'make' before decisions or predictions
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 fires on the word 'make' when it leads into a decision, a prediction or a choice, as in 'to make informed decisions'. When it was turned up, the text drifted toward talk about making decisions.
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.
-
NOAA’s Fisheries Office of Aquaculture. “It is a tool that can be used when evaluating proposed or operational farming sites and gives them a factual basis to make decisions.”⏎In the Executive Summary of the report, NOAA scientists say there is an
-
driving cars.⏎One of the most exciting aspects of AI is its ability to learn and adapt. Machine learning algorithms can analyze vast amounts of data to identify patterns and make predictions, and they improve with experience. This capability has applications
-
decisions and take responsibility for their own financial well-being. Faced with an array of complex financial services and sophisticated products, many graduates lack the knowledge and skills to make rational, informed decisions on the use of their money and planning for future events
-
technology to achieve an ultraprecise lunar landing.⏎Navigation GUI, on the other hand, assists in human-machine interactions, making it possible for ground operators to make quick decisions during the descent process. It is a novel yet straightforward
-
any decision.⏎Keeps You Well-informed⏎If you consider all the facts before making any decision, then you will be well-informed of all the consequences of your decision
-
an infection.⏎Cunningham Panel™ Assists Clinicians in the Diagnosis of PANS and PANDAS⏎The Cunningham Panel of tests can assist clinicians in making the correct diagnosis. The Panel provides insight into how
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 one that I can't do, the one that I can't do. So I'm going to have to say something... So, the answer is 2, which is 22, and 22 is 2
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
with pre-existing anemia mustn't be given garlic.⏎- Dr's Foster and Smith of peteducation.com are in agreement that garlic causes a breakdown of red blood cells in dogs though they permit that garlic in terribly tiny amounts in some industrial pet
-
by sitting down to a hot breakfast. You need to consider what you’re eating. Focus on foods that have nutrients such as calcium, B vitamins, fibre, iron and folate so that you
-
provide about 20-35% of your guideline daily calorie allowance (GDA).⏎Apart from providing energy, breakfast foods are good sources of calcium, iron and B vitamins as well as protein and fibre. Fruit and vegetables are good sources
-
it is the building block of all of our cells – and therefore our muscles, tissues and organs. And it is a tremendous source of vitamin B12 and iron. Due to its high
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
- L4:1785 suppresses
- L4:2528 suppresses
- L3:1669 excites
- L6:1776 weakly suppresses street and urban place names
- L0:748 weakly excites
- L5:1508 weakly excites
- L7:1207 weakly excites
- L6:2633 weakly suppresses
- L0:2721 weakly excites
- L6:784 weakly suppresses
- L4:1419 weakly excites
- L2:2593 weakly suppresses
- L6:3007 weakly excites
- L3:2422 weakly excites
- L5:2079 weakly suppresses
- L2:119 weakly excites
- L5:521 weakly excites
- L1:560 weakly excites
- L2:777 weakly excites
- L7:2126 weakly excites
- L1:1226 weakly suppresses
- L4:2968 weakly excites
- L2:2423 weakly suppresses
- L0:1520 weakly excites
Feeds later layers
- L20:1177 excites
- L16:2743 excites
- L22:587 excites
- L19:2306 excites
- L10:2087 weakly excites
- L13:2628 weakly excites
- L19:2406 weakly excites
- L23:2252 weakly excites
- L23:1187 weakly excites
- L18:1468 weakly excites
- L10:1492 weakly excites
- L25:1294 weakly excites
- L22:361 weakly excites
- L24:9 weakly excites
- L17:3053 weakly excites
- L21:1494 weakly excites
- L20:1581 weakly excites
- L20:2406 weakly suppresses
- L14:2680 weakly excites
- L24:1162 weakly excites
- L22:487 weakly excites
- L23:471 weakly excites
- L23:2909 weakly excites
- L10:2761 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
- precaution
- oise
- discoveries
- ylie
- 抉择
- speculation
- acb
- _callbacks
- _decision
- decisions
Suppresses
- time
- 化
- emia
- 化的
- -time
- piring
- 媚
- erg
- -Time
- ten