"Since" starting a reason clause
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. 8 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
This neuron lights up on the word "Since" when it starts a sentence that explains why something is true, such as "Since they are nocturnal...". It stays quiet on most other words. Turning it up made the model's text drift into unrelated rambling about math problems, so the steering tells us little about this neuron.
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.
-
it can also cause laryngitis and bronchitis, and can trigger or worsen attacks if you happen to have asthma or other chronic lung problems. Since you do not stay completely immune to RSV for life once you have had it
-
. You can get garbage cans with locking lids. If you keep a dish of food outside for your dog during the day, bring that food inside at night. Since they are nocturnal, they will avoid light, so you can keep yard lights
-
no known negative side effects of MRI scans. The only major limitation of MRI machines are that, since they use magnets, patients with permanent metallic fixtures on their bodies, such as artificial
-
al neural networks (CNN). In recent years, CNN-based methods have gained popularity<|endoftext|>typedef unsigned int size_t;⏎This type is used to hold size definitions. Since sizes are always positives, this type is usually defined as an unsigned long, but
-
minutes, but may take more or less time. The actual duration of treatment depends on the individual case After root canal treatment, the tooth is usually painless. Since he is no longer supplied with blood, he has also become unstable. In most
-
ermilion is an opaque and heavy reddish orange pigment similar to the color Scarlet. It is more stable in oils than in other forms and brushes out well. Since the 1920's, it has largely been replaced by Cadmium
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 number of 2010-1000-2010=3000 would be increased by 1000*2010=3000, where 201
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Elimination Systems (SPDES) permit program.<|endoftext|>The opossum is an animal most of us have only seen at the zoo. After all, they are a marsupial animal, which means the females have a pouch in which to
-
930s.⏎Images: (Left) via the National Media Museum; (Right) via Getty Images<|endoftext|>Because RNA is such an important part of how the cell builds proteins, some scientists speculated that this was its
-
variety of foods such as ready meals, tinned goods and pasta sauces. Consider using natural foods, herbs and spices to flavour your food, and give your taste buds time to get used to the new flavours.⏎As we age our digestion can become
-
mutually beneficial—after all, the ants could be doing this purely to protect themselves and their colony—but it does
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
- L7:884 excites
- L7:754 suppresses
- L7:658 suppresses
- L7:336 suppresses
- L7:1183 excites
- L7:1425 suppresses
- L6:431 excites
- L7:1074 excites
- L7:1923 excites
- L7:2331 suppresses
- L7:469 excites
- L7:2514 suppresses
- L7:1112 excites
- L7:1964 excites
- L5:2239 excites
- L7:530 suppresses
- L7:61 suppresses
- L7:635 suppresses
- L7:508 excites
- L7:267 suppresses
- L7:989 suppresses
- L7:2387 suppresses
- L7:501 excites
- L6:2530 suppresses
Feeds later layers
- L13:2252 strongly excites
- L10:721 excites
- L11:1574 suppresses
- L12:2240 excites
- L16:541 suppresses
- L13:1431 suppresses
- L10:753 excites
- L13:350 excites
- L12:956 excites
- L13:1702 excites
- L11:1645 excites
- L12:1262 excites
- L12:244 excites
- L14:2116 excites
- L17:1233 excites
- L10:534 excites
- L15:1312 excites
- L12:162 excites
- L15:2499 excites
- L12:1830 excites
- L17:2403 excites
- L11:1792 excites
- L13:1130 excites
- L13:149 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
- {_
- -Americans
- {i
- ','',
- LOUD
- due
- /Users
- stu
- 排毒
- 得太
Suppresses
- nels
- pled
- narr
- ètres
- angan
- raf
- setDescription
- verdad
- 描述
- arence