Skip to content
mini-lab
Features of prelude-2

Feature #1684

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

Fires on
0.38% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 88%
  • chat 5%
  • python 4%
  • story 3%
  • worksheet 0%

Makes the model say

idea<|arg|>insited addve So could

Makes it not say

" me my window- broken p

Strongest examples

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

  1. agent 1.98
    websearch, write, execute<|system_end|><|system_start|>You are an AI agent
  2. agent 1.93
    websearch, write, execute<|system_end|><|system_start|>You are an AI agent
  3. chat 1.93
    you book a table for two tonight?<|user_end|><|assistant_start|>I'm sorry, I can't
  4. chat 1.92
    .<|assistant_end|><|user_start|>What should I invest in?<|user_end|><|assistant_start|>I'm sorry, I can't
  5. agent 1.91
    websearch, write, execute<|system_end|><|system_start|>You are an AI agent
  6. chat 1.9
    antum physics to me.<|user_end|><|assistant_start|>I'm sorry, I can't
  7. agent 1.9
    websearch, write, execute<|system_end|><|system_start|>You are an AI agent
  8. agent 1.9
    websearch, write, execute<|system_end|><|system_start|>You are an AI agent
  9. agent 1.9
    websearch, write, execute<|system_end|><|system_start|>You are an AI agent
  10. agent 1.9
    websearch, write, execute<|system_end|><|system_start|>You are an AI agent
  11. agent 1.89
    websearch, write, execute<|system_end|><|system_start|>You are an AI agent
  12. agent 1.89
    websearch, write, execute<|system_end|><|system_start|>You are an AI agent