'time' in 'at the time'
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. 9 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 'time' in the phrase 'at the time', as in 'at the time of its deployment' or 'at the time Campbell began'. When it was turned up, the replies did not show a clear shift in topic.
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.
-
Ann Lund, from the university, carried out research into the kinds of scoliosis treatments available at the time Richard was alive.⏎They included looking at 11th century Persian polymath
-
913, by the Geographic Board of Canada. The work done by the American Bureau was monumental, well informed and incorporated the most advanced scholarship available at the time. In many respects, the information is still useful today, although prudence should
-
took place on April 22, 1970 brought 20 million Americans – 10% of the U.S. population at the time
-
course, and a spa.⏎The construction of the hotel was no easy feat. The building was constructed using steel and concrete, which was a new technique at the time. The dome, which is the most notable feature of the hotel, was constructed
-
0, 1909, the June festivities drew over 300,000 people (larger than the population of Queens at the time) to see the bridge lit
-
in the wrong soil, they won’t grow,” Campbell hypothesizes. “But in Polk County, maybe they can.”⏎Despite this diversity, at the time Campbell began his work in 2011, Polk was one of
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: "Yes, I like it. It's not so bad. It's not bad." So far, there's a negative influence, but it's not so bad. So far, there's a negative influence, but it's not so
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
: Signs to Watch For⏎Would you recognize the early signs of Alzheimer’s disease in yourself or a loved one? An Alzheimer Society of Canada survey of baby boomers has revealed a surprising lack of awareness of symptoms.⏎Only 50 per cent
-
- Rapid heart rate⏎- Sweating, clammy skin⏎- Tremor of the hands or other body parts⏎A severe form of alcohol withdrawal called delirium tremens can cause:⏎Exams and Tests⏎Your health care provider will
-
to Delta-9-tetrahydrocannabinol, or THC, the main psychoactive ingredient in marijuana.⏎The nonhuman primates, all of reproductive age with a history of successful conception, received a THC edible once daily over the course
-
, while the ship was halfway across the Pacific Ocean. The liner continued to Auckland, New Zealand, where two of the passengers – a couple in their 70's – were airlifted to hospital. More than 1,10
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
- L6:2854 excites
- L7:2232 excites
- L8:2558 excites
- L8:929 excites
- L8:2860 weakly suppresses
- L8:2322 weakly suppresses
- L8:1916 weakly excites
- L5:2351 weakly excites
- L8:1351 weakly suppresses
- L8:523 weakly excites
- L4:2321 weakly excites
- L8:1739 weakly suppresses
- L7:1534 weakly excites
- L7:2753 weakly suppresses
- L8:2511 weakly excites
- L7:1594 weakly suppresses
- L7:1884 weakly excites
- L0:1629 weakly excites
- L6:1622 weakly excites
- L6:2811 weakly suppresses
- L6:1918 weakly excites
- L3:2595 weakly excites
- L7:2655 weakly excites
- L8:1388 weakly suppresses
Feeds later layers
- L14:2984 excites
- L12:2830 excites
- L10:456 excites
- L12:742 excites
- L15:3033 excites
- L10:1693 excites
- L21:794 excites
- L10:591 excites
- L22:1904 excites ages right before a number
- L16:2301 excites
- L19:1551 excites
- L13:2282 excites
- L13:745 excites
- L23:1612 excites
- L16:1992 excites
- L13:1327 excites
- L18:2274 excites
- L17:3034 excites
- L10:1214 suppresses
- L20:2078 excites
- L11:817 excites
- L16:1815 excites
- L16:583 excites
- L12:2853 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
- 当时的
- currentTime
- AppState
- 那时候
- 时光
- currentState
- 時代
- 的时代
- 当时
- alive
Suppresses
- erator
- ctions
- -sided
- ISTR
- ymmetric
- CTIONS
- цион
- 原因
- zus
- sters