the digit 1 starting a year
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. 24 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 first digit '1' of a year, such as the 1 in 1951 or 1846, when the model is about to finish writing the year. Turned down, it produced strings of equals signs; turned up, the text drifted without a clear theme.
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.
-
castes other than the Scheduled Castes (SCs) and the Scheduled Tribes (STs) have not been enumerated in any of the Census since 1951 till day”. The submission was in response to a plea by the
-
a small telescope and recognized as a planet through its motion among the neighbor stars. Neptune is far too faint to be seen without a telescope. Its discovery in 1846 is the result of mathematical computations on the motion of Uranus whose
-
the other hand, take the case of the now less remembered Royal Indian Navy mutiny (also called the Bombay mutiny) of February 18, 1946. Over 10,000 Indian sailors across
-
, the Ministry of Environment, Forest and Climate Change has come up with the first draft of the amendments to the Indian Forest Act, 1927. The amendments have been laid down with the primary objective of abol
-
alternative to lawsuits. As a result,<|endoftext|>10 / The Natural World⏎|Artist / Origin||⏎Albrecht Altdorfer (German, 1480–1538)⏎Period: 1400
-
signing of the Treaty of Paris in 1763.⏎Colonial Military Policy⏎Understanding the need for military discipline,
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: = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
secretpassword';⏎mysql> GRANT ALL PRIVILEGES ON database_name.* TO user_name@localhost;⏎mysql> FLUSH PRIVILEGES;⏎Step 5: Installation and Configuration of Drupal⏎Open
-
largest wetland in the North Island and is recognised as a Ramsar site – a wetland of international importance.⏎Ō Tū WharekaiCANTERBURY Ō Tū Whare
-
, toward positive holistic wellbeing.⏎Strengthen, support and guide stakeholders, to further advance the self-determined education aspirations, achievements and successes for Kahungunu ākonga, whānau, hapū and iwi.⏎Design
-
cut down on fuel needs and overall costs, which would have a huge impact on the feasibility of manned deep space missions. Not to mention there would likely be immeasurable psychological benefits; although it is difficult to conduct adequate psychological study of astronauts,
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
- L21:253 excites
- L23:1421 excites
- L21:622 excites
- L19:2749 excites
- L20:1098 suppresses
- L20:1985 suppresses
- L20:747 suppresses
- L21:554 excites
- L14:2877 suppresses variables and symbols in formulas
- L16:2751 suppresses
- L23:970 suppresses
- L19:1711 excites
- L18:186 excites
- L19:2671 excites
- L16:456 excites
- L19:1682 weakly suppresses
- L19:668 weakly excites
- L20:270 weakly excites
- L18:465 weakly excites
- L14:89 weakly suppresses
- L17:2012 weakly excites
- L20:772 weakly suppresses
- L23:256 weakly excites
- L3:78 weakly suppresses
Feeds later layers
- L26:606 strongly suppresses
- L27:390 suppresses
- L27:747 suppresses
- L25:798 suppresses
- L26:605 suppresses
- L26:1231 suppresses
- L25:252 suppresses
- L26:499 weakly excites
- L25:17 weakly excites
- L26:48 weakly excites
- L27:398 weakly suppresses
- L25:1898 weakly suppresses
- L27:1589 weakly suppresses
- L26:867 weakly suppresses
- L25:2308 weakly suppresses
- L26:21 weakly suppresses
- L25:1388 weakly suppresses
- L26:2576 weakly suppresses
- L26:227 weakly excites
- L26:1954 weakly excites
- L26:2684 weakly suppresses
- L25:681 weakly suppresses
- L25:2844 weakly suppresses
- L27:1649 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
- finally
- 一些
- ,'"
- 馆
- ,readonly
- ,"
- ,"
- 不至于
- 有时
- ####
Suppresses
- =
- ']=
- ={$
- =");⏎
- =
- =⏎
- =",
- |=
- =$
- = ⏎