the word 'form' in 'formative'
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. 3 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them 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 lights up on the 'ative' part of the word 'formative', mostly in education contexts about assessment and learning. Turning it up made the writing wander without a clear topic, so the link to the word is the main clue.
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.
-
mative Assignments⏎Formative activities and assignments provide practice and feedback to students as they work through the course.
-
learning objectives to reveal how your students are performing and learning within the course.⏎Learning objectives in the Table of Contents:⏎Learning Objectives on a course page:⏎Formative and Sum
-
before the final assessment. It also helps educators gain insight into whether their students understand the content. Rather than testing what students think they know about the topic, formative assessments allow you to determine how well students comprehend content and guide instruction based on that
-
learners is especially apparent for kindergarteners, who have not yet developed independent problem-solving skills or exhibit autonomy over their own study habits.⏎💡 Solution: Formative Assessment⏎There are various ways this can be beneficial, such as allowing students to
-
formative effect of Adventure on early IF, combined with the familiarity of many IF players and
-
nouns, as in aperture, armature, juncture, scripture, texture, fissure, pressure, etc. It is sometimes used as an English formative, as in wafture.⏎I. iii.12 (21
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: a 10-year period. I have a 10-year period. But that's not. So I have a 10-year period. I have a 10-year period. But that's not. So I have
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
the Juulnisse, which are elves that live in peoples' homes, and for Julemanden, the Danish version of Santa Claus.⏎The Finns have a tradition of resting and relaxing on Christmas Day. The night before, on
-
publishing history]⏎|20 Images Online|⏎(48 votes cast)⏎Biggles and the team set out as part of the International Squadron to help the Finns. They stumble upon a dying Pole and become involved in a tussle with
-
.newworldencyclopedia.org/entry/Karl_Marx<|endoftext|>Credit: Andrei Marincas | Shutterstock⏎Atomic Number: 86⏎Atomic Symbol:
-
74 Illingworth accompanied George Armstrong Custer’s expedition into the Black Hills, and he captured over seventy images. Illingworth was taken to court for embezzling government property when he failed to provide the agreed upon copies
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
- L0:2401 weakly suppresses
- L2:2133 weakly suppresses
- L1:2558 weakly excites
- L2:1817 weakly excites
- L2:2523 weakly excites
- L2:1328 weakly excites
- L1:1758 weakly suppresses
- L0:1224 weakly suppresses
- L0:711 weakly excites
- L1:2330 weakly excites
- L0:3058 weakly excites
- L1:2088 weakly excites
- L2:2061 weakly suppresses
- L0:1557 weakly excites
- L2:493 weakly suppresses
- L2:1321 weakly suppresses
- L1:1029 weakly excites
- L1:2253 weakly excites
- L1:2160 weakly suppresses
- L1:1674 weakly excites
- L2:1899 weakly excites
- L2:2927 weakly excites
- L2:483 weakly suppresses
- L1:2545 weakly suppresses
Feeds later layers
- L5:2799 excites
- L14:2474 weakly suppresses
- L8:439 weakly suppresses
- L10:2225 weakly suppresses
- L11:1047 weakly suppresses pieces of software and library names
- L10:2309 weakly suppresses
- L18:689 weakly suppresses
- L19:40 weakly excites
- L18:62 weakly excites
- L4:3003 weakly excites
- L4:516 weakly excites
- L23:1470 weakly excites
- L15:851 weakly suppresses
- L26:3068 weakly excites
- L21:2951 weakly excites
- L15:1360 weakly suppresses
- L22:1435 weakly suppresses
- L16:1777 weakly suppresses
- L16:2039 weakly excites
- L9:2612 weakly suppresses
- L16:498 weakly excites
- L17:2679 weakly suppresses
- L18:2868 weakly suppresses
- L10:2539 weakly excites
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
- Goa
- ache
- 落到实处
- isbury
- essen
- 잇
- قصد
- ubishi
- geme
- endDate
Suppresses
- rans
- entic
- leaps
- unnel
- 也只有
- earch
- 迷
- umph
- Brace
- istrate