A small DuplexJev connector: the Whisper-small encoder (frozen) feeds a frozen Qwen3-4B-2507, and every typed question is read as a
single-token distribution over its options, with zero decode steps and many questions per forward pass. Trained with the
same recipe as the Qwen3-32B connectors released with the paper.
|
|
| Audio encoder (frozen) |
openai/whisper-small |
| LLM (frozen) |
Qwen/Qwen3-4B-Instruct-2507 |
| Connector |
last encoder layer → frame stacking + SwiGLU projector (B), 30.4 M trainable parameters |
| Audio tokens |
50 Hz encoder frames, 8 stacked → 6.25 audio tokens/s |
| Total parameters |
4141 M (encoder + LLM + connector) |
| This variant |
Content alignment only (R1–R2, transcript distillation). No paralinguistic training: gender and emotion are at chance. A starting point for your own decision fine-tuning. |
Quick start
pip install "duplexjev[speech]>=0.2.2"
from duplexjev import Decider, Question
d = Decider.from_pretrained("adventists-ai/DuplexJev-B-Whisper-small-Qwen3-4B-2507")
d.decide("utterance_en.wav", [
Question("turn", "Has the user finished speaking?", ["finished", "not finished"], lang="en"),
Question("intent", "What does the user want?", ["climate", "media", "navigation", "phone", "chit-chat"], lang="en"),
], lang="en")
d.decide("utterance_zh.wav", [ # Chinese speech: Chinese question and options
Question("turn", "用户说完了吗？", ["说完了", "还没说完"], lang="zh"),
Question("intent", "用户想做什么？", ["空调", "媒体", "导航", "电话", "闲聊"], lang="zh"),
], lang="zh")
Results
| Evaluation (%) |
paper protocol |
duplexjev 0.2.2 defaults |
| qa100 (spoken QA) |
53 |
61 |
| qa100, Qwen3-4B-2507 reading the transcript |
84 |
– |
| ZJU-ML (spoken QA, real + TTS) |
51 |
– |
| Easy-Turn (4-way turn state, 800 clips) |
56.5 |
– |
| Gender, 800 real utterances |
55.1 |
52.9 |
| Emotion, 4-way, 800 utterances |
31 |
30.8 |
Leaderboard scores (see the GitHub README): main language (mean of qa100, ZJU-ML, Easy-Turn) 53.5,
paralinguistics (mean of gender, emotion) 43.0.
Paper protocol: the evaluation prompts of the paper (single option order, one H200). duplexjev defaults: the package's
own prompts with lang matched to the clip. Spoken QA with a small LLM is bounded by the LLM itself (see the
transcript-reading row); use these connectors for short decisions and speaker cues rather than open knowledge questions.
Latency of one decision event (10 questions, one 4.5 s clip, duplexjev 0.2.2, plain PyTorch): 5.2 s on 8 CPU threads (Xeon Platinum 8558, fp32, not quantized); about 145 ms on one H200 (bf16, measured while the GPU was shared with a training job).
Training
Frozen encoder and LLM; only the connector is trained: R1 and R2 transcript distillation on two disjoint 0.5 M-utterance
packs of the Ultravox v0.6 mixture (32k steps each, global batch 16).
Code and scripts: https://github.com/adventists-ai/duplexjev.
Limitations
- Evaluated on read or acted speech and small test sets (100–800 items); not on streaming input.
- Small LLMs answer knowledge questions poorly even from the transcript.
- Emotion labels come from acted corpora; do not use the outputs to make decisions about individuals.
- The connector only works with the encoder and LLM listed above.
License
Apache-2.0 for the connector weights. Some training corpora (e.g. WenetSpeech, CoVoST 2) are licensed for non-commercial use only; check them before commercial use. The frozen models keep their own licences: Qwen3-4B-2507 (Apache-2.0) and the encoder (Apache-2.0).
Citation
@inproceedings{jin2027duplexjev,
title = {Batched Speech Decisions Without Decoding: Single-Token Supervision Lets a Frozen {LLM} Hear Beyond the Transcript},
author = {Jin, Jie and Ma, Ziyin and Yin, Min and Chen, Jinyu and Song, Haigang and Pang, Zhikun and Zhang, Xiaowen},
booktitle = {Submitted to IEEE ICASSP},
year = {2027}
}
Built by Adventists.ai. Claude (Anthropic) assisted with code.
