Round r70-rve-task8v2
Round 70 task 8 v2 smoke lane
Source-reported narrativeRound README and results text are source material. Numeric values that are not tied to a claim ledger record remain source-reported.
Round source README · rounds/r70-rve-task8v2/README.md
Round date: HISTORICAL (before 2026-10-09) Publication badge: INCOMPLETE; KEPT FOR AUDIT Why not VALID: The historical inventory retains INCOMPLETE because of NO_OUTCOME, NO_SCORED_RESULTS, SMOKE_GATE. See the round evidence and limitations below. Recomputation status: OBSERVED SOURCE ONLY
Status
INCOMPLETE
- NO_SCORED_RESULTS — The README says scored cells were not run.
- SMOKE_GATE — The v2 smoke did not pass its required admission gate.
- NO_OUTCOME — No scored task-8 comparison exists.
STATUS: INCOMPLETE for scored cells. Analysis: DESCRIPTIVE; the two smoke results are not a language comparison.
This is the v2 prompt cohort for Rust and Go only. The separately recorded v1 smoke cohort remains in task 8. No scored v2 cells were released. The two v2 smoke cells both returned 1/25 and failed the predeclared release gate. This is an observed smoke outcome, not a scored task-8 result.
Cohort accounting
| Lifecycle field | n | Basis |
|---|---|---|
| Maximum planned v2 executions | 4 | One smoke per stack and at most one follow-up per stack if its smoke passed. |
| Started | 2 | Rust and Go v2 smoke IDs only. |
| Finished smoke executions | 2 | Both returned an official smoke grade. |
| Officially graded smoke cells | 2 | Both are marked not scored in the test-count ledger. |
| Scored cells | 0 | The release gate failed for both stacks; no bulk cells followed. |
| Scored ITT denominator | 0 | No scored cells were released. Smoke outcomes remain outside that denominator. |
| Task ID | Exact arm | Cell ID | Smoke result | Test fraction | Scored |
|---|---|---|---|---|---|
| r70-8-rust-v2 | rust | codex__gpt-6-luna__max__default__r70-8-rust-v2__r1 |
fail | 1/25 | no |
| r70-8-go-v2 | go | codex__gpt-6-luna__max__default__r70-8-go-v2__r1 |
fail | 1/25 | no |
The v2 admission controls are separate from scored outcomes: one reference passed and one no-op failed for each stack. Exact rows are in the controls ledger under cohort r70-task8-v2. The public receipts do not include the versioned task-8 prompt or grader version/hash, so the prompt-based diagnosis is not independently checked here.
Reproduction record
- Kogen commit: not applicable; these are direct Codex task builds, not Kogen runs.
- Harness commit / CLI version: no exact per-cell harness Git commit or Codex CLI release is recorded for this v2 smoke cohort. The design says the runner and sandbox follow R70; that does not establish an exact per-cell revision.
- Model and effort: direct Codex,
gpt-6-luna,max. - Task IDs:
r70-8-rust-v2andr70-8-go-v2. - Historical execution command: not retained in the public record.
- Reproduce the published outcomes and control counts: run
python3 rounds/r70-rve-task8v2/reproduce.pyfrom the repository root. This reads ledger rows and does not launch cells. - Raw records: test-count ledger, cohort
r70-task8-v2-smoke; controls ledger, cohortr70-task8-v2.
The prompt wording was changed from v1 for the retry-body/cache-key rule. Because v2 never passed the smoke gate, this lane establishes no language or prompt-effect result.
Recomputed from raw records
`recomputed.json`, mined cell records, and the token audit back observed official outcomes, per-arm pass counts/rates, manifest token counters, and wall medians. The source contains 2 rows (fail 2); overall pass rate is 0.0% (0/2) among pass/fail grades. Per-arm detail and coverage counts are in the JSON.
No shared cell IDs are present in the committed public grade export; the source-only cohort is checked against the page's reported figures below. No shared public-export cell IDs are available for a paired outcome comparison. No complete published metadata/manifest token pairs are available for this round (0 missing manifest usage; 2 missing published metadata). Missing counters remain unknown, not zero. This round also has 2 raw-only or non-public rows; they are kept separate from the published-export denominator.
Captured run records
Standard-schema records assigned to this round in the indexed snapshot.
No per-round file is listed. Check the run-record index for the complete partition list.
Reproduction inventory
Published analysis replay and full experiment replay are reported separately.
- Replay type
analysis-only- Script
- rounds/r70-rve-task8v2/reproduce.py
python3 rounds/r70-rve-task8v2/reproduce.py - Inputs
- Full execution replay
- Unavailable
- Scope
- Extracts the two v2 smoke rows and controls; it does not launch scored cells.
View the complete per-round reproduction inventory · Repository reproduction guide