Overview / P1

P1 · refund-policy Q&A

20 runs declaring these stages, oldest first as this console received them.

Compare two runs · Drift across its runs

  1. load_docs0
  2. split1
  3. embed0
  4. retrieve0
  5. answer0

Under each stage: how many of this pipeline's 2 stored comparisons first differ there, as each one's stored report reads. 1 found no difference.

Show: All · Records agree · Need attention

5 of 20 runs shown.

Each mark is one run's "Records agree", as its own page reads it:

  • Records agree
  • Contradicts its record
  • Couldn't check
  • Not checked
  • No verdict recorded
  1. 8 p1-08-candidate-set2

    Records agree

    First difference · stage 2 · split

  2. 14 p1-14-candidate-set3-byte-changed

    Contradicts its record

    Not compared

  3. 15 p1-15-candidate-set4-file-deleted

    Contradicts its record

    Not compared

  4. 16 p1-16-candidate-set5-no-artifacts

    Not checked (held without its output files required)

    Not compared

  5. 20 p1-20-candidate-set6-answer-failed

    Records agree

    Not compared