the total after "N of" in fractions
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. 8 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
Fires on the blank spot right after "of" in phrases like "Page 1 of", "out of" or "10 of the", where the model is about to write the total number. The strongest firing in the set comes from these spots. Turning it up did not bring out a matching theme in the generated text.
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.
-
2 – August 20.<|endoftext|>Genetics as it applies to evolution, molecular biology, and medical aspects.⏎10 posts • Page 1 of 1⏎Hope somebody can help, I've got an A level biology exam on
-
of bone that are not subject to great mechanical stress. It makes up much of ... (100 of 310 words)<|endoftext|>How do crabs usually die?⏎How do cr
-
24 attendees – aged one to 56. Twenty-two of the victims needed hospitalization. As of May 2nd, 10 of the 22 were still in hospital and in stable condition.⏎Kalikot District,
-
analysis. The number of answer matches between any pair of exam papers can be compared with the expected number of answer matches. Example: John (62 correct out of 100) and Barbara (67 correct out of 10
-
be introduced. Generalized anxiety disorder is estimated to occur in 5% of the general population. Women are generally more affected than men .⏎Page 1 of 2⏎Dating Advice For Men<|endoftext|>Feb 5, 2020
-
025.SS UNSPSC Code 30260000 See more.Customers who bought this item also bought.Page 1 of 1 Start over Page 1 of 1 .Previous page.1/4
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: the history of the development of the United States, the years 1965–1900, and the Truman and postwar years. This is the only basis for your analysis. What do you mean? — I can
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
know about the dangers of fires at home. Many households take active steps toward protecting family members from the risks of fire. This includes installing smoke detectors, keeping fire extinguishers charge, and establishing an emergency plan in case family members need to exit
-
and the personal pursuit of happiness. Undue taxation that provides pork for a<|endoftext|>Find All Factors⏎A sure way to find all factors of a number using prime factorization.⏎Algebra.com [offsite link]⏎Algebra.com has a variety
-
SE) = 10 log10(MAX/(MSE)^(1/2))^2 = 20 log10(MAX/(MSE)^(1/2)). Therefore, PSNR = <|endoftext|>In my current position as
-
102 ≈ 0.97. The expected number of powers of 2 without zeroes starting from the power N is cxN/(1-x) ≈ 40 ⋅ 0.97N.⏎Let
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
- L4:1432 excites
- L5:1006 excites
- L7:1476 weakly excites
- L5:110 weakly suppresses
- L6:419 weakly excites
- L2:2044 weakly excites
- L5:1212 weakly suppresses
- L1:652 weakly suppresses
- L2:2188 weakly suppresses
- L6:922 weakly suppresses
- L6:1046 weakly suppresses
- L1:440 weakly suppresses
- L7:1428 weakly suppresses
- L0:1423 weakly excites
- L2:2701 weakly excites
- L5:504 weakly suppresses
- L2:217 weakly excites
- L5:27 weakly excites
- L6:92 weakly excites
- L5:1159 weakly excites
- L1:437 weakly suppresses the digit 9 at a fragment start
- L1:2161 weakly excites
- L7:1982 weakly suppresses
- L0:55 weakly excites
Feeds later layers
- L9:2439 excites
- L21:2855 excites
- L10:1700 excites
- L22:1627 excites
- L11:2679 excites
- L15:2436 excites 'out of' in ratios and rankings
- L18:2519 excites
- L12:2889 weakly excites
- L15:2428 weakly excites
- L19:1851 weakly excites
- L12:196 weakly excites
- L17:2906 weakly excites
- L12:472 weakly excites
- L17:2984 weakly excites blank before a number in prevalence claims
- L9:679 weakly excites
- L12:3001 weakly excites
- L10:1329 weakly excites
- L18:1329 weakly suppresses
- L10:2016 weakly excites
- L12:2289 weakly excites 'of' after part-whole nouns
- L14:2674 weakly excites
- L15:3034 weakly excites
- L15:1926 weakly excites
- L10:1926 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
- 总量
- setContent
- 主流
- 共
- issent
- 总数
- 톱
- 大全
- 占比
- 全程
Suppresses
- eday
- nger
- responseData
- schlä
- agged
- iode
- 颛
- awai
- ��
- ffiti