Neuron Explorer

Qwen3-0.6B

technical word fragments, splicing and isotopes

A neuron in layer 6 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron reacts to parts of longer technical words, like 'splicing' in biology, 'isotopes' in physics, or 'allotropes' in chemistry. Turned up, the writing stayed vague and drifted into numbers and circuits.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    ⏎on the Hidden Markov Model of probabilities (HMMgene) also can locate donor and acceptor sites.⏎- Splice variants: Alternative splicing, resulting in different combination of exons, has been found to occur⏎in

  2. #2 peak

    carbyne, an infinite sp1 hybridized carbon chain. As a result, much less has been explored about carbyne as compared to other carbon allotropes such as fullerenes, nanotubes and graphene. Although end-c

  3. #3 peak

    disrupted, an alarm will sound. It is most effective to warn about a smoldering fire. Ionization smoke detectors contain a very small amount of americium-241 within an ionization chamber. They create an electric current between two

  4. #4 peak

    device that at both ends converts gases of deuterium and tritium (isotopes of hydrogen) into plasmoids - plasma contained by a magnetic field through

  5. #5 peak

    ⏎is derived from internal sequences called "exons" which alternate with the introns in the longer, nuclear mRNA.⏎(Figure of splicing mechanisms)⏎Splicing of nuclear mRNA is catalyzed by specialized RNA-protein complexes called small nuclear

  6. #6 peak

    pressure processes which are descriptive of "ballotechnic" reactions, i.e. shock- induced reactions in heterogeneous ma- terial mixtures.⏎- ISOTOPIC SEPARATION AND ENRICHMENT⏎- “ballotechnic”

Steering

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.

turned down a lot
Prompt

Here is something I have been thinking about lately: analog audio, 12 to 12 ohm, and this is the audio output for a speaker. You can use a 12 ohm resistor to connect to a 12 ohm resistor directly. This is a

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

  1. #1 peak

    4. Why is data bus bi-directional?⏎5. What is the function of accumulator?⏎6. What is flag, bus?⏎7. What are tri-state devices and why they are essential in a bus oriented system?⏎8. Why are

  2. #2 peak

    with 1/4" TS and TRS connectors, XLR connectors, and other plugs and jacks. Headphone connectors and other 1/8" plugs are also TRS connectors, where the two signals share a common ground connection

  3. #3 peak

    .6µm Bipolar Complementary Metal Oxide Semiconductor, or BiCMOS, is typically implemented as dual NAND gates with coupled inputs and a BJT driving the output. BiCMOS is an adaptation of older CMOS technologies, and

  4. #4 peak

    video about signal flow,…there are three main categories…of the strength or level…of the electrical analog signal.…⏎Mic level, line level, and speaker level.…Mic level is the weakest of the three.…This is usually somewhere around…

Connections

Show on map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

  1. L4:2392 excites
  2. L4:1478 suppresses
  3. L4:1968 excites
  4. L2:1670 weakly excites
  5. L1:1340 weakly suppresses
  6. L5:46 weakly excites
  7. L4:2941 weakly excites
  8. L4:1066 weakly excites
  9. L4:2254 weakly suppresses
  10. L0:2583 weakly suppresses
  11. L5:1462 weakly excites
  12. L3:322 weakly excites
  13. L0:2464 weakly suppresses
  14. L5:707 weakly suppresses
  15. L1:2922 weakly excites
  16. L4:112 weakly suppresses
  17. L5:216 weakly suppresses
  18. L5:1123 weakly excites
  19. L2:2328 weakly suppresses
  20. L4:1174 weakly excites
  21. L5:169 weakly excites
  22. L5:341 weakly suppresses
  23. L3:1065 weakly excites
  24. L4:1560 weakly excites

Feeds later layers

  1. L7:2495 suppresses
  2. L9:1337 suppresses
  3. L8:2474 suppresses
  4. L8:2022 suppresses
  5. L8:1600 weakly excites
  6. L7:2271 weakly suppresses
  7. L7:1252 weakly suppresses
  8. L7:2770 weakly suppresses
  9. L22:1275 weakly suppresses
  10. L8:2849 weakly excites
  11. L7:2883 weakly suppresses
  12. L8:1358 weakly suppresses
  13. L11:688 weakly suppresses
  14. L7:565 weakly excites
  15. L26:1648 weakly suppresses
  16. L16:1186 weakly suppresses
  17. L23:752 weakly excites
  18. L8:1369 weakly suppresses
  19. L9:2035 weakly suppresses
  20. L8:2464 weakly suppresses
  21. L23:1243 weakly excites the spot right before a year's last digits
  22. L9:1605 weakly suppresses
  23. L26:2377 weakly suppresses
  24. L24:1019 weakly suppresses

Activation histogram

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.

1 values between -9.16 and -8.881 values between -3.40 and -3.112 values between -2.53 and -2.252 values between -1.38 and -1.091 values between -1.09 and -0.806 values between -0.80 and -0.52146 values between -0.52 and -0.232375 values between -0.23 and 0.061302 values between 0.06 and 0.3565 values between 0.35 and 0.643 values between 0.64 and 0.931 values between 0.93 and 1.211 values between 1.21 and 1.501 values between 1.79 and 2.081 values between 2.08 and 2.37offnegativepositiveoften

Logit lens

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 layers instead.

Promotes

  1. حة
  2. SPELL
  3. пу
  4. SpringApplication
  5. теля
  6. quarry
  7. 疾
  8. зыва
  9. جم
  10. Commerce

Suppresses

  1. يان
  2. ieee
  3. -output
  4. ewire
  5. ��态
  6. ckett
  7. wires
  8. 絜
  9. ası
  10. illi