unclear: words like 'different', 'close', 'all'
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 fairly consistent: several of them 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 a mixed set of ordinary words, such as 'different', 'close', 'all' and 'agree', that tend to sit before a qualifier like 'levels' or 'walks of life'. It does not show one clear topic, and turning it up made the model drift into unrelated text rather than anything tied to these words.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
ke?⏎Even players with the same language background and goals can play together. In other words, though both my daughter and I desire to learn Norwegian and are at different levels ourselves, we can successfully play the game together and learn from one another.
-
we think this problem is relevant for Professor Finnegan's class at WSU.⏎What textbook is this problem found in?⏎Our data indicates that this problem or a close variation was asked in Chemistry: An Atoms First Approach - Zumdahl At
-
monastery in sixth-century Egypt is impossible, the lives of late antique Roman elites and Egyptian monks did intersect. Desert Fathers, their monks, and their visitors came from all walks of life and practiced asceticism with varying degrees of rigor. Certainly,
-
0 states with the biggest one-year increases in CO2 emissions, with Georgia, Arizona, California, Pennsylvania, Michigan, Iowa, Illinois, Virginia and North Carolina close behind.⏎The top three states – Texas, Georgia and Arizona – had the greatest
-
gleaming, cheap , efficient and connects delhi end to end along with its neighboring cities.⏎Delhi has been home to many people for centuries and people from all part of India have stayed here and have enriched the city with their ethos and customs
-
?⏎In order for a sentence to be grammatically correct, the subject and verb must both be singular or plural. In other words, the subject and verb must agree with one another in their tense. If the subject is in plural form, 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: I know that, on any two any two 2008 2008 2008 2008 2008 2008 2008 200
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
, e-mail FormulateThis@allured.com.⏎Question: I’m trying to understand cosmetic emulsion inversion. What actually happens inside the emulsion at the phase borders?⏎Answer: First, remember the HLB system. For ethoxyl
-
As the temperature is lowered, the HLB increases and there is a point at which the emulsifiers then orient to form an o/w emulsion (the Phase Inversion Temperature or PIT), resulting in an inverse emulsion.⏎From a theoretical
-
a phase boundary, for instance, gas-oil, oil-water, and<|endoftext|>Green kitchen
-
they do add up to equal the total, but a voltmeter doesn't provide phase angle measurements, only amplitude.⏎Try measuring voltage dropped across both resistors at once
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
- L18:2885 suppresses
- L7:653 weakly suppresses
- L19:2520 weakly excites
- L8:1337 weakly excites
- L20:635 weakly excites
- L13:102 weakly excites
- L9:1994 weakly excites unclear: sun angles, geometry and road-safety terms
- L19:1160 weakly excites
- L9:153 weakly excites
- L6:398 weakly excites
- L8:1098 weakly suppresses
- L3:1524 weakly excites
- L6:2162 weakly suppresses
- L19:2904 weakly excites
- L8:2996 weakly suppresses
- L0:2694 weakly suppresses
- L21:149 weakly excites
- L8:689 weakly excites
- L3:208 weakly suppresses
- L16:2314 weakly excites
- L10:1553 weakly excites
- L14:2550 weakly excites
- L21:1759 weakly suppresses
- L10:558 weakly excites
Feeds later layers
- L24:423 excites
- L23:2358 weakly excites
- L27:1164 weakly excites
- L27:1525 weakly excites
- L23:1527 weakly excites
- L25:1612 weakly excites
- L27:1704 weakly excites
- L24:835 weakly excites
- L24:2486 weakly excites
- L25:2597 weakly excites
- L27:2064 weakly excites
- L27:903 weakly excites
- L23:1015 weakly suppresses
- L26:2513 weakly excites
- L24:1942 weakly suppresses
- L23:2774 weakly suppresses
- L25:1049 weakly excites
- L26:842 weakly excites
- L25:432 weakly suppresses
- L25:2565 weakly excites
- L24:1660 weakly excites
- L24:1629 weakly excites
- L23:989 weakly excites
- L27:1844 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
- 物理
- sociale
- soci
- phys
- sociology
- sociales
- 哲学
- geography
- socially
- 地理
Suppresses
- LoginForm
- pez
- 某
- 贴
- Luxembourg
- .lazy
- licit
- lik
- 查看全文
- ılan