Neuron Explorer

Qwen3-0.6B

'time' in 'at the same time'

A neuron in layer 2 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 fires on the word 'time' in the phrase 'at the same time'. Its weak opposite side is the word 'Summary' in 'Executive Summary'. Turning it up did not produce text that clearly reflects the phrase.

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

    are also clearly harmful species. At the same time, the danger posed by non-native species as a group often has been exaggerated and

  2. #2 peak

    within the U.S. Department of Labor, workplace fatalities have been cut in half and occupational injury and illness rates have declined 40 percent. At the same time, U.S. employment has nearly doubled.⏎Under the OSH Act, employers

  3. #3 peak

    . At the same time, it manages the communication network of the muscles in the waist, leg and foot

  4. #4 peak

    . At the same time, the perceptions expressed by respondents suggest they are becoming more vulnerable to shocks and stress

  5. #5 peak

    the scavenge air amount, and thereby an increase of the exhaust gas temperature level, which leads to reduced efficiency of the diesel engine itself, at the same time creates a remarkable potential for increased power output on both power turbines and steam turbines.

  6. #6 peak

    precincts) was to provide local seats of government where citizens could record documents, such as deeds or wills, and participate in court proceedings. At the same time, the sheriff was provided with a home base from which to fulfill his basic responsibilities

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: I have been thinking about lately, and the first thing to do The first thing to do is a big box. The second to do The third to do The fourth to do The fifth to do The first thing

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

    998. Name: South Korea: Restoration of the Cheonggyecheon River in Downtown Seoul: Executive Summary: Since the 1970<|endoftext|>CS610 Handouts pdf

  2. #2 peak

    aculture. “It is a tool that can be used when evaluating proposed or operational farming sites and gives them a factual basis to make decisions.”⏎In the Executive Summary of the report, NOAA scientists say there is an opportunity to apply the best available

  3. #3 peak

    avirus (p = 0.008, OR<|endoftext|>A Global Index of Sustainable Well-Being⏎Source: The New Economics Foundation⏎From the Executive Summary:⏎There is a growing global consensus that we need new measures of progress. It

  4. #4 peak

    bad influences have moved into a new battlefield– the Internet. The latest studies show that people, especially teenagers, who frequently visit their social media account increase the chances of participating in unwanted habits, such as smoking

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. L0:1301 weakly suppresses
  2. L0:2745 weakly excites
  3. L0:178 weakly suppresses
  4. L0:1224 weakly suppresses
  5. L1:121 weakly suppresses
  6. L1:1973 weakly suppresses
  7. L0:351 weakly excites
  8. L0:285 weakly suppresses
  9. L1:2230 weakly excites
  10. L1:925 weakly suppresses
  11. L1:1712 weakly excites
  12. L1:2486 weakly excites
  13. L0:1774 weakly suppresses
  14. L1:1985 weakly suppresses
  15. L1:272 weakly excites
  16. L1:2877 weakly suppresses
  17. L0:476 weakly suppresses
  18. L0:3010 weakly suppresses
  19. L1:1249 weakly excites
  20. L0:589 weakly excites
  21. L1:2047 weakly excites
  22. L1:1328 weakly excites
  23. L1:243 weakly suppresses
  24. L1:130 weakly excites

Feeds later layers

  1. L4:202 excites
  2. L9:605 excites
  3. L7:389 excites
  4. L4:194 excites
  5. L17:1455 excites
  6. L4:183 excites
  7. L18:2912 suppresses
  8. L3:1093 weakly excites
  9. L15:410 weakly excites
  10. L9:497 weakly excites
  11. L15:2722 weakly suppresses "and" or "or" in a pair or list
  12. L3:452 weakly excites
  13. L15:385 weakly suppresses
  14. L8:2671 weakly suppresses
  15. L9:2140 weakly excites
  16. L3:1296 weakly suppresses
  17. L5:238 weakly suppresses
  18. L7:851 weakly suppresses
  19. L15:325 weakly suppresses
  20. L6:968 weakly suppresses
  21. L7:879 weakly excites
  22. L16:1699 weakly excites
  23. L26:1415 weakly excites 're' prefix fragment
  24. L3:2339 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.

1 values between -0.69 and -0.632 values between -0.63 and -0.582 values between -0.53 and -0.481 values between -0.48 and -0.438 values between -0.38 and -0.3322 values between -0.33 and -0.2862 values between -0.28 and -0.23121 values between -0.23 and -0.18262 values between -0.18 and -0.13514 values between -0.13 and -0.08806 values between -0.08 and -0.03876 values between -0.03 and 0.02582 values between 0.02 and 0.07353 values between 0.07 and 0.12156 values between 0.12 and 0.1784 values between 0.17 and 0.2232 values between 0.22 and 0.2712 values between 0.27 and 0.327 values between 0.32 and 0.373 values between 0.37 and 0.431 values between 0.43 and 0.481 values between 1.28 and 1.33offnegativepositiveoften

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. apon
  2. eners
  3. laden
  4. UIFont
  5. unes
  6. SEEK
  7. toThrow
  8. ⽉
  9. etSocketAddress
  10. été

Suppresses

  1. .wikipedia
  2. 一站式
  3. 突破口
  4. milestones
  5. rest
  6. 里程碑
  7. .gov
  8. .ViewModel
  9. ving
  10. 最后一个