ArtifactBench: Evaluating AI Music Detectors Under Distribution Shift
A new paper introduces ArtifactBench, a lineage-aware evaluation suite designed to test AI-generated music detectors under distribution shift. The benchmark groups recordings by content identity and separates calibration from testing to provide a more rigorous assessment of detector performance. Results show that while ArtifactNet achieves high accuracy, other public detectors perform poorly under the shifted cohort, revealing limitations hidden by aggregate scores.
- ▪ArtifactBench is a lineage-aware evaluation suite that measures detector behavior across different generator families, versions, and real-music domains.
- ▪The benchmark separates calibration from final testing and records inference failures independently from classification errors to report source-level performance with uncertainty.
- ▪On the 562-track common-success test intersection, ArtifactNet achieved an AUROC of 0.982 and balanced accuracy of 0.918, significantly outperforming the public Deezer detector.
- ▪Detectors such as SpecTTTra and CLAM fell below 0.30 AUROC under the shifted cohort, highlighting substantial generator and real-domain shifts.
- ▪The study demonstrates that aggregate scores alone can conceal significant performance variations caused by leakage control, cohort availability, and model-specific missingness.
Hacker News (AI / LLM) files mainly under ai. We currently carry 5,919 of its stories.
Story provenance
Source · retrieval · rights · ranking — open for full record
inspect →
Story provenance
Attribution is not the same as permission. This drawer separates discovery metadata, excerpts, WeSearch-generated summaries, reuse status, and whether the publisher receives the visit. Nothing here claims a legal grant the publisher has not made.
Record
| Original publisher | arXiv.org |
| Canonical URL | https://arxiv.org/abs/2609.23550 |
| Publication time | Tue, 22 Sep 2026 03:25:08 +0000 |
| Retrieval time | 2026-09-22T03:28:49.488Z |
| Last seen | 2026-09-22T03:28:49.488Z |
| Headline source | Publisher (no WeSearch rewrite) |
| Excerpt source | publisher body |
| Excerpt method | First ~120 words (~800 chars) of extracted publisher body, fair-use limited. |
| Summary | WeSearch · cerebras-chat (WeSearch summarizer) |
| Summary source text | contentText |
| Citation coverage | Summary is a WeSearch-generated derivative; primary citation is the original publisher URL. |
| Cluster | eqpJEPMKRLxE · 1 stories |
| Cluster logic | Grouped by semantic title/content similarity across sources within a rolling window. Same-publisher template collisions are excluded from coverage comparison. |
| Ranking reason | Story pages are not engagement-ranked. Hub feeds use recency, with optional source-diversified chronological ordering (cap consecutive stories per source). No personalized ranking. |
| Publisher visit | Yes — open original |
| Substitutes article? | No — link-out required for full text |
Rights status (four layers)
WeSearch handling by dimension
| Indexing | May the item be indexed (stored, ranked, made findable)? | Allowed |
| Snippet | May a short excerpt of the publisher's text be shown? | Allowed |
| AI summary | May WeSearch generate its own short summary of the article? | Limited |
| Retrieval / RAG | May the content be exposed for third-party retrieval-augmented generation? | Not asserted |
| Model training | May the content be used to train AI models? | Not asserted |
| Commercial reuse | May the content be reused commercially? | Not permitted |
Basis: Derived from the published RSS/Atom feed. Contact: [email protected]. Reviewed: 2026-07-24.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Sound arXiv:2609.23550 (cs) [Submitted on 20 Sep 2026] Title:ArtifactBench: Lineage-Aware Evaluation of AI-Generated Music Detectors under Distribution Shift Authors:Heewon Oh View a PDF of the paper titled ArtifactBench: Lineage-Aware Evaluation of AI-Generated Music Detectors under Distribution Shift, by Heewon Oh View PDF HTML (experimental) Abstract:AI-generated music detectors are commonly compared using aggregate scores on benchmarks whose training overlap, generator lineage, source provenance, and audio-transformation history are only partially observable. This paper introduces ArtifactBench, a lineage-aware evaluation suite for measuring detector behavior across generator families and versions, real-music domains, collection-cohort shift, and inference coverage.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv.org.
Discussion
0 commentsMore from arXiv.org
