Reproduce “The Right Source Is Not Enough: Understanding History Interference in Speech-LLM Memory” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
HongJin Song, Weiqiao Shan, Runwu Shi, Yifei Wu, Yujin Wang, Jiale Luo, Jiasheng Kuang, Xinyu Yang and 6 more
The authors test how MOSS-Audio and MiMo-Audio use remembered facts in multi-turn speech question answering. They block access to irrelevant history and intervene on role and answer-value signals. Blocking greatly reduces open-ended errors but leaves many multiple-choice errors; combined diagnostic shifts recover 79.1% of selected MOSS errors, not a deployable fix.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.