"but" after a negated clause
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. 1 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
The neuron fires on "but" when it comes right after a "not X," phrase, as in "not X, but Y". Turning it up gave text that did not follow this pattern, so the steering adds little to the explanation.
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.
-
running. Slow to<|endoftext|>South-eastern Australia is one of the most bushfire-prone areas in the world.⏎The risk of bushfires cannot be eliminated, but their impact can be reduced by effective planning and bushfire management.⏎The National Parks
-
ask for the Negro is not benevolence, not pity, not sympathy, but simply justice.”⏎The 13th Amendment, ratified later in 18
-
as well as for psychological problems such as anxiety and panic. People often misunderstand the goal of therapy and especially mindfulness. Relaxation and happiness are not the aim, but rather a "freedom from the tendency to get drawn into automatic reactions to thoughts
-
that many jumping spiders, crab spiders, wolf spiders and other garden spiders are not indiscriminate diners, but in fact develop rather narrow tastes in prey. For example, a crab spider may
-
Rabbi Emil Fackenheim in his book, "God's Presence in History." "Nevertheless, not messengers, not angels, not intermediaries, but God himself acts in human history."⏎God hears the cries of the people Israel and
-
use of zero tolerance in schools since its inception in the 1980s. It found that not only does zero tolerance not achieve its stated goals, but that its most common punishments, suspension and expulsion, lead to increased drop
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 need to be sure I will be in a place that I need to be sure I will be in a place that I need to be sure I will be in a place that I need to be sure I will be in a place that
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
required for digestion of food, exercise, cleaning the house, everything. Definition of terms that create your total calorie burn:⏎- RMR=resting metabolic rate-the amount of calories you need to exist without any activity-ie:bedrest⏎-
-
Roosevelt dedicated the United States to the Policy of the Good Neighbor-"the neighbor who resolutely respects himself and, because he does so, respects the rights of others-the neighbor who respects his obligations and respects the sanctity of his agreements in and with
-
be 10 0 10 why am I getting 10 0 20.I am confused can some one help me in this regard.thanks in advance. kamesh. here is the code from Marcus Mock exam
-
respect to the horizontal. (seaplane) trim-The angle with the horizontal surface of the water assumed by the float or hull under given conditions. trimming moment-The moment about a reference point, usually the center of gravity, exerted by the
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
- L0:972 weakly excites
- L0:386 weakly suppresses
- L0:148 weakly suppresses
- L0:1513 weakly suppresses
- L0:84 weakly suppresses
- L0:1030 weakly excites
- L0:1721 weakly excites
- L0:463 weakly excites
- L0:606 weakly suppresses
- L0:97 weakly excites
- L0:1855 weakly suppresses
- L0:151 weakly suppresses
- L0:893 weakly excites
- L0:2389 weakly excites
- L0:1038 weakly suppresses
- L0:2808 weakly excites
- L0:1501 weakly excites
- L0:1915 weakly excites
- L0:1111 weakly excites
- L0:1790 weakly suppresses
- L0:490 weakly suppresses
- L0:441 weakly suppresses
- L0:2989 weakly excites
- L0:1713 weakly excites
Feeds later layers
- L2:829 suppresses
- L2:163 excites
- L2:263 excites
- L2:179 suppresses comma after a clause subject
- L25:1655 weakly excites
- L13:2141 weakly excites end-of-text token after finished articles
- L24:667 weakly excites
- L9:160 weakly suppresses
- L15:1959 weakly suppresses
- L10:492 weakly excites
- L27:1961 weakly excites
- L12:1206 weakly excites
- L20:850 weakly suppresses
- L10:144 weakly suppresses
- L24:1653 weakly excites
- L6:165 weakly excites
- L27:1664 weakly excites
- L27:2595 weakly excites
- L14:2459 weakly excites
- L27:785 weakly excites
- L10:650 weakly excites
- L4:2519 weakly excites
- L12:1485 weakly excites
- L8:613 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
- 放射
- 綦
- -secret
- SED
- Ề
- sez
- singly
- ировки
- 三位一体
- 𝑉
Suppresses
- ect
- 比分
- eur
- egot
- iger
- QUIRE
- .cgColor
- ftype
- autical
- �认