the variable t for 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. 5 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 letter t when it stands for time in a formula, such as sin(ωt) or the t in t-tests. When turned up, the model's writing did not clearly change in a way tied to time, so the evidence is weak.
Written by Claude from this neuron's activations and steering outputs. Confidence: medium. 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.
-
⏎t0=0; %set the initial time to zero⏎tend=1; %set the final time to 100⏎t
-
.m.⏎Inferential Statistics with SPSS⏎Using an SPSS data file, you'll test hypotheses with t-tests and regression analysis to predict risks and trends.⏎Tuesday, Oct. 1
-
power is then written as⏎Here the average of sin2 ωt over a cycle is 1/2 and that of sin ω t cos
-
it that way. That just means that I have to make sure that I use gravity as a negative number, since gravity always acts down.⏎vf = vi + at⏎= 56.3m/s + (-9.81m
-
is averaged over a complete cycle.⏎The alternating voltage and alternating current in the series RLC circuit at an instant are given by⏎v=Vm sin ωt and i = I m sin( ω t + φ)⏎where ϕ
-
the commands you give and the M-file are somewhat disparate. The process is illustrated using a single, simple differential equation:⏎Integrate this equation from t = 0 to t = 1. The exact solution can be found by
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: To get the most out of the product, the company's current capacity to process it, the company is currently looking for a profit, the company is trying to find a path to the next level, the next level requires a lot of work
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
. Four state agencies committed up to $2.4 million to this facility, which is expected to reduce carbon dioxide emissions by 30,000 tons per year.⏎“We are proud to recognize Landfill Methane Outreach Program partners who
-
partner and is working towards greener solutions for utility demands. Each year, Seattle’s stormwater carries more than 8,200 tons of toxic metals and volatile chemicals into the city’s waterways. In addition,
-
energy resource prevents greenhouse gas emissions and reduces landfill odors.⏎This year's LMOP winning projects will avoid the emissions of 546,000 tons of carbon dioxide equivalent per year, the equivalent of annual greenhouse gas emissions from nearly
-
dangerous"!⏎When you are interviewing potential hvac contractors, you will hear phrases such as total cooling capacity, sensible and latent capacity, btu, and tonnage.⏎The Air Conditioning Contractors of America (ACCA) have a consumer area
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
- L4:613 excites
- L2:1585 weakly excites
- L0:2925 weakly suppresses
- L4:3047 weakly excites
- L2:493 weakly suppresses
- L0:632 weakly suppresses
- L4:2089 weakly suppresses
- L0:2812 weakly suppresses
- L3:343 weakly suppresses
- L4:876 weakly suppresses
- L4:977 weakly excites
- L2:1399 weakly excites
- L0:246 weakly excites
- L2:2406 weakly suppresses
- L3:1319 weakly excites
- L3:2665 weakly suppresses
- L2:2749 weakly excites
- L3:2873 weakly excites
- L3:1723 weakly excites
- L2:1155 weakly excites
- L4:958 weakly suppresses
- L4:916 weakly suppresses
- L2:2923 weakly suppresses
- L4:2003 weakly excites
Feeds later layers
- L7:451 excites
- L22:1074 excites
- L9:383 excites
- L16:583 excites
- L20:1335 excites
- L25:565 excites
- L10:637 excites
- L14:1088 excites
- L12:2853 weakly excites
- L26:2065 weakly excites
- L12:2338 weakly excites
- L19:1929 weakly excites timestamps and dates in text
- L19:2242 weakly excites
- L8:1257 weakly excites
- L22:2601 weakly excites
- L26:1324 weakly excites
- L8:233 weakly excites
- L9:2117 weakly suppresses
- L8:488 weakly excites
- L10:456 weakly excites
- L19:1937 weakly suppresses
- L10:2050 weakly suppresses
- L19:1390 weakly suppresses
- L22:2445 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
- 时间
- time
- 一天
- 时间和
- Ờ
- 时间段
- поя
- (time
- TIM
- coh
Suppresses
- APTER
- metis
- abet
- ave
- .setLevel
- itary
- .amazonaws
- deaux
- aceutical
- 尽