Skip to content
mini-lab
Features of prelude-2

Feature #2717

Fires on “<|tool_call_start|>” (not labeled yet: read the examples)

Fires on
1.15% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 84%
  • story 8%
  • worksheet 4%
  • python 3%
  • chat 2%

Makes the model say

k mom importantill get<|tool_call_start|> cloth called

Makes it not say

off⏎ ⏎ upon_(Hello "

Strongest examples

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

  1. agent 2.71
    keep_even([7, 6, 0]))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  2. agent 2.66
    4: print(opposte(8))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  3. agent 2.6
    ⏎3: print(decrrement(6))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  4. agent 2.52
    ⏎6: print(howw_many(SCORES))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  5. agent 2.52
    ))⏎5: print(ngeate(7))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  6. agent 2.52
    print(raise_tto(1, 3))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  7. agent 2.48
    (WORD))⏎5: print(twice(2))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  8. agent 2.44
    ]))⏎5: print(count_items(NUMBERS))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  9. agent 2.43
    (NUMBERS))⏎7: print(cube(6))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  10. agent 2.43
    print(lpus(12, 10))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  11. agent 2.4
    ⏎⏎ Line 5: print(fact(1))⏎<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path
  12. agent 2.4
    ⏎15: print(hlaf(7))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path