Experiment ·
DEEPNORAG
Do the knowledge packs still matter for the deep tier with thinking 512: arm N (DT512 config with config.kit = []) vs control H = DT512 slice rows on V-para (diagnostic)?
From the record
- What was tested
- DEEPNORAG
- Result
- −0.045 (−0.083…−0.011)
- Verdict
- closed: rule failed, CI excluded the hoped-for gain, a step-0 gate failed, or the mechanism was falsified
- Detail
- closed — RETRIEVAL-MATTERS (diagnostic; stage 3 not needed since Δ ≤ −0.03)
- Set
- V-para (V-para) slice 20; v0.1 (smoke, 2 items)
- Judge
- Claude Haiku (L1)
- n
- 20 (signal20-vpara slice)
- Seeds
- 1 (s42)
- Pre-registered
- yes, sha256-frozen
- Reopen if
- n/a (diagnostic)
Linked facts · 2
- R33 Remove retrieval from deep tier: slice 20R33 Remove retrieval from deep tier: slice 20−0.045 (−0.083…−0.011)Kinda record, not a measurement (set, status, judge, n, seeds, date: not recorded)Caveatwhy closed: diagnostic: packs staySourceresearch record
reports/CLOSED.md:52Open the data file - Retrieval removal on deep tier (DEEPNORAG): no packs → −0.045 (packs stay (diagnostic))Retrieval removal on deep tier (DEEPNORAG): no packs → −0.045 (packs stay (diagnostic))Kinda record, not a measurement (set, status, judge, n, seeds, date: not recorded)Sourcenot in the public repositories yetOpen the data file