"just" as an emphasis word
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. 18 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 the word "just" when it means "exactly" or "simply", as in "just that" or "just how far". It goes negative on nouns like "need" or "possibility" that come before "of". Turning it down or up made the text repeat words like "possibility" and "ability".
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.
-
skill sets they need to improve.⏎We have spent thousands of hours working with students and developing strategies to help them to succeed on the test. Our track record demonstrates just that. If you are not sure if the
-
some similarities between the two have been the source of conflict for thousands of years.⏎Both the Jewish and Muslim faith believe in one God. Being monotheistic means just that, the belief in one God. This is the most obvious similarity between the
-
protective equipment, oxygen, and treatments.⏎“Vaccines have never been the way out of this crisis on their own, but this current wave is demonstrating again just what a powerful tool they are to battle back against this virus.⏎“Delta is
-
is clear from all that is known of the<|endoftext|>Obese & Fat German Shepherd⏎Like humans, dogs can become obese for several reasons, and the effect is just as severe in them. German Shepherds are prone to becoming obese, and
-
she could be both a scientist and mom.⏎Strange to think of a time when women literally weren’t allowed to study. This fact alone definitely makes you grateful for just how far women have come since. When Rubin was elected to the National Academy of
-
the cooler climate and higher pH levels on top of Round Mountain in Saluda provide just the right habitat for seeds dropped by migrating birds.⏎“Maybe these birds are def
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 fact that the presence of the need to have the fact of the fact that to have the fact of the fact that to have the fact of the fact to the fact of the fact of the fact of the fact of the fact of the
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
06-6214.⏎Atomic Force Microscopy⏎One of the primary issues with electron microscopy techniques for the analysis of aerosol particles lies in the necessity of operating under high vacuum conditions which can fundamentally alter the properties of aerosol particles
-
on the environment, and dismissing the impacts that are known. Historically, the US military's reporting of training operations has been patchy - even activists admit the need for some secrecy - but there's reason to think the noncombat environmental impact is
-
° F.⏎- Discuss heat safety precautions with members of your household. Have a plan for wherever you spend time— home, work and school—and prepare for the possibility of power outages.⏎- Check
-
Findings suggest that young writers may overlook the possibility that a reader may have a different interpretation, or a false interpretation, even though
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
- L17:1605 suppresses
- L15:2551 suppresses
- L15:802 excites
- L16:2948 excites
- L15:835 excites
- L16:2023 weakly excites
- L17:1429 weakly suppresses
- L8:645 weakly excites
- L17:2107 weakly suppresses
- L7:1182 weakly suppresses
- L15:2899 weakly suppresses
- L15:1450 weakly suppresses
- L16:2204 weakly excites
- L17:2270 weakly excites
- L15:1483 weakly excites
- L13:1995 weakly suppresses
- L15:2708 weakly suppresses
- L7:1923 weakly excites
- L11:2719 weakly suppresses
- L16:1582 weakly excites 'it' in 'as it sounds' idioms
- L6:285 weakly suppresses
- L2:12 weakly excites
- L16:2419 weakly suppresses
- L13:1382 weakly excites
Feeds later layers
- L23:914 suppresses
- L19:2762 suppresses
- L19:2893 suppresses
- L21:2749 suppresses
- L27:1493 suppresses
- L20:233 excites
- L20:1539 suppresses
- L26:2278 suppresses
- L20:1956 suppresses
- L25:2373 suppresses
- L20:2010 suppresses
- L20:2692 suppresses
- L20:1560 suppresses
- L20:242 excites
- L20:2883 suppresses
- L19:2070 weakly suppresses
- L20:2187 weakly suppresses
- L19:1796 weakly suppresses
- L21:1441 weakly suppresses
- L21:2338 weakly suppresses
- L24:480 weakly suppresses
- L25:301 weakly suppresses
- L27:1199 weakly suppresses
- L20:2266 weakly suppresses 'the' before 'idea', 'view' or 'impression'
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
- gregate
- ends
- itu
- reviewed
- 형
- consumed
- реж
- -la
- carried
- (clicked
Suppresses
- existence
- onacci
- jednoc
- existence
- fact
- ationToken
- presence
- emploi
- unsch
- необходимости