Skip to content

v0.2.1 — Separate the Signal

Latest

Choose a tag to compare

@deghosal-2026 deghosal-2026 released this 29 Aug 06:57
· 2 commits to main since this release

v0.2.1 — Separate the Signal

Mistral effect confirmed, pipeline invariants, false-negative measurement.

Key Findings

  • Mistral effect confirmed — DeepSeek+GPT separating experiment (0.246 convergence) disproves diversity-of-training-objective hypothesis. Mistral is the unique variable driving productive debate.
  • Pipeline integrity — 5 row-count invariant assertions at every pipeline seam. 2,333→359 collapse class eliminated.
  • False-negative measurement — 1.7–3.4% missed-issue rate across 59 known-bad PRs.
  • 0 flaky artifacts, 0 theatre, 87.4% MATCH ground truth (3,008/3,440).

Model Selection

  • Default production pair: GPT+Mistral (0.536 convergence)
  • Validation pair: DeepSeek+Mistral (0.572 convergence)
  • Recommendation: include Mistral — not just any diverse pair

Links