"the" after "over" in time spans
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. 20 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 responds to "the" in phrases like "over the past decade" or "over the years". Turned up, the model produced strings of "the first" and "last" words.
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.
-
existing finish rather than create a new one in the restoration process. This process may risk the loss of the remaining<|endoftext|>With the development of computers and the internet over the past several decades, communication and written language have undergone drastic transformations. Until a couple
-
date backwards, using B.C. dates, to the time of the creation of the world. Provisionally, based on the studies published in Biblical Chronology over the past two years, I have come to the year 3932 B
-
colonies are disappearing fast. According to the nonprofit organization Pollinator Partnership, the United States has lost over 50 percent of its managed honeybee colonies over the past 10 years.⏎Pollinator gardens are just one of the many ways
-
. “The shellfish results indicate the strength of the growing aquaculture sector.”⏎Supply and demand of farmed oysters have risen exponentially in North Carolina over the past
-
proximity to our brains, many people are worried that they might be associated with an increased risk of brain tumors. There have been around 30 research studies over the past decade that have looked at this issue and most have found that:⏎- People
-
of large balloons designed to beam the Internet to unwired places—developed the technology over the last 18 months. Google’s blog reports:⏎“Over the years, many scientists have investigated various body fluids—such as tears—in the hopes
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 am a 36-year-old, 43-bmd, 56-bmd, 56-bmd, 44-bmd, 46-bmd, 46-bmd, 44
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
. Does not equal residence. Can only have one domicile at a time.⏎ii. Person starts out life with domicile of origin:⏎1. determined by parents.⏎iii. Acquire a domicile of choice when abandon the domicile of origin; the domicile
-
of methane emissions. According to the Environmental Protection Agency, methane can remain in the atmosphere for up to 15 years and is 20 times more likely to trap heat
-
Know the Difference⏎- Excessive Heat Watch⏎– Conditions are favorable for an excessive heat event to meet or exceed local Excessive Heat Warning criteria in the next 24 to 72 hours.⏎- Excessive Heat Warning⏎– Heat
-
2x√(1+x6)⏎Comment. There is no need to integrate and then differentiate by x.⏎If you want to type x6 instead of x^6 simply type x6, highlight only 6 and click on the icon (
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
- L19:1263 excites
- L19:3065 excites 'for' after a span of time
- L16:1027 excites
- L17:1912 suppresses
- L17:2544 suppresses
- L19:254 suppresses
- L5:2700 excites
- L19:2789 suppresses
- L15:1825 excites
- L18:291 suppresses
- L18:1487 weakly suppresses
- L17:1842 weakly suppresses
- L13:2164 weakly excites
- L19:2563 weakly excites
- L9:1575 weakly excites
- L10:2995 weakly excites
- L19:1419 weakly excites
- L19:2151 weakly excites
- L18:3013 weakly suppresses
- L18:2750 weakly suppresses
- L17:2446 weakly suppresses
- L16:1898 weakly excites
- L17:2887 weakly excites
- L17:2908 weakly excites
Feeds later layers
- L21:807 excites
- L26:2398 excites
- L27:910 excites
- L25:869 excites
- L24:1775 excites
- L22:1470 excites
- L26:1094 excites
- L23:892 excites "the" in "over the past" time spans
- L22:3024 excites
- L21:1797 excites
- L23:1339 excites
- L23:982 excites
- L21:1248 excites
- L23:3063 excites
- L24:2546 excites
- L24:2353 weakly excites
- L24:549 weakly excites
- L21:2207 weakly excites
- L22:2987 weakly excites
- L21:2283 weakly excites
- L21:2559 weakly excites
- L24:995 weakly excites
- L24:1835 weakly excites
- L22:2156 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
- last
- last
- first
- 最后一个
- Last
- first
- (last
- [last
- .last
- 第一个
Suppresses
- yours
- 可能です
- 属实
- onden
- ours
- ่าง
- ancement
- itud
- RoutingModule
- 습