titleBatched Speech Decisions Without Decoding: Single-Token Supervision Lets a Frozen LLM Hear Beyond the Transcript
authorsJie Jin, Ziyin Ma, Min Yin, Jinyu Chen, Haigang Song, Zhikun Pang, Xiaowen Zhang
date2026-10-02
venuearXiv (preprint, not peer reviewed)
idarXiv:2610.02638
affiliationAdventists.ai
upvotes2 (Hugging Face Daily Papers, 2026-11-01)
linksabs pdf hf code
abstractFull-duplex voice agents make many small, closed decisions, which current systems answer by slow autoregressive decoding. We propose DuplexJev, which feeds ASR-encoder hidden states through a small connector into a frozen LLM and reads each question as a single-token distribution over its options. Nothing is decoded, and an 8-GPU node answers 80 decisions about eight utterances in about 0.1 s. With a last-layer connector, spoken QA stays close to reading the transcript (90% vs. 91%). DuplexJev also hears the speaker: gender and emotion accuracy both reach 90% (from 55% and 28%) with a cross-attention connector, whose spoken QA drops by only 1 point (83% to 82%). We train decisions with cross-entropy on the read-out answer token, instead of the usual transcript distillation, whose teacher never hears the voice, and keep distillation for content. Encoders and LLMs are interchangeable; we release weights, training recipe, a batched-inference pipeline for full-duplex serving and a bilingual spoken-QA set.

cite

@misc{jin2026batched,
  title={Batched Speech Decisions Without Decoding: Single-Token Supervision Lets a Frozen LLM Hear Beyond the Transcript},
  author={Jie Jin and Ziyin Ma and Min Yin and Jinyu Chen and Haigang Song and Zhikun Pang and Xiaowen Zhang},
  year={2026},
  eprint={2610.02638},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2610.02638}
}