Neuron Explorer

Qwen3-0.6B

'in' as a word-start fragment

A neuron in layer 25 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron responds to the letters 'in' when they start a longer word that has been split into pieces, as in 'in|coherent', 'in|rush' or 'in|humane'. Turning it down made the model get stuck repeating 'in the in the', which shows it is tied to that little 'in' piece.

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

    fighters to deter pursuit).⏎Several U.S. aircraft were shot down as the first wave approached land; one at least radioed a somewhat incoherent warning. Other warnings from ships off the harbor entrance were still being processed

  2. #2 peak

    concomitant users of warfarin and a skeletal muscle relaxant and, as pre-specified negative control comparators, among concomitant users of an in

  3. #3 peak

    n.) Hence: To carry on a process of deduction or of induction, in order to convince or to confute; to formulate and set forth propositions and the inferences from them; to argue.⏎7. (n.) To converse; to

  4. #4 peak

    20 V. Using additional external components, higher voltages can be achieved. The ADP1613 has an adjustable soft start function to prevent inrush current when the device is enabled. The pin-selectable switching frequency and PWM

  5. #5 peak

    ............... . This enables the nephrons to maintain a stable GFR even when arterial blood pressure changes.⏎Which mechanism of renal autoregulation responds to an INCREASE in BP which causes an increased STRETCH in the afferent arter

  6. #6 peak

    2) (Learn how and when to remove this template message)|⏎Sendust is a magnetic metal powder that was invented by Hakaru Masumoto at Tohoku Imperial University in Sendai, Japan, about 1936 as

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: In in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in

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

    skills pay dividends beyond academic writing. By practicing persuasive writing, students also build a solid foundation in the kind of critical thinking and reading necessary for them to be well-informed citizens.<|endoftext|>"You're a strong skater, Piet, and you have

  2. #2 peak

    forest protection by buying tracts of land, but which often fail to address the needs or rights of local Indigenous peoples. Uninhabited forest areas often go unprotected, even if purchased through a buy-an-acre

  3. #3 peak

    Fauj’ in 1943 which was initially formed in 1942 by Rash Behari Bose. He had visited London during pre-independence period to discuss the

  4. #4 peak

    harvesting and selective harvesting at high and extra-high intensities<|endoftext|>I know you are used to reading about those famous and well-known freedom fighters who existed in the pre-independence regime; the likes of Chandra Bose, Abrindra Kumar, Kh

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. L22:535 excites
  2. L24:1546 suppresses
  3. L9:2415 excites
  4. L23:1359 suppresses
  5. L23:1960 suppresses
  6. L21:2137 excites
  7. L19:1626 weakly suppresses
  8. L22:2526 weakly suppresses
  9. L17:1766 weakly suppresses
  10. L23:369 weakly excites
  11. L20:3029 weakly suppresses
  12. L21:2256 weakly suppresses
  13. L24:2514 weakly suppresses
  14. L6:2649 weakly excites
  15. L15:2931 weakly excites
  16. L17:706 weakly excites
  17. L15:3030 weakly excites
  18. L22:1112 weakly excites
  19. L24:2806 weakly suppresses
  20. L19:966 weakly excites
  21. L22:2162 weakly suppresses
  22. L18:939 weakly excites
  23. L8:1862 weakly excites
  24. L23:126 weakly excites

Feeds later layers

  1. L27:998 strongly suppresses
  2. L27:873 suppresses
  3. L27:918 suppresses
  4. L26:786 suppresses
  5. L27:1229 suppresses
  6. L27:1031 weakly suppresses
  7. L26:2107 weakly suppresses
  8. L26:2314 weakly excites
  9. L26:2078 weakly excites
  10. L27:1715 weakly excites
  11. L26:12 weakly excites
  12. L27:2701 weakly excites
  13. L26:79 weakly suppresses
  14. L27:2577 weakly excites
  15. L26:1478 weakly suppresses
  16. L26:2133 weakly excites
  17. L27:210 weakly suppresses
  18. L26:164 weakly suppresses
  19. L26:3009 weakly excites
  20. L26:1620 weakly excites
  21. L26:1189 weakly excites
  22. L27:1600 weakly suppresses
  23. L27:1523 weakly excites
  24. L27:882 weakly excites

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.

2 values between -13.86 and -12.903 values between -10.98 and -10.021 values between -10.02 and -9.061 values between -8.10 and -7.146 values between -7.14 and -6.1813 values between -6.18 and -5.2211 values between -5.22 and -4.2620 values between -4.26 and -3.3027 values between -3.30 and -2.3444 values between -2.34 and -1.38487 values between -1.38 and -0.422927 values between -0.42 and 0.54241 values between 0.54 and 1.5058 values between 1.50 and 2.4626 values between 2.46 and 3.4214 values between 3.42 and 4.385 values between 4.38 and 5.347 values between 5.34 and 6.304 values between 6.30 and 7.265 values between 7.26 and 8.223 values between 8.22 and 9.181 values between 9.18 and 10.151 values between 10.15 and 11.111 values between 23.59 and 24.55offnegativepositiveoften

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. .lwjgl
  2. acement
  3. assist
  4. 的认可
  5. 用量
  6. 助け
  7. �认
  8. roduction
  9. pNet
  10. 潢

Suppresses

  1. 包括
  2. include
  3. including
  4. includes
  5. Including
  6. Includes
  7. including
  8. Include
  9. 还包括
  10. Including