math word problem wording
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. 15 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
It fires weakly on the wording of maths and word problems, such as 'in a geometric sequence', 'the same' or the full stop after a number in a practice question. It is quiet overall, and turning it up produced vague, story-like text rather than maths.
Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens an email to the maker with the neuron already filled in).
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.
-
make at the end of the third week?⏎3. Solve for the following.⏎a. In a geometric sequence, and . Find .⏎b. In a geometric sequence, and . Find .⏎In
-
One of the most common uses of this calculation is when purchasing a product on sale.⏎Thus, if the marked price is 1,000 and has a discount rate of 25% should take the following relationship: 1.
-
model within an Earth system model, we analysed the yield responses of six major crops to three SG technologies (SGs) and emissions reduction when they provide roughly the same reduction in radiative forcing
-
same slope, m = 2, and we know that,⏎Lines with the same slope are parallel.⏎Hence, the given lines
-
elementary problem solving series into printable worksheets:⏎Milk Street — Boston, 1706⏎Ben helped his father make 650 tallow candles. After selling some, they had 39 candles left.
-
. Supported by the Birmingham Buddhist Vihara, the pagoda underwent a large amount of planning. The circumference of the pagoda is 153 ft. and is divided into eight sections. The pagoda is in the Burmese
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: if I were a question to the person that I would be a question to the person that I would be a question to the person that I would be a question to the person that I would be a question to the person that I would be
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
of 0.02). Plug your population size and required margin of error into the formula. The result will be the number of samples you need to take.⏎For example, suppose that you have a
-
make at the end of the third week?⏎3. Solve for the following.⏎a. In a geometric sequence, and . Find .⏎b.
-
liu.<|endoftext|>Bond Price Calculator⏎With this bond price calculator, we aim to help you calculate the bond price issued by a government or a corporation. Finding out the current bond price
-
= 9 – 5.⏎Made = 650⏎Sold = ?⏎Leftover = 39⏎Leftover = Made – Sold⏎Which means also that:⏎Sold = Made – Leftover⏎Sold = 650
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
- L13:1197 excites
- L14:265 suppresses
- L14:1982 suppresses
- L14:1417 excites
- L13:2962 excites
- L12:2199 excites
- L14:2911 excites
- L12:2945 suppresses
- L11:346 weakly excites
- L12:528 weakly excites
- L14:1968 weakly suppresses
- L12:2026 weakly suppresses
- L11:143 weakly excites
- L9:1978 weakly suppresses
- L3:465 weakly excites
- L12:471 weakly suppresses
- L11:1599 weakly suppresses
- L12:268 weakly excites
- L14:368 weakly suppresses
- L11:1992 weakly excites
- L9:91 weakly excites
- L14:1053 weakly suppresses
- L8:808 weakly excites
- L13:2577 weakly excites
Feeds later layers
- L16:2999 suppresses
- L16:2846 suppresses
- L17:1220 suppresses
- L19:3020 suppresses
- L21:2920 excites
- L16:887 weakly suppresses
- L16:982 weakly suppresses
- L16:2633 weakly suppresses
- L18:763 weakly suppresses
- L19:1488 weakly suppresses
- L16:1593 weakly suppresses
- L17:2815 weakly suppresses
- L20:1996 weakly suppresses
- L17:1449 weakly excites
- L20:2260 weakly excites
- L16:2175 weakly suppresses
- L18:2259 weakly suppresses
- L19:2174 weakly suppresses
- L18:2672 weakly suppresses
- L19:2844 weakly suppresses
- L21:2227 weakly suppresses
- L17:1019 weakly suppresses
- L24:2496 weakly suppresses
- L16:2961 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
- .Hit
- raphics
- -render
- utive
- ERRUPT
- fried
- enie
- .blit
- .IS
- _sent
Suppresses
- InstanceOf
- ham
- laws
- Up
- h
- 袜
- down
- 询问
- uated
- haul