Skip to content
mini-lab
Features of prelude-2

Feature #2555

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

Fires on
1.30% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 71%
  • python 24%
  • story 3%
  • chat 1%
  • worksheet 1%

Makes the model say

([_ positiveab r seaOhe

Makes it not say

:// sheanceonewayAdded])Rename

Strongest examples

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

  1. agent 3.13
    import is_odd, raise_to⏎⏎ Line 5: assert raise_to(1
  2. agent 2.92
    , raise_to, sum_to⏎⏎ Line 4: assert raise_to(5
  3. agent 2.86
    1: from textlib import final_letter⏎⏎ Line 3: assert final_letter("lemon
  4. agent 2.79
    import max_of, next_number⏎⏎ Line 3: assert max_of(1
  5. agent 2.69
    import double_all, max_item⏎⏎ Line 5: assert max_item([3
  6. agent 2.67
    1: from textlib import num_chars⏎⏎ Line 3: assert num_chars("pizza
  7. agent 2.66
    , add_all, mean_list⏎⏎ Line 3: assert count([7,
  8. agent 2.63
    from items import has_item, first⏎⏎ Line 3: assert has_item([9
  9. agent 2.63
    import first, includes, max_item⏎⏎ Line 5: assert max_item([6
  10. agent 2.53
    1: from arith import positive, minus⏎⏎ Line 3: assert positive(-1) ==
  11. agent 2.52
    1: from listlib import max_item⏎⏎ Line 3: assert max_item([8
  12. agent 2.51
    1: from helpers import smaller, count⏎⏎ Line 5: assert count([6,