commas inside lists
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. 14 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
This neuron fires on commas that separate items in a list, usually right before the last item or before 'and'. Turning it up made the model produce loops of numbers and unrelated text, so the steering does not show a clear link to lists.
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.
-
subject area. These pocket-sized books are the perfect way to get ahead in a new subject quickly. Our expert authors combine facts, analysis, perspective, new ideas, and enthusiasm to make interesting and challenging topics highly readable.<|endoftext|>The MPI Multid
-
ades are one of the few animals to have survived all five of Earth’s major extinction events, and are known to tolerate extreme heat, cold, radiation, pressure, and the absence of food, water and oxygen.⏎In order to do this,
-
. You can’t find the right words, the right phrase, the right style, the right topic, the right approach, the right content, the right angle, the right thing to write. The cursor does what it does etymologically best
-
ricula. In addition, a bioenergy literacy search tool is available to help education professionals find and link bioenergy teaching resources like videos, datasets, images, articles, software and curricula to STEM based standards. The Imagine Tomorrow competition challenges teams of
-
eman, Duvendack, and Loubere, 2019), claims the article contains a worrying number of omissions, errors, inconsistencies, and that it also employs flawed methodologies. It lays out concerns that it legitimises
-
reminder that this product is for A SINGLE USER ONLY. That means that copying or sharing any part of this product with another teacher, grade level, school, department, district,
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 can't understand math, but I can't find something" " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " "
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
of equipment in indoor spaces that have been sealed tightly to block out cold temperatures and wind. Symptoms of carbon monoxide exposure can include everything from headaches, dizziness and drowsiness to nausea, vomiting or tightness across the chest. Severe
-
built environment around them. Ask students to think of ways that humans influence their environment (streets, houses,<|endoftext|>Food packaging is significant for a few reasons. As a matter of first importance
-
Important Part Of The Solution⏎The more we recycle, the less garbage winds up in our landfills and incineration plants. By reusing aluminum, paper, glass, plastics, and other materials, we can save production and energy costs,
-
⏎What does STEM have to do with Wall Street?⏎When you think about CAREERS on Wall Street, what usually comes to mind are Brokers, Traders, Investment Bankers, and other jobs typically filled by those with degrees in Finance and
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
- L13:2931 excites
- L13:264 excites
- L13:1789 excites
- L13:592 suppresses
- L13:2742 suppresses
- L11:116 excites
- L10:163 suppresses
- L13:749 weakly excites
- L12:1243 weakly excites
- L13:926 weakly excites
- L13:1938 weakly suppresses
- L8:1104 weakly suppresses
- L9:2726 weakly excites
- L12:2057 weakly suppresses
- L12:1235 weakly suppresses commas inside numbers and lists
- L13:199 weakly excites
- L0:1633 weakly excites
- L12:2981 weakly suppresses
- L13:767 weakly suppresses
- L10:499 weakly excites
- L13:2641 weakly suppresses
- L13:1020 weakly suppresses
- L0:2199 weakly excites
- L13:2625 weakly excites
Feeds later layers
- L15:2314 suppresses
- L16:798 suppresses
- L19:1414 suppresses
- L16:1139 excites 'and' in lists of places or names
- L16:913 excites
- L16:2327 weakly suppresses
- L16:2279 weakly suppresses
- L15:677 weakly suppresses
- L16:1809 weakly suppresses comma inside a list of names or acronyms
- L16:852 weakly excites
- L15:2979 weakly suppresses
- L20:1114 weakly excites
- L15:3020 weakly excites
- L15:2525 weakly suppresses
- L16:1163 weakly excites
- L16:2360 weakly suppresses
- L18:1207 weakly suppresses
- L20:610 weakly suppresses
- L16:2922 weakly excites
- L16:1749 weakly suppresses
- L16:634 weakly suppresses
- L15:1504 weakly excites
- L21:2676 weakly suppresses
- L18:1693 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
- prez
- (pre
- pres
- PRE
- сов
- PRES
- ewing
- (ip
- 瓜
- 三种
Suppresses
- unky
- &)⏎
- 读后
- oky
- çı
- hti
- xlsx
- 小时前
- licken
- achelor