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

Feature #2026

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

Fires on
0.21% of tokens
Mostly in
chat
Layer
after block 4

Where it fires

  • chat 44%
  • story 41%
  • agent 13%
  • python 2%
  • worksheet 0%

Makes the model say

too<|arg|>_ra herg of again

Makes it not say

won name favorite? printodd⏎4

Strongest examples

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

  1. chat 4.9
    tell his friends.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  2. chat 4.78
    when things were hard.<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  3. chat 4.65
    work they had done!<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  4. chat 4.64
    meadow some more.<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  5. chat 4.6
    full. The end.<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  6. chat 4.59
    6250.<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  7. chat 4.58
    sunny and clear day!<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  8. chat 4.58
    another story soon.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  9. chat 4.56
    a great day.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  10. chat 4.52
    and add numbers.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  11. chat 4.5
    time to go home.<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>
  12. chat 4.5
    ing gray shoes.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>