arXiv：cs.AI· Peiwei Chang, Li Zhang, Mark D. Plumbley, Philip J. B. Jackson, Wenwu Wang·· 20 小时前
AMemOps：评估长音频流中的记忆更新
AMemOps: Evaluating Memory Updates in Long Audio Streams
AI 导读
研究者提出 Audio Memory Operations（AMemOps），一个面向长音频流的基准，用于测试最新属性检索、未变属性保持与缺失来源拒识。配对控制实验显示，在测试乐器的其他录音上训练，可让参考 delta-rule 记忆相对匹配注意力机制的更新优势提升 13.4 个百分点；在 120 秒检索延迟下，匹配历史可区分一致性重复的收益与早期值冲突的代价，加性记忆控制则识别出擦除带来的收益。
正文
Abstract:Audio memory must recognise recurring sources across acoustic changes while keeping their attributes up to date. Overall accuracy does not reveal how these demands interact. We introduce Audio Memory Operations (AMemOps), a benchmark for latest-attribute retrieval, preservation of unchanged attributes and absent-source rejection in long audio streams. Paired controls vary recordings, training identities and earlier events to examine their effects on updating. Training on other recordings of the test instruments increases a reference delta-rule memory's update advantage over matched attention by 13.4 percentage points. At retrieval delays of 120\,s, matched histories separate benefits of consistent repetition from costs of conflicting earlier values, while additive-memory controls identify benefits of erasure.
| Subjects: | Sound (cs.SD); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.18652 [cs.SD] |
| (or arXiv:2610.18652v1 [cs.SD] for this version) | |
| https://doi.org/10.48550/arXiv.2610.18652 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Peiwei Chang [view email]
[v1]
Tue, 13 Oct 2026 16:16:19 UTC (279 KB)
来源：arXiv：cs.AI · arxiv.org