cycles and periods, before the word 'cycle'
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. 22 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. 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 talk about repeating cycles, such as tides, moon phases, seasons and sleep rhythms, especially at the spot where the word 'cycle' would come next. When it was turned up, the model kept writing the word 'cycle' over and over.
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.
-
current. In an AC circuit, the voltage and current vary continuously with time. Let us first calculate the power at an instant and then it is averaged over a complete cycle.⏎The alternating voltage and alternating current in the series RLC circuit at an
-
. These charged particles collide with gas molecules in the atmosphere. The energy released in these collisions is given off as light.⏎Geomagnetic storms follow the 11-year solar
-
, it’s our phones, TVs, and laptops that are impairing our sleep.⏎In healthy adults, our biological clock follows a 24-hour wake-sleep cycle. As the sun rises and daylight gradually enters through the curtains, the body
-
.<|endoftext|>In ancient China one’s Zodiac sign was fixed by the position of the planet Jupiter during the year of one’s birth (it takes twelve years for Jupiter to orbit the sun). On this level, Ai Weiwei Circle of Animals/Zodiac
-
all still reading this post, right?⏎Well, let’s start off with a bit of backstory: For generations, a woman’s monthly cycle has been tied to the moon cycle likely because they’re about the same length–with a “typical
-
repeating event per unit time. It is also referred to as temporal frequency, which emphasizes the contrast to spatial frequency and angular frequency. The period is the duration of one cycle in a repeating event, so the period is the reciprocal of the frequency.⏎
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 can only be a good agent on a first- and second- or third-mention key to a particular problem at a particular time. I can only be a good agent when I have a good number of other agents at work that I
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
laws of physics and chemistry (a mixture of chance and necessity or "natural law") and not by design. This is the naturalistic explanation. Darwinism (evolution based on the hypothesized natural selection of random mutations) is the mechanism that
-
chance of failing to pass on a beneficial mutation ).⏎As noted elsewhere (see ‘Giving up on reality’), Lenski seemed to have given up on ‘evolution in the lab’ and resorted to<|endoftext|>Paul McCartney heard the melody to
-
herbs,shrubs,trees). Animals were grouped according to their habitat and physical differences. The problem with this system was it did not group organisms into their evolutionary history.⏎How did Aristole classify life? What was the problem with his system
-
claims:⏎‘Lenski’s experiment is also yet another poke in the eye for anti-evolutionists, notes Jerry Coyne, an evolutionary biologist at the University of Chicago
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
- L21:1710 excites
- L19:3004 suppresses
- L16:2979 suppresses
- L18:1837 suppresses
- L12:1879 excites
- L8:2650 excites
- L10:1707 suppresses
- L9:2310 suppresses
- L12:2912 suppresses
- L10:2870 suppresses
- L17:1571 excites
- L9:2937 excites
- L18:1157 excites
- L18:1149 suppresses
- L20:2982 excites
- L11:2586 suppresses
- L11:350 excites
- L15:2543 weakly excites
- L19:562 weakly excites
- L9:2069 weakly suppresses
- L17:2600 weakly excites
- L12:2782 weakly suppresses
- L15:1829 weakly excites
- L21:2065 weakly excites
Feeds later layers
- L26:2155 excites
- L23:2078 excites
- L23:2557 excites
- L27:2128 excites
- L26:2396 excites
- L24:2394 excites
- L25:1915 excites
- L27:3070 weakly excites
- L27:495 weakly excites
- L24:906 weakly excites
- L24:814 weakly excites
- L25:2678 weakly excites
- L23:2496 weakly excites
- L27:2304 weakly excites
- L24:89 weakly excites
- L25:3037 weakly excites
- L27:3046 weakly excites
- L26:519 weakly excites
- L26:1448 weakly excites
- L26:1862 weakly excites
- L26:1400 weakly excites
- L25:1558 weakly excites
- L25:2505 weakly excites
- L26:1001 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
- cycle
- cycles
- Cycle
- cycle
- Cycle
- cycles
- _cycle
- 周期
- -cycle
- 循环
Suppresses
- Katz
- 墒
- VIDIA
- Hearts
- marg
- 觚
- arrass
- ancias
- Proud
- Disabilities