| <html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width, initial-scale=1"><title>WAVEBench</title><style>html{color-scheme:light dark}body{font:16px/1.65 system-ui,sans-serif;margin:0;padding:12px 20px;color:#1f2937;background:transparent}h1{font-size:28px;line-height:1.2;margin:8px 0 18px}h2{font-size:19px;margin:22px 0 8px}p{margin:12px 0}ul{padding-left:22px}li{margin:9px 0}a{color:#2563eb;text-underline-offset:3px}@media(prefers-color-scheme:dark){body{color:#d1d5db}a{color:#93c5fd}}</style></head><body><h1>WAVEBench</h1><p>WAVEBench is a benchmark for evaluating evolutionary generalization in genome language models through viral-versus-cellular classification of wastewater metagenomic reads. It combines taxonomic cross-validation, fixed evaluation reads, and composition controls to measure performance across levels of taxonomic novelty.</p><h2>Resources</h2><ul><li><a href="https://huggingface.co/datasets/WAVEBench/wavebench" target="_top">Benchmark data</a>: Kraken2 reports, taxonomic splits, read manifests, taxonomy, and frozen evaluation reads at three tiers.</li><li><a href="https://huggingface.co/datasets/WAVEBench/wavebench-figures" target="_top">Figures and results</a>: saved evaluation outputs, figure-generation inputs, and current manuscript and regenerated figures. The dataset card documents missing historical inputs and reproduction limits.</li></ul><p>The datasets are currently private and available to authorized collaborators. Full raw and deduplicated corpora and original per-read Kraken2 outputs are not distributed. Data provenance and preprocessing methods will be described in the accompanying paper (link forthcoming).</p></body></html> | |