Skip to content
mini-lab
Features of prelude-2

Feature #4040

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

Fires on
0.62% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 81%
  • story 15%
  • python 2%
  • chat 1%
  • worksheet 1%

Makes the model say

ouname thoughtigIfowed `us

Makes it not say

Tim It() to it day together later

Strongest examples

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

  1. agent 3.58
    websearch, write, execute<|system_end|><|system_start|>You
  2. agent 3.58
    websearch, write, execute<|system_end|><|system_start|>You
  3. agent 3.58
    websearch, write, execute<|system_end|><|system_start|>You
  4. agent 3.58
    websearch, write, execute<|system_end|><|system_start|>You
  5. agent 3.58
    websearch, write, execute<|system_end|><|system_start|>You
  6. agent 3.58
    bsearch, write, execute<|system_end|><|system_start|>You
  7. agent 3.58
    bsearch, write, execute<|system_end|><|system_start|>You
  8. agent 3.58
    bsearch, write, execute<|system_end|><|system_start|>You
  9. agent 3.58
    bsearch, write, execute<|system_end|><|system_start|>You
  10. agent 3.51
    arch, write, execute<|system_end|><|system_start|>You
  11. agent 3.51
    arch, write, execute<|system_end|><|system_start|>You
  12. agent 3.51
    arch, write, execute<|system_end|><|system_start|>You