'to' in 'access to'
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 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 fires on the word 'to' when it comes after 'access', as in 'access to technology'. The activations are weak compared with the other neurons here. Turning it up gave unrelated, rambling text.
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.
-
and understanding the subject thoroughly to provide required feedback at school. What really is the hurdle here?⏎Access to technology: Not every student has internet at his home. Though it is a pretty
-
years has been how to apply the time-honored purposes of the Communications Act to our 21st Century communications platform—broadband Internet—access to which is generally provided by the same companies that provide telephone and cable television
-
Organised group of people within a larger group.⏎- Asylum - Refugee. ( Government's permission to stay in their country).⏎- Kid-gloves - To treat with extreme tact
-
a host of countries.⏎This article cannot deal with every issue on the summit agenda. What follows are a few examples of the challenges and innovations under discussion.⏎Access to Education⏎At the most fundamental level, the world has still not achieved the commitment
-
promoting the sport and achieving. American swimmers hold themselves in high standards and take every opportunity to change the otherwise monotonous sport to an exciting sport. Access to facility is the greatest obstacle but the efforts to contacting new facilities at local levels are
-
nectar.⏎|Science Recorder Pro|⏎|Free trial. No ads. Exclusive interviews. Access to all articles. Just $4.99/yr. Subscribe|⏎New research
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 Power of a Power" is a well-known, but not the most. But it's a good one. And it's a good one. I am not sure about. I am not sure about. So, I am not
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
ὔτε γὰπ τὸ χωρίον τοῦτο, ἐν ᾧ τὸ ἰερὸν φυλάττ<|endoftext|>Iron and Stone by the
-
. This should not be confused with the anniversary of the Internet, which happened in 1969 as a Defense Department project
-
⏎As the UK Parliament prepares to celebrate the 150th anniversary of its world-famous clock, E&T examines the history of Big Ben.⏎With
-
the story of Baldr's death. In Gylfaginning, Snorri Sturluson tells how Loki, fearing the wrath of the Æsir
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
- L19:858 weakly excites
- L10:90 weakly suppresses
- L27:1129 weakly suppresses
- L23:2746 weakly excites
- L16:1918 weakly excites
- L23:326 weakly suppresses
- L2:274 weakly excites
- L12:1693 weakly excites
- L22:1922 weakly suppresses
- L22:2878 weakly suppresses
- L16:1115 weakly excites
- L25:2828 weakly suppresses
- L11:2196 weakly suppresses
- L21:971 weakly suppresses
- L5:3030 weakly excites
- L12:1229 weakly suppresses
- L26:1331 weakly excites
- L12:1563 weakly suppresses
- L21:1499 weakly suppresses question marks ending a question
- L2:2842 weakly excites
- L26:973 weakly excites
- L15:1815 weakly suppresses
- L21:21 weakly excites
- L9:1276 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
- Tür
- RecognitionException
- iếu
- związane
- WebpackPlugin
- enção
- ollen
- pien
- 殂
- IMO
Suppresses
- ández
- ittest
- aces
- nghi
- acet
- idable
- Originally
- conde
- clid
- aight