unclear: 'purpose' and words near it
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. 12 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 is hard to pin down. It fires on words like 'secure', 'support' and 'found' in academic and descriptive text, and it goes strongly negative on 'purpose' and the word 'be' right before it. Turning it up or down did not show a clear theme.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. Think it is wrong? Flag it (opens an email to the maker with the neuron already filled in).
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.
-
railroad, Sacramento quickly saw its population swell in the second half of the 19th century as blue-collar laborers poured into the city in droves to secure work as machinists, painters, carpenters and boilermakers for
-
aries). But they differ more than they do resemble each other, when one looks at the details. Comparing them can give us insight into what the Framers found important in 1781, and what they changed their minds on by
-
Discussion Area. You should quote or paraphrase from specific pages in the course text to support your claims about philosophical ideas and concepts, using correct APA citation. Start reviewing and
-
urative Language Practice Cards provide an easy way to help students develop a critical reading comprehension skill. Each card consists of a leveled passage that contains a form of figurative language. Questions ask students to choose
-
the foreign good since it is not available locally, however, they will reluctantly do so.⏎Based on the general theory of ethnocentrism, consumer ethnocentrism is specific to consumers of a particular country. Ethnocentrism refers to
-
on where and how the curvature occurs.⏎Kyphosis (also called kyphoscoliosis) is a deep front-to-back curve typically occurring at
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 have been working on this task, and I have been working on this task. I have been working on this task. I have been working on this task. I have been working on this task. I have been working on this task
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
into laboratory kits that are complete, well-organised, and properly targeted for temperature-controlled shipping and storage.⏎First, the purpose of the kit or kits must be firmly defined. This is a critical stage, as imprecision in
-
in Houston, Texas.⏎Chartered in 1958 by the Secretary of Defense, DDMS was originally formed with the express purpose of providing much needed DoD support to the U.S. initial manned space flight
-
, a nine-year venture that existed between 1817 and 1826, whose purpose was to educate and train non-white Christian missionaries. The book focuses upon the evangelical
-
saying Social Security and Medicare are bad? The fact of the matter is these programs (along with welfare) were created to do only one thing, take money away from the people who have it, and give it
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
- L8:1890 suppresses
- L9:2504 excites
- L8:2139 suppresses
- L11:2555 weakly suppresses
- L9:200 weakly suppresses
- L9:1505 weakly excites
- L10:2075 weakly excites
- L11:2819 weakly suppresses
- L8:2827 weakly excites
- L7:2710 weakly suppresses
- L9:2710 weakly excites
- L9:1160 weakly excites
- L7:2106 weakly suppresses
- L9:2451 weakly excites 'for' giving a reason or cause
- L3:2607 weakly excites
- L11:1790 weakly suppresses
- L9:1000 weakly excites
- L11:2620 weakly suppresses
- L10:2507 weakly excites
- L10:1242 weakly suppresses
- L7:2843 weakly suppresses
- L4:1490 weakly suppresses
- L2:2906 weakly excites
- L11:255 weakly excites
Feeds later layers
- L14:522 excites
- L17:1097 suppresses
- L16:2424 suppresses
- L18:2659 suppresses
- L15:515 suppresses word before 'that', 'to' or 'for' continuation
- L17:1670 suppresses
- L22:2856 suppresses
- L14:1306 excites
- L15:1887 suppresses
- L18:2469 suppresses
- L19:2827 suppresses
- L17:2014 excites
- L16:1449 suppresses
- L19:597 suppresses
- L19:1912 suppresses
- L15:672 weakly suppresses
- L22:2793 weakly suppresses
- L27:2279 weakly suppresses
- L13:1835 weakly suppresses
- L20:1861 weakly suppresses
- L19:1110 weakly excites
- L15:2174 weakly suppresses
- L20:928 weakly excites
- L19:2993 weakly suppresses
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
- delay
- erview
- ahoma
- imes
- urrence
- ippet
- stagger
- permalink
- langue
- krit
Suppresses
- purpose
- 目的
- Purpose
- Purpose
- purposes
- task
- 用途
- tasks
- 目标任务
- (tasks