unclear: mid-sentence words before a choice
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. 21 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 fires on a word that sits just before a choice or a continuation, such as 'within', 'or', 'categories' or an opening bracket. When it was turned up, the text drifted into rambling, mostly unrelated wording, so no clear theme came through.
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.
-
smooth variations which are precise to the unique paintings.⏎Sextus Empiricus' Outlines of Pyrrhonism is among the most crucial and influential texts within the historical past of Greek philosophy. within the calls for of cause Casey Perin
-
desires, the right kind method to pursue happiness, and the correct plan for an excellent lifestyles.⏎Read or Download Aristotle for Everybody: Difficult Thought Made Easy PDF⏎Best greek &
-
total footprint of around 84 acres can be selectively opened up and intensified with vegetation.⏎Right: Microforests aim to connect existing patches of vegetation<|endoftext|>Related categories 2⏎Atmospheres of Neutron Stars⏎An interactive archive of neutron
-
to abcteach⏎Free Account (Settings)⏎An 8 page study unit for Ron Roy's A-Z Mystery,
-
opportunity to develop into the individual he or she is destined to be by virtue of their own efforts.⏎Pure Fun - Yoga
-
critical areas: (1) developing understanding of and applying proportional relationships; (2) developing understanding of operations with rational numbers and working with expressions and linear equations; (3) solving problems involving scale drawings and informal geometric constructions, and working. K
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 method of the short-tangent is a method that is part of the algebra that is used to the method of determining the tangent to a function at the (e) the instant of the it is the method of determining the tangent at the
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
, it should let people know the importance of saving energy. Second, efforts should be made to develop new energy resources, such as solar energy, geothermal(地热) energy and so on. We have reason to be optimistic in this respect
-
in the Nihon Shoki 日本書紀 (720) 9 and in the Hizen-no-kuni fudoki 肥前國風土記 (first part of the 8th century) 10
-
. Because terrain shadowing effects are not considered, areas with high camera angles may also appear blurred if they are located (viewed from the camera position) behind mountain ridges or peaks.⏎i.photo.rectify can be run directly, specifying options
-
the archipelago’s land surface. The fold in the Earth’s crust that formed the Kornati mountain range was produced by the collision of the Adriatic and European tectonic plates
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
- L19:1118 excites
- L17:257 weakly suppresses
- L19:53 weakly suppresses
- L18:2998 weakly excites
- L16:461 weakly excites
- L14:392 weakly excites
- L0:0 weakly excites
- L1:2596 weakly suppresses
- L7:547 weakly suppresses
- L18:569 weakly excites
- L18:533 weakly excites
- L10:485 weakly suppresses
- L7:732 weakly excites
- L3:707 weakly suppresses
- L2:2036 weakly suppresses
- L8:328 weakly excites
- L13:91 weakly excites
- L15:313 weakly suppresses decimal point inside numbers
- L17:1652 weakly excites
- L6:8 weakly suppresses
- L3:421 weakly suppresses
- L10:157 weakly suppresses end of document and web boilerplate
- L20:417 weakly suppresses
- L16:296 weakly suppresses
Feeds later layers
- L22:2618 excites
- L23:471 weakly excites
- L24:766 weakly excites
- L27:2726 weakly excites
- L23:892 weakly excites "the" in "over the past" time spans
- L25:1263 weakly excites
- L22:2272 weakly excites end of text token before new document
- L22:2745 weakly excites
- L25:269 weakly excites
- L23:577 weakly suppresses
- L24:2563 weakly suppresses
- L24:2240 weakly suppresses
- L26:696 weakly suppresses
- L26:2055 weakly suppresses
- L23:2063 weakly suppresses
- L22:1887 weakly excites
- L22:1422 weakly excites
- L23:2671 weakly excites
- L25:117 weakly excites
- L22:1387 weakly excites
- L24:2718 weakly suppresses
- L25:2486 weakly excites
- L24:2009 weakly excites
- L25:1234 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
- 区
- .default
- 区管委会
- 县
- bled
- IPH
- default
- íveis
- иона
- //#
Suppresses
- ...'
- ệu
- chnitt
- -quarters
- ')'
- inance
- ETO
- @Api
- fuss
- .boost