Benchmark
Adaptive computational coverage (stopping rule)
Simulated
- Result
- 200 / 200 EXACT at the stopping point; 0 false terminations; 41.7 % (v4-balanced) and 31.2 % (v7-lowcov) fewer reads than the smallest fixed coverage with equal EXACT (% reads avoided)
- Classification
- Software-generated DNA strands passed through a software channel model. Not a laboratory result.
- Methodology
- Rule fixed in the pre-registration (§6). Acceptance: 0 false success and a Wilson 95 % upper bound on false terminations ≤ 0.05 per profile (achieved 0.037).
- Dataset
- EVAL seeds 91000-91099; each seed a pool at mean coverage 15 (NB dispersion 4), consumed in 0.5× batches.
- Configuration
- Decoder E; rule frozen on DEV seeds 90000-90009 at level `all`, margin m = 0 (FREEZE.json).
- Environment
- Development host: shared VPS, Intel Xeon Gold 6240 @ 2.60 GHz, 4 cores / 8 threads, Ubuntu 24.04, Python 3.12, CPU only.
- Limitations
- Decides when a read pool already drawn is enough to decode. It does not steer a sequencer and says nothing about sequencing cost.
- Version
- V9
- Source
- v9.0.0/experiments/v9/coverage/results/eval.jsonl
- Reproduce
experiments/v9/reproduce.sh coverage· how to reproduce