Research map / Continual learning / Class-incremental
TAPE: Task-Adaptive Prototype Evolution in Audio-Language Models for Fully Few-shot Class-incremental Audio Classification
Yunlong Gao, Wenxin Liang, Guanglu Wang, Senqi Guan, Linlin Zong, Dongyu Zhang, Xinyue Liu
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 Prototype Evolution to baseline (ablation) | prototypical-network baseline · LBS-100 | Average Accuracy: 40.30 -> 83.54 | Improved |
| add Prototype Evolution to baseline (ablation) | prototypical-network baseline · FSC-89 | Average Accuracy: 48.12 -> 67.88 | Improved |
| add Task-Adapter to baseline (ablation) | prototypical-network baseline · LBS-100 | Performance Dropping rate: 61.94 -> 15.49 | Improved |
| add Task-Adapter to baseline (ablation) | prototypical-network baseline · Nsynth-100 | Performance Dropping rate: 28.89 -> 5.83 | Improved |
| add Task-Adapter on top of Prototype Evolution (ablation) | Prototype Evolution variant · FSC-89 | Performance Dropping rate: 21.19 -> 20.53 | Improved |
| add Prototype Evolution on top of Task-Adapter (ablation) | Task-Adapter variant · LBS-100 | Average Accuracy: 82.20 -> 84.53 | Improved |
| use frozen CLAP encoder with prototypes instead of prompt tuning (TAPE vs PALM) | PALM · Nsynth-100 | Average Accuracy: 51.65 -> 94.78 | Improved |
| zero-shot CLAP with class names | CLAP · LBS-100 | accuracy: about 10% average accuracy | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- iCaRL — iCaRL: Incremental Classifier and Representation Learning (CVPR 2017) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- UniAudio 1.5: Large Language Model-Driven Audio Codec is A Few-Shot Audio Task Learner (NeurIPS 2024 · similarity 0.86)
- Few-Shot Audio-Visual Class-Incremental Learning with Temporal Prompting and Regularization (AAAI 2025 · similarity 0.85)
- Measuring Audio's Impact on Correctness: Audio-Contribution-Aware Post-Training of Large Audio Language Models (ICLR 2026 · similarity 0.85)
- Synthio: Augmenting Small-Scale Audio Classification Datasets with Synthetic Data (ICLR 2025 · similarity 0.84)
- AudioCALM: Continuous Autoregressive Language Modeling for Universal Audio Generation (NeurIPS 2026 · similarity 0.84)
- What to Remember: Self-Adaptive Continual Learning for Audio Deepfake Detection (AAAI 2024 · 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.
