Skip to content
mini-lab
Log in Sign up
Features of prelude-2

Feature #529

Fires on “ent” (not labeled yet: read the examples)

Fires on
1.33% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 74%
  • python 13%
  • story 10%
  • chat 2%
  • worksheet 1%

Makes the model say

s] justM as closer named became

Makes it not say

my: +ishanpython " `"

Strongest examples

On text the dictionary never trained on. The underlined token is where it fires most.

  1. agent 2.72
    grep, question, read, shell, skill, subagent, webf
  2. agent 2.71
    grep, question, read, shell, skill, subagent, webf
  3. agent 2.7
    grep, question, read, shell, skill, subagent, webf
  4. agent 2.69
    grep, question, read, shell, skill, subagent, webf
  5. agent 2.69
    grep, question, read, shell, skill, subagent, webf
  6. agent 2.68
    grep, question, read, shell, skill, subagent, webf
  7. agent 2.67
    grep, question, read, shell, skill, subagent, webf
  8. agent 2.67
    grep, question, read, shell, skill, subagent, webf
  9. agent 2.67
    grep, question, read, shell, skill, subagent, webf
  10. agent 2.67
    grep, question, read, shell, skill, subagent, webf
  11. agent 2.67
    grep, question, read, shell, skill, subagent, webf
  12. agent 2.66
    grep, question, read, shell, skill, subagent, webf