通过分析模型的空标记，发现可有效抑制语音与翻译中的无中生有现象。
The Null Token Knows: Reducing Message-Free Hallucination in ASR and NMT
- 利用模型内置空标记得分作为拒绝生成的信号。
- 提高空标记分数能显著降低幻觉，但可能误删有效内容。
- 提出应同时评估抑制幻觉与误删代价，更全面衡量方法效果。
现代编码器-解码器系统在输入无可用信息时仍能生成流畅文本。本文通过分析自动语音识别（ASR）和神经机器翻译（NMT）模型中的预留空标记（null token），探究其是否携带可被利用的拒绝生成信号。我们在Whisper等模型中审计了原始空标记得分及标量对数偏移，并探查解码器状态，对比监督行编辑与传统外部门控机制。结果表明，多数模型确实存在可用的拒绝信号，但标准解码策略未能可靠利用。提升空标记得分可显著抑制虚构内容，但过度干预也会删除有效语音或缩短合法翻译。该研究将空标记转化为诊断幻觉的工具，强调应以幻觉抑制与误删代价双重指标评估拒答方法，而非仅关注幻觉减少。
原文摘要 · Abstract (English)
Modern encoder-decoder systems can produce fluent text even when their input contains no recoverable message. We study this failure in ASR and NMT through the models' reserved null tokens, asking whether the score for ending generation already carries a usable abstention signal. Across speech recognizers and translation models, we audit native null-token scores and scalar logit shifts. In Whisper, we additionally probe decoder states and compare supervised row edits with conventional external gates. The evaluated models often expose a useful abstention signal, but stock decoding does not reliably act on it. Raising the null-token score can sharply suppress fabrication, but aggressive intervention also deletes valid speech or shortens legitimate translations. These findings turn the null token into a diagnostic lens on hallucination and motivate evaluating abstention methods by both suppression and deletion costs, rather than by hallucination reduction alone.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。