the 0 in decades like 1980s
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. 20 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 final zero in a decade such as 1950s or 1980s, the part that says which decade is meant. Turning it up did not produce a clear theme in the replies.
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.
-
, geographic locations are commonly identified by a postcode. This is a two-part set of letters and digits separated by a space, designed in the 1950s to assist
-
less threatening to some who might react with greater hostility to discussions focused on the actual world. Star Trek’s discussion of issues of race in the 1960s through the use of science fiction is an excellent example of this sort of approach
-
a “Helper’s High”⏎In the 1980s, researchers coined a term called the “helper’s high.” It referred to
-
of the seagrass initiative set forth in the International Decade of Ocean Exploration in the 1970s, the concerns raised by scientists at the ISBW 1 in Komin
-
Reverend Dimmesdale, and burrs to add onto the overall theme of guilt.⏎Nathaniel Hawthorne was a writer in the 1800s, an anti
-
Roma intellectuals in the 1970s and 80s. He was an assiduous participant of camps
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 different way the company is being used, and the different way the company is being valued. For instance, when we have seen, in the past, a different [insert common] [insert common] [insert common] of value,
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
UN CERD Committee is an independent body of experts that monitors participating countries’ implementation of the Convention on the Elimination of All Forms of Racial Discrimination. In 2008
-
traditionally ‘female’ industries and occupations could be equally recognised in terms of the value of its contribution. ⏎After much public debate, Australia ratified the Convention on the Elimination of All Forms of Discrimination Against Women (CEDAW), which was adopted
-
are often so excited to post<|endoftext|>Chapter 16 Risk Management for Drugs and Biologics in the United States Chapter 16RISK MAnAgEMEnT FOR DRugS AnD bIOLOgICS In THE un
-
also in helping with anxiety disorders. Side effects such as sedation, weakness, and tiredness might be a limiting factor.⏎Botulinum toxin injections might be useful for certain disabling localized tics. The effect may
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
- L19:1739 excites
- L15:1770 suppresses
- L18:1193 excites
- L19:592 excites
- L19:626 excites
- L19:1331 suppresses
- L4:1045 excites
- L16:1336 weakly suppresses
- L6:177 weakly excites
- L19:1909 weakly excites
- L15:2998 weakly excites
- L19:1744 weakly excites 'the' in mid-sentence noun phrases
- L18:1806 weakly suppresses
- L17:2556 weakly suppresses
- L9:122 weakly excites
- L8:423 weakly suppresses
- L16:243 weakly suppresses
- L19:254 weakly suppresses
- L14:815 weakly excites
- L19:2461 weakly suppresses
- L15:2146 weakly suppresses
- L2:2493 weakly suppresses
- L5:1127 weakly excites
- L17:647 weakly excites
Feeds later layers
- L27:250 strongly excites
- L26:1017 strongly excites
- L25:1685 strongly excites
- L25:813 strongly excites
- L22:2151 strongly excites
- L25:986 strongly excites
- L25:812 excites
- L24:443 excites
- L25:335 excites
- L25:614 excites
- L21:1213 excites
- L25:1963 excites
- L27:60 suppresses
- L22:533 excites
- L24:434 excites
- L27:957 excites
- L24:109 suppresses
- L23:146 excites
- L21:1400 excites
- L21:883 excites
- L25:2260 excites
- L26:856 excites
- L27:831 excites
- L26:111 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
- ies
- 们
- 们的
- ities
- ages
- es
- ations
- 們
- ologies
- ções
Suppresses
- ADVERTISEMENT
- ูล
- bite
- égorie
- ่ม
- 其他
- iento
- umba
- 霰
- gördü