'to' after 'agreed'
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. 1 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' right after the verb 'agreed' or 'agrees', as in 'agreed to exchange' or 'agrees to receive'. Turning it up did not produce an obvious theme in the sample text.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
be used in the service of scientific research.⏎Article II: Establishes freedom of scientific research and encourages cooperation in the service of science.⏎Article III: Parties agree to exchange research and program plans, scientific results, and scientific personnel, in the spirit
-
trading over futures.⏎- Futures Market⏎The bitcoin futures market is similar to that of other assets. By purchasing a security with a futures contract, a buyer agrees to receive a security at a specified price and date. By buying a futures contract,
-
–2, 3–5, 6–8, 9–12⏎Important events during the presidency of Jimmy Carter.⏎President Jimmy Carter agreed to answer seven questions selected from several hundred posted by Scholastic students.⏎A time
-
, high-level U.S. officials assumed that a Warsaw Pact conventional or nuclear attack on NATO Europe would invite a U.S. nuclear response (unless the Soviets agreed to limit fighting to conventional weapons). Rejecting the idea of "no first use,"
-
second director). On March 3, 1879, the 45th Congress and President Hayes agreed to promote governmental economy and efficiency by discontinuing the three ongoing geological and geographical surveys and
-
points mean for adaptation support? A key recommendation is the need to learn from current local level processes of change – the changing risks<|endoftext|>On Friday Australian energy ministers agreed to an investigation into the country’s energy security, stability, and affordability to be led
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 following 199700000000000000000000000000000000000000000
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
of the Romanov Dynasty⏎Role of Stalin in the 1917 Revolution⏎He was released from exile and became part of the editorial board of Pravda.⏎He was elected to the
-
is similar to use of arrowleaf balsamroot (pronghorn, bighorn sheep, and mule deer). Arrowleaf bals
-
bread; All power to the soviets.⏎He was caught up in a struggle in the Bolshevik party and called a betrayer of socialism because of the Pravda publishing support for the war effort and the provisional gouvernment.⏎He changed
-
soldiers to join the fight.⏎As the struggle against the Confederate South raged on, the newly-formed 54th Massachusetts battalion prepared to attack the seemingly impregnable fortress of Fort Wagner. The odds were heavily against them; the fort had
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
- L0:2065 weakly excites
- L0:2976 weakly suppresses
- L0:2694 weakly suppresses
- L0:1655 weakly suppresses
- L0:2680 weakly suppresses
- L0:1040 weakly suppresses
- L0:2910 weakly suppresses
- L0:2151 weakly suppresses
- L0:2365 weakly excites mid-word pieces like 'advent' and 'precaution'
- L0:554 weakly suppresses
- L0:1301 weakly suppresses
- L0:1248 weakly suppresses
- L0:1379 weakly excites
- L0:807 weakly suppresses
- L0:1822 weakly suppresses
- L0:2541 weakly excites
- L0:2702 weakly excites
- L0:1025 weakly excites
- L0:2934 weakly suppresses
- L0:2284 weakly excites
- L0:1105 weakly suppresses
- L0:2390 weakly suppresses
- L0:1987 weakly suppresses
- L0:241 weakly suppresses
Feeds later layers
- L9:2365 excites
- L6:287 excites
- L14:1818 weakly excites
- L26:697 weakly suppresses
- L5:2742 weakly excites
- L20:722 weakly suppresses
- L15:2433 weakly suppresses
- L6:1133 weakly excites "to" before an infinitive verb
- L3:3013 weakly suppresses
- L19:1665 weakly suppresses
- L16:2304 weakly excites
- L25:2496 weakly suppresses
- L8:723 weakly excites
- L18:1790 weakly excites
- L2:1000 weakly suppresses
- L19:1930 weakly suppresses
- L13:2862 weakly excites
- L22:492 weakly suppresses
- L6:2310 weakly suppresses
- L5:1022 weakly suppresses
- L5:1582 weakly suppresses
- L22:268 weakly suppresses
- L13:2025 weakly excites
- L21:1928 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
- oko
- обы
- long
- 绞
- azz
- /e
- ingt
- ulation
- ations
- thermal
Suppresses
- landers
- trailers
- lename
- adviser
- Territories
- hardest
- 录像
- 닝
- lander
- 纪录片