unclear: word right after a modifier
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. 0 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
The neuron tends to fire on a word that comes right after another word that describes it, as in 'joint pain', 'wind speed' or 'ground stations'. It is weakly negative on the '&' or 'of' inside organisation names. Turning it up or down did not produce a clear theme in the text that followed.
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.
-
Illinois Learning Standards prepare students for the challenges of. The Core Knowledge Preschool Sequence and Teacher Handbook provides
-
Six Jewish musicians from the philharmonic were murdered by the Nazis in Austria and 11
-
- Trouble concentrating⏎- Joint pain and stiffness⏎- Muscle cramps⏎- Impaired memory⏎- Can’t
-
a typhoon if it happened in the northwestern Pacific.⏎WHAT’S THE DIFFERENCE BETWEEN A HURRICANE AND A TROPICAL STORM?⏎Wind speed. When tropical cyclones are just starting out as general areas of low pressure with
-
MinneapolisSaint Paul metropolitan place (known as the “Twin Cities”). This place has the largest immersion
-
coastal dishes, oyster sauce, fish sauce and Clamato juice⏎Seafood, Molluscs, Crustaceans<|endoftext|>WikipediaRead full entry⏎Mitsukurina is a genus of mackerel shark in
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: this is A the this is A the This IS A THIS is A this IS A this IS a this IS A. this IS A this IS a This IS a this IS A
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
conducted at the Oregon National Primate Research Center at Oregon Health & Science University, monitored the reproductive systems of healthy female nonhuman primates following exposure
-
San Tan Allergy & Asthma can help determine whether your (or your child’s) symptoms may be
-
. The Sisters of Notre Dame de Namur are becoming more intentionally aware of this responsibility in communities and
-
see a lot of diabetes in my clients." exclaims Dr. William H. Wilson, Professor of Psychiatry and Director of Inpatient Psychiatric Services at Oregon Health & Science University.⏎A simple statement that means so much. Considering that Dr. Wilson
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
None: this is the first layer.
Feeds later layers
- L4:176 strongly excites
- L2:381 excites
- L3:169 suppresses
- L1:71 suppresses
- L10:516 excites
- L5:1286 excites
- L7:524 suppresses
- L16:482 suppresses
- L18:2369 suppresses
- L3:455 excites unclear: mid-name word fragments
- L16:614 excites
- L2:3 suppresses
- L1:762 suppresses
- L20:360 excites
- L2:1445 excites
- L20:474 suppresses
- L1:1154 suppresses
- L1:122 suppresses 'World War II' and the 'II' after 'War'
- L22:315 suppresses
- L17:513 suppresses
- L16:158 suppresses
- L17:1993 suppresses
- L19:903 excites
- L3:137 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
- ellung
- urga
- 十三五
- 担
- 惠民
- bor
- 握
- CLU
- erca
- ówn
Suppresses
- isposable
- gone
- _None
- Signature
- Signature
- writes
- recht
- yses
- reeze