Run record · computational telemetry only

Phase 103 SeqQA GPU Decode Repair Eval

completedevalHOT_EVAL

phase-103-phase-103-seqqa-gpu-decode-repair-eval-5223314

Claim boundary

frontier_0_1_certified = falsescorecard_metric_eligible = false

This is computational observability telemetry only. Nothing here is a Frontier 0.1 certification, a scorecard metric, or biological proof. Computational results are never biological proof — wet-lab and provider validation is a separate, gated process.

Run details

Run id: phase-103-phase-103-seqqa-gpu-decode-repair-eval-5223314
Phase: 103
Run mode: HOT_EVAL Live eval against a held set — a harness statistic, not a scorecard.
Backend (model): unavailable
Device / location: unavailable
Commit: 5223314aa5a3eacb57095eaf8a1f1406bd065a94
Tag: sprint-102-seqqa-gpu-result-diagnostic-accepted
Started: unavailable
Ended: unavailable
Duration (s): unavailable
Records attempted: 25
Records completed: 25
Records scored: 25

What this proves

The fraction of scored eval records that matched exactly under the harness for this run.

What this does NOT prove

Does NOT certify the model, is NOT scorecard-eligible, and is NOT biological proof. Unscored records are excluded, never counted as zeros.

Metrics

seqqa_exact_match

0.0000

source: metric-artifact

aim_run_metric

unavailable

source: aim

records_completed

25.0000

source: summary

records_scored

25.0000

source: summary

Blocked-vs-completed counts are shown as evidence. A missing value reads "unavailable", never 0.

Artifacts & checksums

report[external_artifact_store]

sha256: 6b643ae774442e884137631f143ba08394c9960bc3f5b61c13ced5a93815cfbe

size: 5,364 bytes

checksum_card[external_artifact_store]

sha256: 2f25ea17b5ec406b5010bba2cdce7eeea31f04034237b88b1cae3c931a3c2fc0

size: unavailable

Storage location and topology are never shown — only the sanitized external_artifact_store label and the content checksum.