can have negative impacts on your mental health and self esteem.⏎Some actions and thoughts to consider;⏎Don't compare yourself to others,
-
will is crushed by suppressive parents, they could end up depressed, anxious or ineffective in their life. A thwarted free will can significantly impact the child’s self-worth.⏎During earlier stages of human society, compliance was a positive and necessary
-
give us something to look forward to, provide a social outlet, allay worries about health and fitness, increase energy levels, and improve self-esteem. All these factors can lessen the severity of anxiety and depression. Exercise also
-
UNICEF show that participation in athletics reaps countless long-term benefits for girls. For example, playing sports has immense psychological benefits for girls, such as improved self image, increased self confidence, positive feelings and knowledge about one’s body, and
-
of suicide. Even years after the bullying has stopped, researchers have found that adults who were bullied as teens have higher levels of depression and poorer self-esteem than other adults.⏎A recent study that<|endoftext|>|Cherokee County, Texas
-
1,148% growth rate.⏎These children endure stigmatization and/or bullying on a daily basis. Many are emotionally shattered; most have a poor self-image and ALL are vulnerable. In addition, many of the
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 the call bously called the test, the test had a call from the call from the call, which called the call back, then the call from the call from the call, which called back the call, then the call from the
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
CE) saw the rise of Zen Buddhism in Japan, which greatly influenced Japanese art. Zen-inspired ink paintings and tea ceremony ceramics became popular during this time.⏎The Edo period (1603 CE to 1868 CE)
-
aren’t supposed to be greedy for gain (1 Tim. 3:8). That may suggest they were responsible for the finances of the church
-
the young pastor (1 Timothy 1:18). God continues to use children today in their schools, homes
-
a genre of Japanese art that flourished during the Edo period (1603-1868). It is a
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
- L4:1924 weakly suppresses
- L5:2636 weakly excites
- L19:1385 weakly excites
- L20:2513 weakly suppresses
- L21:1539 weakly excites
- L17:2391 weakly excites
- L18:1376 weakly suppresses the word 'case' in set phrases
- L23:2629 weakly suppresses
- L19:929 weakly excites
- L19:130 weakly suppresses
- L23:2090 weakly excites
- L13:1731 weakly suppresses
- L18:336 weakly suppresses
- L10:411 weakly suppresses
- L8:920 weakly excites
- L23:649 weakly excites
- L22:2555 weakly suppresses
- L6:1206 weakly excites
- L5:218 weakly suppresses
- L7:406 weakly suppresses
- L11:816 weakly excites
- L9:1024 weakly excites
- L21:381 weakly excites zeros inside long number codes
- L4:101 weakly excites
Feeds later layers
- L27:1316 strongly excites
- L25:1584 excites art and artist names in lists
- L26:558 weakly excites
- L25:147 weakly suppresses
- L25:1895 weakly suppresses first digit of a year like 2020
- L27:902 weakly excites
- L27:768 weakly excites
- L26:1218 weakly suppresses
- L25:322 weakly suppresses
- L27:1235 weakly excites
- L26:1562 weakly suppresses
- L25:342 weakly excites
- L27:2371 weakly excites
- L25:995 weakly suppresses
- L25:845 weakly suppresses
- L27:1389 weakly suppresses
- L26:834 weakly excites
- L27:352 weakly excites
- L27:2066 weakly excites
- L27:2150 weakly excites
- L25:224 weakly suppresses
- L27:167 weakly suppresses
- L26:1384 weakly suppresses
- L27:1525 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
- image
- Image
- images
- image
- 形象
- (image
- 图像
- →image
- -image
- _image
Suppresses
- thesize
- boil
- retty
- ัน
- 줄
- Dot
- adal
- Middleware
- 鹩
- assen