Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
VocSim A Training-free Benchmark for Zero-shot Content Identity in Single-source Audio
Maris Basha, Anja T Zai, Sabine Stoll, Richard Hahnloser
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add transductive PCA (D100) (ablation) | Whisper-L-v3 EWMTF · VocSim public sets | P@1 (%): 61.5 -> 66.8 | Improved |
| add transductive PCA (D100) (ablation) | CLAP · VocSim public sets | P@1 (%): 61.7 -> 63.7 | Improved |
| add transductive PCA (D100) (ablation) | WavLM-Large · VocSim public sets | P@1 (%): 62.8 -> 64.1 | Improved |
| add transductive PCA (D100) (ablation) | Whisper-L-v3 · VocSim blind sets | P@1 (%): 11.5 -> 11.5 | No clear effect |
| use Whisper-Large-v3 instead of Whisper-Small (ablation) | Whisper · VocSim public sets | P@1 (%): 57.4 -> 66.8 | Improved |
| use Whisper-Large-v3 instead of Whisper-Small (ablation) | Whisper · VocSim public sets | GSR (%): 28.7 -> 41.7 | Improved |
| use Whisper-Large-v3 instead of Whisper-Small (ablation) | Whisper · VocSim blind sets | P@1 (%): 5.7 -> 11.5 | Improved |
| use Whisper-Large-v3 instead of Whisper-Small (ablation) | Whisper · VocSim blind sets | GSR (%): 29.1 -> 39.4 | Improved |
| include padding tokens in mean pooling instead of masking (ablation) | Whisper encoder · BC subset | P@1 (%): 41.1% -> 65.5% | Improved |
| add DTW re-ranking to statistical pooling (ablation) | Whisper pipeline · VocSim | P@1 and GSR | No clear effect |
| flip 1-20% of labels (ablation) | GSR metric · VocSim | P@1 and GSR | Hurt |
| vary encoder layer across all 32 layers (ablation) | Whisper encoder · VocSim | P@1 / GSR | No clear effect |
Closest papers by meaning
- Massive Sound Embedding Benchmark (MSEB) (NeurIPS 2025 · similarity 0.86)
- Weakly-supervised Audio Separation via Bi-modal Semantic Similarity (ICLR 2024 · similarity 0.84)
- BirdSet: A Large-Scale Dataset for Audio Classification in Avian Bioacoustics (ICLR 2025 · similarity 0.84)
- ZeroSep: Separate Anything in Audio with Zero Training (NeurIPS 2025 · similarity 0.84)
- Unified Speech Recognition: A Single Model for Auditory, Visual, and Audiovisual Inputs (NeurIPS 2024 · similarity 0.84)
- SSLAM: Enhancing Self-Supervised Models with Audio Mixtures for Polyphonic Soundscapes (ICLR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
