unclear: product and model names
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. 10 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
It fires on the piece of a product or model name that comes right before its version number, such as the dash in MiG-15 or the word after Arduino. Its strongest negative hits are in plain tech wording like "template engine" or "spell check program". Turning it up gave rambling text with no clear theme.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
language. Octave has extensive tools for solving common numerical linear algebra problems, finding the roots of nonlinear equations, functions written in Octave's own language, or by using dynamically loaded modules written in C, C++, Fortran,
-
ouflaged MiG-15bis, serial number 2315393 (3
-
parts — a modular system developed back in the early 1950s.⏎At the heart of the machine lies a Raspberry Pi Compute Module along with Arduino Mini. The Compute Module actuates a series of motors and grippers, while also
-
that the haplotypes of the sequenced SARS-CoV-2 viruses were diverse and changed over time, but that the percentage of infected participants
-
North Atlantic air traffic. Then a young scientist working for the federal government, she’d been assigned the task following a near-miss incident involving a KLM Boeing 747, which had flown into an ash cloud over Alaska and lost all
-
race relations during the 1960s are important parts of each other’s histories. To the region’s most ardent segregationists, the Douglas DC-4s and Lockheed Electras landing from faraway cities on shimmering southern run
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 would like to use a matrix to manage the project, and I need to include a task manager, and I need to include a risk management tool. The task manager, such as Jira, and the risk management tool, such as
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
grow in complexity with time.⏎The global pattern to follow is MVC and the layering:⏎(View & Controller)⏎|View||HTML templates usually generated using a template engine|⏎|Input Controller||Filter and validate data|⏎||Application Controller||Application flow
-
the most widely used server-side scripting language⏎At time of this writing, PHP is the most used server-side language (
-
, particularly if you’re applying for college credit or other kinds of credit. You do not want your newspaper to appear grammatically challenged, therefore use a spell check program to be certain everything is correct. It may seem small, however this can be
-
. While some people consider this changes as something bad paying attention to the fact that such mass distribution deprives new albums and songs their value, the others believe that streaming helps to fight the piracy which exploded in the early 2000s
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
- L9:379 suppresses
- L5:1480 excites
- L9:1024 excites
- L8:391 suppresses
- L6:878 weakly excites
- L9:2959 weakly excites
- L9:1270 weakly excites
- L9:2346 weakly suppresses
- L9:1873 weakly excites
- L6:1506 weakly excites
- L9:1933 weakly excites
- L1:1910 weakly excites
- L7:1737 weakly suppresses
- L6:2759 weakly excites
- L8:1925 weakly excites
- L5:489 weakly excites
- L8:1307 weakly suppresses words for stretching out or extending
- L4:2018 weakly suppresses
- L7:2805 weakly suppresses
- L4:2096 weakly excites
- L8:2577 weakly suppresses
- L0:2377 weakly suppresses
- L8:2765 weakly excites
- L1:157 weakly suppresses
Feeds later layers
- L12:2096 suppresses
- L12:1184 excites
- L13:2227 suppresses
- L21:1476 suppresses
- L16:2057 suppresses
- L12:2314 suppresses
- L13:2520 suppresses
- L11:1766 suppresses
- L11:1190 suppresses
- L14:2069 excites
- L11:1385 suppresses
- L11:1937 excites
- L13:2000 suppresses
- L13:1546 suppresses
- L18:1931 suppresses
- L13:327 suppresses
- L13:1257 suppresses
- L18:2768 suppresses
- L16:2367 suppresses
- L13:809 suppresses
- L11:1221 weakly suppresses
- L11:1500 weakly excites
- L11:2525 weakly suppresses
- L13:2839 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
- phem
- бот
- neau
- \Blueprint
- ture
- อน
- SPATH
- ろ
- umph
- uario
Suppresses
- providers
- provider
- companies
- _provider
- achi
- _PROVIDER
- provider
- breaches
- providers
- 公司