unclear: weak scattered triggers, finance on negative side
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. 6 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
This neuron responds only weakly and to a mixed bag of words, such as years in dates and the word 'find'. Its strongest negative reactions come on money and bank-account words, and turning it down pushed the writing toward finance questions about interest and savings.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
your identity safe.⏎- Only Make Purchases On Trusted Sites⏎When deals seem too good to be true, they just may be - you might be paying as a victim of identity theft when you make purchases on websites that aren't secure. There
-
Collectors’ Circle, Jim and Ann Ach ’65, Peggy Phelps, Jane Hurley Wilson ’64 and Michael G. Wilson, and the Getty Foundation.⏎Dani
-
Atoll) martix. It is an extension of Bullseye matrix problem.⏎Who has power to do everything in this world?⏎Close MATLAB with keyboard⏎Find the treasures in MATLAB Central and discover how the community can help you!⏎Choose a
-
Jan 09, 2017 08:57 AM EST⏎MIT Professor Is Building Bio
-
enter a 13.5 out of 15 into a calculator. The calculator will do the rest, including converting the decimal to a percentage.⏎To find the percentage value of a mark, you’ll need to know its total mark.
-
72,000 km/hour (45,000 miles/hour) when it hit the atmosphere.
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: I have a 30-year loan. The interest rate is 5.5% and the number of periods is 4. I have a 10-year loan with a 5% interest rate. What is the future value
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
than others to save for.⏎Decide how much money will go into each account when your child receives money (either through an allowance, gifts, or paid work
-
bond price calculator:⏎Face value– The coupon's face value;⏎Coupon rate– The annual coupon rate;⏎Frequency– The coupon frequency, i.e., the number of times the coupon is distributed in a year;⏎n–
-
sums of money being spent to protect against them. The image of some nefarious character plotting in
-
prepare Cash Budget⏎• Understand and apply Profit Planning concepts to prepare Pro-forma Statement⏎Capital Budgeting & Investment Analysis⏎• Understand and apply time value of money concepts to evaluate different streams of cash flow⏎• Describe the characteristics of
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
- L3:1446 excites
- L3:2577 weakly suppresses
- L5:60 weakly excites
- L4:182 weakly suppresses
- L4:12 weakly suppresses
- L5:941 weakly excites
- L1:2796 weakly suppresses
- L4:1305 weakly excites
- L4:23 weakly excites
- L5:69 weakly excites
- L1:1973 weakly suppresses
- L4:1937 weakly suppresses
- L5:2932 weakly suppresses
- L4:245 weakly excites
- L5:1176 weakly excites
- L4:2730 weakly suppresses
- L3:231 weakly excites
- L3:59 weakly excites
- L4:2606 weakly suppresses
- L1:160 weakly suppresses
- L5:222 weakly suppresses
- L5:158 weakly suppresses
- L5:2128 weakly excites
- L2:2947 weakly suppresses
Feeds later layers
- L9:2045 suppresses
- L22:2624 suppresses
- L9:2793 suppresses
- L7:2666 suppresses 'a' in 'once a week' frequency
- L7:1062 suppresses
- L7:2466 suppresses
- L18:2554 suppresses
- L20:1586 suppresses
- L20:630 suppresses
- L7:1832 excites
- L21:2047 suppresses
- L10:2273 suppresses
- L7:1293 suppresses
- L7:1029 excites
- L7:3054 excites
- L7:565 suppresses
- L21:873 suppresses
- L7:2140 excites
- L24:2372 suppresses
- L20:700 suppresses
- L20:1700 suppresses
- L9:996 suppresses
- L8:2170 suppresses
- L22:408 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
- verg
- erva
- yl
- +xml
- unta
- erv
- جاز
- pole
- acons
- nü
Suppresses
- 账户
- buch
- 分期
- 源泉
- 受益
- utschen
- odings
- 偿还
- 还款
- 人生