unclear: 'lesson' and 'lique' word pieces
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. 2 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 seems to fire on the start of words like 'Lessons' and 'liquefied', and on pieces of technical words. The examples do not share a clear theme, so the pattern is uncertain. Turning it up produced rambling text about temperatures and sums.
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.
-
energy needs. For the majority of off-grid operations, this means a choice between oil, LPG (liquefied petroleum gas) or LNG (liquefied natural gas) for high
-
of the UNLIKE.⏎Sir M. FOSTER, K.C. B., M. D., and LEWIS E. SHOKÉ,⏎LESSONS IN ELEMENTARY PHY-⏎PHYSIOLOGY. By Sir MICHAEL F
-
, the institution of Marriage — and how these wrong decisions can be avoided and remedied. Sign up now while there is still space in this special course!⏎Lessons cover issues such as the acknowledgment of God, Marriage, Life, Obscenity
-
(non-polarized), inexpensive, and durable. Avoid capacitors with any kind of polarity marking, as these will be destroyed when powered by AC!⏎Lessons In Electric Circuits, Volume 2, chapter 1: "Basic AC
-
the Arduino. It can be found here at arduino.cc for Mac, Linux, and PC.⏎One cannot play the Can Can without finesse!<|endoftext|>Lessons From the Past The following i⏎The following is an excerpt from the recommended standards
-
0 seconds of the French folksong "Au Clair De La Lune" on April 9, 1860.⏎Édouard-Léon Scott de Martinville recorded the voice by using a "phonautograph" to
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 patient has been. I am the patient. I am the patient. I am the patient. I am the patient. I am the patient. I am the patient. I am the patient. I am the patient.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
why: Doing so stimulates the brain the same way sex, food, and cash do. Researchers scanned subjects' brains and found that parts linked to those pleasures lit up when subjects chatted about themselves.⏎Researchers also offered study subjects up to 4
-
BC to AD 500, consisting of thousands of large stone jars placed on hills within the central plain of the Xiangkhoang Plateau in Laos.⏎According to local Lao legend, the jars were created by a race of
-
180 - 197.⏎Kwong-loi Shun (1997). Mencius and Early Chinese Thought.
-
:⏎Unit Objectives: At the completion of this unit the⏎Chapter Fifteen Summary⏎The police frequently use three procedures to help by witnesses identify criminals: lineups, show-up, and photographic lineups. Lineups involve showing the witness a
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
- L1:2755 weakly suppresses
- L1:1899 weakly excites
- L0:1934 weakly suppresses
- L1:1298 weakly suppresses
- L1:1659 weakly excites
- L1:1870 weakly suppresses
- L1:3038 weakly excites
- L1:1583 weakly excites
- L1:1099 weakly excites
- L0:432 weakly excites
- L1:1973 weakly suppresses
- L0:1123 weakly suppresses
- L0:178 weakly suppresses
- L0:2787 weakly excites
- L1:2506 weakly excites
- L0:3058 weakly excites
- L0:1224 weakly suppresses
- L0:1301 weakly suppresses
- L0:1644 weakly suppresses
- L0:960 weakly suppresses
- L1:1953 weakly suppresses
- L0:53 weakly excites
- L1:2330 weakly suppresses
- L0:42 weakly excites
Feeds later layers
- L21:252 excites
- L5:1294 excites
- L18:198 excites
- L19:593 excites
- L26:2235 weakly excites word-start 'l' after a space
- L10:904 weakly suppresses
- L3:1929 weakly suppresses
- L16:671 weakly excites
- L18:331 weakly excites
- L3:2309 weakly suppresses
- L27:1748 weakly suppresses
- L5:2321 weakly excites
- L4:2870 weakly excites
- L21:2568 weakly excites
- L3:2040 weakly excites
- L3:1036 weakly suppresses
- L27:1934 weakly suppresses
- L24:2459 weakly suppresses
- L26:2668 weakly excites
- L4:1762 weakly excites "this" in time and context phrases
- L3:2216 weakly excites
- L5:1093 weakly excites
- L10:273 weakly excites
- L6:2766 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
- ={$
- absorbs
- cope
- ═
- TEE
- WC
- راق
- >{$
- parole
- ousel
Suppresses
- elman
- roud
- azz
- builders
- rawing
- .styleable
- 'B
- LOGGER
- 募集
- DrawerToggle