unclear: word fragments across languages
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. 21 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 fires on partial word pieces in many languages, such as Russian, Finnish and Persian, often in the middle of a word. Its pattern is not clear, and turning it up produced garbled, repeated text.
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.
-
евой колонке, с их английскими эквивалентами, данными в правой колонке⏎1) legislative branch a) Член Верховного Суда⏎2) civil and criminal cases b)
-
techniques and processes that were developed *before* the idea of genetic engineering which is the biochemical creation of specific advantageous mutations in the genome of a plant. Before the advent of GMO farmers and plant breeders had to wait for nature to create a natural
-
yväislinkit|⏎|Tila||Julkaistu - 1 syysk. 2021|⏎|OKM-julkaisutyyppi||A1 Julkaistu artikkeli, so
-
it. What this implies is that caffeine is doing something to your internal brain chemistry to make this happen – and where there’s neurochemistry, there’s a ton of crazy internal mechanisms that actually make that effect happen.⏎One can easily and efficiently
-
and the Zoological Society of London’s Institute of Zoology and published this week in the journal Nature Communications⏎The African straw-coloured fruit bat (Eidolon helv
-
garrot, gowdy, pied widgeon, whiteside, curre, and doucker. Barrow's golden-eye of America (Glaucionetta Islandica) is
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: there is a one-of-100-100-100 There is a one-of-100-100-100 probabilities that there is either of them, either of them, or
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
(i.e. Varro's description of the extent of Roma quadrata) mit dem unten erörterten Pomerium und der Auffassung des Dionys. I 88: περιγράφ
-
Antalya-Türkiye’de doğdu. Lisans öğrenimini MSÜ Deniz Harp Okulunda tamamlamıştır.⏎Yüksek Lisans Öğrenim
-
ментарии при наличии в них нецензурной брани, оскорблений или спама<|endoftext|>Conflict and Solid
-
их английскими эквивалентами, данными в правой колонке⏎1) legislative branch a) Член Верховного Суда⏎2) civil and criminal cases b) окружной суд⏎3
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
- L18:893 excites
- L16:864 excites
- L18:2342 excites
- L18:2676 excites
- L20:2873 suppresses
- L19:939 suppresses
- L18:550 suppresses
- L19:2507 excites
- L17:500 suppresses
- L18:491 excites
- L16:1231 excites
- L7:122 excites
- L7:76 excites
- L20:2800 suppresses
- L20:789 excites
- L20:1796 excites
- L19:2880 excites
- L19:1025 suppresses
- L20:1356 suppresses
- L19:1584 excites
- L19:1581 suppresses
- L19:2640 weakly excites
- L19:3044 weakly excites
- L20:677 weakly excites
Feeds later layers
- L22:2745 strongly excites
- L27:2096 strongly suppresses
- L24:359 strongly suppresses
- L26:1320 suppresses
- L25:2973 suppresses
- L23:1390 excites
- L27:1582 excites
- L23:258 suppresses
- L23:2526 excites
- L27:2605 excites
- L25:284 suppresses
- L26:711 suppresses 'years' after an age number
- L22:142 excites 'of' after a quantity or noun phrase
- L22:171 excites
- L27:1260 weakly suppresses
- L26:2643 weakly excites
- L27:1737 weakly suppresses
- L22:1425 weakly excites
- L22:1523 weakly suppresses
- L22:309 weakly excites
- L27:1764 weakly suppresses
- L22:1993 weakly suppresses
- L25:917 weakly suppresses
- L26:507 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
- يون
- ость
- voie
- ные
- reek
- такие
- деятельность
- تان
- жение
- różne
Suppresses
- ального
- енного
- ului
- ического
- arters
- ителей
- емых
- ющей
- альных
- телей