Artificial Intelligence · arXiv:2611.25773 v1 · submitted 19 Nov 2026 · 71 references
Duopra: From Full-Duplex Conversation to Full-Duplex Agency
The paper says
“Full-duplex interaction models exhibit strong conversational agency, enabling natural, low-latency turn-taking, backchannels, and interruptions. However, completing tasks further requires task agency: timely tool invocation and argument revision following user corrections.”
more of the abstract
Frontend-backend agents separate initiation from execution: the frontend decides whether and what to request, while the backend determines how. Execution can be offloaded, but unsent requests cannot be executed, and incomplete ones can misdirect the backend. On Full-Duplex-Bench v3 (FDB v3), most failures of Realtime-Venus-Audio + Harness in our streaming pipeline occur before a domain tool executes. We introduce Duopra, a full-duplex interaction model that learns task execution alongside conversational control. Duopra treats the backend as a tool, accepting a self-contained natural-language request. Direct calls and backend delegation share an action interface and requirements for faithful initiation and correct execution. Its three-stage training combines protocol SFT, full-duplex RL rewarding timely and complete execution, and projector-only repair of talker conditioning after thinker updates. With the same backend, Pass@1 rises from 4.7% to 44.7%; Duopra achieves 47.7% calling tools directly (26.0% after SFT) and 49.3% with a DeepSeek backend. It improves three FDB v1.0 takeover rates over the base model. These results show that task agency can coexist with key conversational agency.
from the abstract
Pith says
Machine review · deepseek-v4-flash · 20 Nov 2026 · read the full text (153,496 characters) · not checked by a person
It can be wrong. Challenge this review or reply to it.
Holds up, with conditions. Duopra shows that a full-duplex voice frontend can learn to initiate tool calls and backend requests, raising task completion from 4.7% to 44.7% on FDB v3, and the paper's analysis cleanly separates the initiation bottleneck from execution quality. This line is wrong
8 minor · 5 open questions
How often objections from reviews like this one are right: not yet measured
The paper, first character to last
The reviewer was given the whole text, 153,496 characters. Not checked: figures, code, what the 71 references say.
8 minor objections
The paper says, at §1, Figure 1
Pith says: minor objection 1 of 8
Figure 1 is described but not visible in the supplied text. The caption is informative, but the figure itself could not be inspected.
The paper says, at §3, Eq. (1)
Pith says: minor objection 2 of 8
The policy equation uses notation $x_{\le t}$, $e_{\le t}$, $y_{<t}$. The definition of $e_t$ (observations) could be clarified earlier; it is introduced in §4.1 but used in §3.
show 6 more
The paper says, at §4.3, Eq. (3)
Pith says: minor objection 3 of 8
The KL estimator $K_{ij}=v_{ij}-\log v_{ij}-1$ is defined in §A.2 but referenced in §4.3. The definition should be given earlier or the reference should be clearer.
The paper says, at §5.2, Table 1
Pith says: minor objection 4 of 8
The table uses superscripts (m, h, x, o) for footnotes, but the footnotes are at the bottom of the table and may be missed. Consider using a more prominent notation.
The paper says, at §A.5, Table 9
Pith says: minor objection 5 of 8
The reward-judge instruction in Table 9 is very long and complex. The paper should summarize the key validation steps that ensure the judge's reliability.
The paper says, at §A.8, Table 10
Pith says: minor objection 6 of 8
The table reports 'Passes by calls' for 1, 2, 3 required calls. The total number of sessions for each call count is 198, 54, 48 respectively (from 300 sessions total). This should be stated in the caption or a note.
The paper says, at §A.9, Table 11
Pith says: minor objection 7 of 8
The backchannel TOR and frequency are reported as 0.0 for all post-trained checkpoints. The paper should clarify whether this is because the model never produces backchannels or because the evaluation metric does not detect them.
The paper says, at §6, Limitations
Pith says: minor objection 8 of 8
The limitations section is well-written and acknowledges key issues (comparability, model coverage, fixed user input). However, it does not mention the potential circularity of using the same model family for reward and evaluation.
Open questions (5)
Pith raised these and could not confirm them against the text of the paper it read. They are questions for the authors, not findings.
- §5.2, Table 1; §A.6, Table 9 Pith raised a concern here that it could not check against the text of the paper it was given, so it is not published as an objection.
- §5.2, Table 1; §A.6, Table 9 Pith raised a concern here that it could not check against the text of the paper it was given, so it is not published as an objection.
- §5.3, Table 2; §A.8, Table 10 Pith raised a concern here that it could not check against the text of the paper it was given, so it is not published as an objection.
- §4.3, Eq. (3); §A.3, Table 5 Pith raised a concern here that it could not check against the text of the paper it was given, so it is not published as an objection.
- §5.4, Table 3; §A.9, Table 11 Pith raised a concern here that it could not check against the text of the paper it was given, so it is not published as an objection.
What would settle it
If an independent evaluation using a different judge (e.g., GPT-4o as in the original FDB v3 benchmark) or human annotation showed that Duopra's Pass@1 gain over RV-Audio + Harness is substantially smaller (e.g., less than 20 percentage points), the central claim that initiation is the primary bottleneck and that Duopra recovers unserved requests would be weakened.
from Pith's summary of the paper
How the review went
5 passes by deepseek-v4-flash · 20 Nov 2026
- Reader holds up, with conditions
- Skeptic concern not confirmed: listed as an open question
- Referee 8 minor
- Check against the paper kept 0 of 6 objections
- Editor wrote the verdict above
Not run on this paper: the simulated authors' defence.
Pith's passes raised 6 objections. A check against the paper's text kept 0 and left 6 as questions.
The review in full
The review in brief, machine-written
This paper makes a concrete empirical contribution. The central finding is that in frontend-backend voice agents, the bottleneck is not backend execution but frontend initiation: the model fails to send a request at all. Duopra addresses this by training a single full-duplex policy to emit tool calls and backend delegations through a unified action space, using a three-stage recipe of protocol SFT, execution-based RL, and projector-only talker repair. The result is a large and clean gain on FDB v3 (4.7% to 44.7% Pass@1 with the same Qwen3-8B backend), and the decomposition into Call and Pass|call convincingly shows where the improvement comes from.
The experimental design is thorough: three decoding seeds, paired bootstrap intervals, multiple baselines, ablations, and analysis of failure modes. The paper is honest about its limitations, including the use of synthetic speech, prerecorded requests, and a single base model. The loss of spontaneous backchanneling is reported rather than hidden.
The main soft spot is the reliance on the same LLM (Qwen3.8-27B) as both the reward judge during training and the evaluation judge for argument and response scores. The paper uses a detailed structured judgment protocol and validates the judge's output, but the potential circularity is not addressed. If the judge systematically favors the model's outputs, the reported scores could be inflated. This is a real concern, though the Pass@1 metric (which depends on exact call matching, not the judge) provides a partial check. A second concern is that the projector repair evaluation uses selection-set evidence, so the reported CER/WER reductions are not independent generalization estimates. The paper acknowledges this.
Who this is for: researchers working on full-duplex voice agents, tool use in spoken dialogue, and post-training for speech models. The paper deserves a serious referee. The authors should address the judge circularity concern, ideally by evaluating on a subset with a different judge or by providing evidence that the judge's errors are not systematically biased toward Duopra.
The stress test's concern
Not confirmed. Pith could not check this concern against the text of the paper it was given; it is listed as an open question, not a finding.
The referee's report
Not published. Pith's referee wrote this assessment before its objections were checked against the paper, and 5 of its 5 major objections did not survive that check.
Pith's reading of the paper (a machine's summary, not the authors')
The paper investigates why frontend–backend full-duplex voice agents fail to complete tasks. It finds that the bottleneck is not the backend's ability to execute tool calls, but the frontend's failure to initiate a faithful request in the first place. The authors introduce Duopra, a model that learns to make direct tool calls and to delegate through a unified action interface, trained with a three-stage procedure combining supervised fine-tuning, reinforcement learning with time-anchored rewards, and projector-only repair to preserve speech quality. On the Full-Duplex-Bench v3 benchmark, Duopra raises Pass@1 from 4.7% (RV-Audio + Harness) to 44.7% with the same Qwen3-8B backend, and to 49.3% with a DeepSeek backend, while also improving conversational takeover rates. The central claim is that task agency—deciding when and how to act—can be learned alongside conversational agency in a single causal policy.
What carries the argument. The central mechanism is the unified action interface that treats both direct tool calls and backend delegation as instances of the same exttt{call} operation, with the backend represented as a tool exttt{query(task)} accepting a self-contained natural-language argument. This allows the same causal streaming policy $\pi_\theta(y_t \mid x_{\le t}, e_{\le t}, y_{<t})$ to decide turn states, speech tokens, and private actions jointly. The three-stage training procedure is the key machinery: Stage 1 (protocol SFT) teaches schema-based calling and clarification; Stage 2 (full-duplex RL with Dr. GRPO) rewards timely, complete, and grounded execution using time-anchored deadlines and outcome tiers; Stage 3 (projector-only repair) realigns talker conditioning after thinker updates to mitigate speech degradation.
Core claim. The paper establishes that in frontend–backend full-duplex voice agents, task success is limited first by initiation, not execution. The frontend must decide whether and what to request; unsent requests cannot be executed. The authors demonstrate that a full-duplex interaction model can learn to initiate faithful tool calls and backend delegations through a unified action interface, recovering tasks that a baseline frontend never attempts. Duopra's three-stage training—protocol SFT, full-duplex RL with time-anchored deadlines and outcome tiers, and projector-only talker repair—enables the model to achieve 44.7% Pass@1 on FDB v3 with the same Qwen3-8B backend that the baseline uses, compared to 4.7% for the baseline. The gain comes almost entirely from sessions the baseline never executed, confirming that initiation is the primary bottleneck.
The premise it rests on. The reward judge (Qwen3.8-27B) reliably evaluates argument correctness, response grounding, and fabrication in the streaming setting, and the paper relies on this judge for both training rewards and evaluation metrics.
If the paper is right (Pith's inferences, not the authors')
If Duopra's approach generalizes, full-duplex voice agents can be trained to handle multi-step tool use, corrections, and conditional dependencies without relying on a separate dispatcher or backend for initiation decisions.
The finding that initiation is the primary bottleneck implies that improving backend execution alone will yield diminishing returns; effort should focus on frontend task agency.
The three-stage training recipe (SFT + RL + projector repair) provides a template for post-training other full-duplex models to acquire task agency while retaining conversational control.
The unified action interface (direct calls and delegation through the same exttt{call} operation) simplifies the architecture and allows the model to choose the best route without a separate routing module.
The paper's decomposition of Pass@1 into Call × Pass|call provides a diagnostic tool that could be applied to other frontend–backend systems to identify whether failures stem from initiation or execution.
The loss of spontaneous backchanneling after task training suggests a tension between task-oriented and purely conversational behaviors that may require explicit multi-objective optimization or data mixing to resolve.
The projector-only repair technique (aligning talker conditioning after thinker updates) could be adopted by other speech models that fine-tune a language model component and need to preserve speech quality without retraining the full speech generator.
The paper's information-sufficiency criterion (a call is justified only if each required argument has a source) formalizes a key aspect of task agency that could be used as a training signal or evaluation metric in other voice agent systems.
Check this review
What was read
The passes were given all 153,496 characters of the paper's LaTeX source.
Fingerprint of that text (SHA-256): 11cfeba32419278d0b08b13654f7e3ed69a971fbf2a7f296ef13ca5a85cfac05
Pith does not keep that text. Rebuilt from the paper's source or PDF by the same code, it has this fingerprint if it is the same text. From a copy of Pith's code:
PYTHONPATH=. python scripts/review_input_check.py 2611.25773
The passes, in the order they ran
- Reader deepseek-v4-flash · medium effort · $0.0054 · instructions of the Reader, fingerprint f9ead7300382…
- Skeptic deepseek-v4-flash · medium effort · $0.0045 · instructions of the Skeptic, fingerprint 8f88c0f61488…
- Referee deepseek-v4-flash · medium effort · $0.0048 · instructions of the Referee, fingerprint a852d99aea0d…
- Check against the paper deepseek-v4-flash · medium effort · $0.0067 · instructions of the Check against the paper, fingerprint 6c7a6883d551…
- Editor deepseek-v4-flash · medium effort · $0.0046 · instructions of the Editor, fingerprint e7211f6a3d27…
- Circularity check deepseek-v4-flash · medium effort · $0.0045 · instructions of the Circularity check, fingerprint e9946e39d67b…
- Summary for the page deepseek-v4-flash · medium effort · temperature 0.2 · $0.0049 · instructions of the Summary for the page, fingerprint e4f56bf568e7…
The passes cost $0.035 in model calls together.
Version stamped on this review: pith-pipeline@v1.4.1-daily-deepseek-or-medium+deepseek/deepseek-v4-flash+p.d60f54226f00
Each link opens a pass's instructions as plain text, the exact bytes its fingerprint was taken over. Every stored set of instructions.
curl -s https://pith.science/methodology/prompts/f9ead730038216984cfb56d7cb34332bdf691828296a1de4245f3d6667d4bf56 | sha256sum
prints the fingerprint of the reader's instructions. The same works for every pass.
The receipt
The review's fingerprint was taken on 20 Nov 2026. The review you are reading still matches it.
Fingerprint of the review (SHA-256): 3c9fb7fa20ecba3a5d45e71f89cc08bf99f727ab69c5c366ce7efcd23fd90795
curl -s 'https://pith.science/api/v1/verdicts/2611.25773?form=canonical' | sha256sum
prints the fingerprint of the review as it stands now. The rule, in four steps.
A matching fingerprint shows the review has not changed since that date. It does not show that the review is right.
How often reviews like this one are right
Reviews like this one: written by deepseek-v4-flash, from the paper's LaTeX source.
How often objections from reviews like this one are right: not yet measured
Reviews by people
This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.
The paper
Abstract
Full-duplex interaction models exhibit strong conversational agency, enabling natural, low-latency turn-taking, backchannels, and interruptions. However, completing tasks further requires task agency: timely tool invocation and argument revision following user corrections. Frontend-backend agents separate initiation from execution: the frontend decides whether and what to request, while the backend determines how. Execution can be offloaded, but unsent requests cannot be executed, and incomplete ones can misdirect the backend. On Full-Duplex-Bench v3 (FDB v3), most failures of Realtime-Venus-Audio + Harness in our streaming pipeline occur before a domain tool executes. We introduce Duopra, a full-duplex interaction model that learns task execution alongside conversational control. Duopra treats the backend as a tool, accepting a self-contained natural-language request. Direct calls and backend delegation share an action interface and requirements for faithful initiation and correct execution. Its three-stage training combines protocol SFT, full-duplex RL rewarding timely and complete execution, and projector-only repair of talker conditioning after thinker updates. With the same backend, Pass@1 rises from 4.7% to 44.7%; Duopra achieves 47.7% calling tools directly (26.0% after SFT) and 49.3% with a DeepSeek backend. It improves three FDB v1.0 takeover rates over the base model. These results show that task agency can coexist with key conversational agency.
The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.
abstract compiled · 2026-11-20
The authors' main claim, as the classifier picked it out 1.00
“These results show that task agency can coexist with key conversational agency.”
How the main claim is stated: asserted 1.00
What the abstract itself states, as probabilities: claims first 0.07 · states limitation 0.25 · human participants 0.11 · controlled comparison 0.90 · randomized assignment 0.02 · states quantitative result 0.99 · compares with prior results 0.96 · mentions code or data availability 0.04
The abstract, sentence by sentence
This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.
| Sentence | Role | The authors' own work | States the main conclusion | Strength |
|---|---|---|---|---|
| Full-duplex interaction models exhibit strong conversational agency, enabling natural, low-latency turn-taking, backchannels, and interruptions. | background 1.00 | 0.09 | 0.04 | asserted 1.00 |
| However, completing tasks further requires task agency: timely tool invocation and argument revision following user corrections. | gap 0.80 | 0.10 | 0.05 | asserted 1.00 |
| Frontend-backend agents separate initiation from execution: the frontend decides whether and what to request, while the backend determines how. | background 0.98 | 0.09 | 0.03 | asserted 1.00 |
| Execution can be offloaded, but unsent requests cannot be executed, and incomplete ones can misdirect the backend. | background 0.82 | 0.19 | 0.03 | asserted 0.87 |
| On Full-Duplex-Bench v3 (FDB v3), most failures of Realtime-Venus-Audio + Harness in our streaming pipeline occur before a domain tool executes. | result 0.66 | 0.93 | 0.04 | asserted 1.00 |
| We introduce Duopra, a full-duplex interaction model that learns task execution alongside conversational control. | contribution 0.98 | 0.87 | 0.09 | asserted 0.99 |
| Duopra treats the backend as a tool, accepting a self-contained natural-language request. | method 0.85 | 0.68 | 0.06 | asserted 1.00 |
| Direct calls and backend delegation share an action interface and requirements for faithful initiation and correct execution. | method 0.86 | 0.24 | 0.05 | asserted 1.00 |
| Its three-stage training combines protocol SFT, full-duplex RL rewarding timely and complete execution, and projector-only repair of talker conditioning after thinker updates. | method 0.99 | 0.90 | 0.04 | asserted 0.98 |
| With the same backend, Pass@1 rises from 4.7% to 44.7%; Duopra achieves 47.7% calling tools directly (26.0% after SFT) and 49.3% with a DeepSeek backend. | result 1.00 | 0.95 | 0.12 | asserted 1.00 |
| It improves three FDB v1.0 takeover rates over the base model. | result 1.00 | 0.91 | 0.10 | asserted 1.00 |
| These results show that task agency can coexist with key conversational agency. | interpretation 1.00 | 0.82 | 0.85 | asserted 1.00 |
What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.99
What was examined, and what was not
Sources read
- arXiv listing abstract v1
Not examined
- full text
- figures and tables
- references and citing sentences
- relations to other papers
Levels run
- P abstract
Not run
- S sections
- T body sentences
- C citations
- R cross-paper relations
How the labels were made
Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-20: 1 request, 8,798 input tokens.
Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON
Figures (5)
Figures are the authors', reproduced from the paper.
References (71)
- [1] Lin, Guan-Ting and Lian, Jiachen and Li, Tingle and Wang, Qirui and Anumanchipalli, Gopala and Liu, Alexander H. and Lee, Hung-yi , booktitle =. 2025 , address =. doi:10.1109/asru65441.2025.11433838 , url = on Pithdoi:10.1109/asru65441.2025.11433838
- [2] Lin, Guan-Ting and Kuan, Shih-Yun Shan and Wang, Qirui and Lian, Jiachen and Li, Tingle and Watanabe, Shinji and Lee, Hung-yi , booktitle =. 2026 , address =. doi:10.1109/icassp55912.2026.11463576 , url = on Pithdoi:10.1109/icassp55912.2026.11463576
- [3] Lin, Guan-Ting and Kuan, Shih-Yun Shan and Shi, Jiatong and Chang, Kai-Wei and Arora, Siddhant and Watanabe, Shinji and Lee, Hung-yi , booktitle =. 2026 , month = jul, address =. doi:10.18653/v1/2026.acl-short.4 , url = on Pithdoi:10.18653/v1/2026.acl-short.4
- [4] Lin, Guan-Ting and Chen, Chen and Chen, Zhehuai and Lee, Hung-yi , year =. 2604.04847 , archivePrefix = on PithPith's reviewarXiv
- [5] 2026 , month = aug, howpublished = on Pith
- [6] Huang, Muye and Zhang, Lingling and Yu, Xingyu and Shi, Lei and Ma, Zhanyu and Xu, Jun and Gao, Jiuchong and Hao, Jinghua and He, Renqing and Liu, Jun , year =. 2606.09186 , archivePrefix = on PithPith's reviewarXiv
- [7] Yao, Dingyu and Zhou, Junhao and Yang, Chenxu and Qin, Chuanyu and Zeng, Xiangyu and Li, Yifei and Hou, Haowen and Liang, Zheming and Wang, Congcong and Tuo, Kaiwen and Zhang, Jun and Zhu, Yuhan and Cao, Yuhang and Ye, Shenglong and Xie, Shuai and Gu, Shuhuan and Huang, Haoyan... on PithPith's reviewarXiv
- [8] A frontend-backend architecture for tool calls in full-duplex speech models , author =. 2026 , eprint = on Pith
- [9] Balam, Jagadeesh and Bartley, Travis and Casanova, Edresson and Chauhan, Sanjay and Chen, Chen and Chen, Zhehuai and Chen, Zijia and Ciannella, Francesco and De Mello, Shalini and Deng, Slyne and others , year =. 2609.21967 , archivePrefix = on PithPith's reviewarXiv
- [10] Chien, Chung-Ming and Orsini, Manu and Kharitonov, Eugene and Zeghidour, Neil and Livescu, Karen and D. Proceedings of the 43rd International Conference on Machine Learning , series =. 2026 , publisher = on Pith
- [11] Zhang, Haoyang and Chen, Jun and Wu, Donghang and Li, Yuxin and Zhang, Yuxin and Zhang, Xiangyu Tony and Liu, Che and Lin, Qingjian and Peng, Yizhou and Liu, Hexin and Chng, Eng Siong and Yan, Chao and Wu, Boyong and Huang, Yechang and Yang, Xuerui and Tian, Fei , year =. 2605... on PithPith's reviewarXiv
- [12] Bao, Lujia and Chen, Qian and Cheng, Luyao and Deng, Chong and Kong, Yuxiang and Li, Xiangang and Li, Xu and Liu, Jiaqing and Tan, Chao-Hong and Wang, Haoyu and Wang, Wen and Wang, Xilou and Xu, Haoxiang and Xu, Junhao and Yi, Liang and Zhang, Binbin and Zhang, Qinglin and Zha... on PithPith's reviewarXiv
- [13] Introducing. 2025 , howpublished = on Pith
- [14] Interaction Models: A Scalable Approach to. Thinking Machines Lab: Connectionism , year =. doi:10.64434/tml.20260511 , url = on Pithdoi:10.64434/tml.20260511
- [15] Cui, Junbo and Xu, Bokai and Wang, Chongyi and Yu, Tianyu and Sun, Weiyue and Xu, Yingjing and Wang, Tianran and He, Zhihui and Ma, Wenshuo and Cai, Tianchi and others , year =. 2604.27393 , archivePrefix = on PithPith's reviewarXiv
- [16] Wang, Xiong and Li, Yangze and Fu, Chaoyou and Zhang, Yike and Shen, Yunhang and Xie, Lei and Li, Ke and Sun, Xing and Ma, Long , booktitle =. 2025 , publisher = on Pith
- [17] Efficient and Direct Duplex Modeling for Speech-to-Speech Language Model , author =. Interspeech 2025 , pages =. 2025 , doi = on Pith
- [18] Yu, Wenyi and Wang, Siyin and Yang, Xiaoyu and Chen, Xianzhao and Tian, Xiaohai and Zhang, Jun and Sun, Guangzhi and Lu, Lu and Wang, Yuxuan and Zhang, Chao , booktitle =. 2025 , publisher =. doi:10.52202/085713-0829 , url = on Pithdoi:10.52202/085713-0829
- [19] Fang, Qingkai and Guo, Shoutao and Feng, Yang , year =. 2606.14528 , archivePrefix = on PitharXiv
- [20] Yao, Yiqun and Li, Xiang and Jiang, Xin and Fang, Xuezhi and Yu, Naitong and Ma, Wenjia and Sun, Aixin and Wang, Yequan , year =. 2509.02521 , archivePrefix = on PitharXiv
show 51 more
- [21] Chen, Qian and Cheng, Luyao and Deng, Chong and Li, Xiangang and Liu, Jiaqing and Tan, Chao-Hong and Wang, Wen and Xu, Junhao and Ye, Jieping and Zhang, Qinglin and Zhang, Qiquan and Zhou, Jingren , year =. 2512.20156v1 , archivePrefix = on PitharXiv
- [22] Zhou, Zhiyang and Shang, Yingxin and Wang, Zhou and Cai, Hongwei and Wang, Weixu and Zhou, Shuran and Zhao, Shuofeng and Fan, Wenke and Guo, Qingxiang and Yang, Dawei and Yang, Lin and Song, Yang , year =. 2609.29217 , archivePrefix = on PithPith's reviewarXiv
- [23] Chiang, Cheng-Han and Wang, Xiaofei and Li, Linjie and Lin, Chung-Ching and Lin, Kevin and Liu, Shujie and Wang, Zhendong and Yang, Zhengyuan and Lee, Hung-yi and Wang, Lijuan , booktitle =. 2026 , month = jul, address =. doi:10.18653/v1/2026.acl-long.404 , url = on Pithdoi:10.18653/v1/2026.acl-long.404
- [24] Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models , author =. Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing , pages =. 2026 , month = oct, address =. doi:10.18653/2026.emnlp-main.2331 , url = on Pithdoi:10.18653/2026.emnlp-main.2331
- [25] Hsiao, Chi-Yuan and Lu, Ke-Han and Fu, Yu-Kuan and Lin, Guan-Ting and Hung, Hsiao-Tsung and Lee, Hung-yi , year =. 2604.10065 , archivePrefix = on PithPith's reviewarXiv
- [26] Ray, Soham and Dhandhania, Keshav and Barres, Victor and Narasimhan, Karthik , booktitle =. 2026 , publisher = on Pith
- [27] Bogavelli, Tara and Melan. Findings of the Association for Computational Linguistics: EMNLP 2026 , pages =. 2026 , month = oct, address =. doi:10.18653/2026.findings-emnlp.2182 , url = on Pithdoi:10.18653/2026.findings-emnlp.2182
- [28] Pahwa, Ramit and Beedu, Apoorva and Priye, Parivesh and Gandhi, Rutu and Takawale, Saloni and Baijal, Aruna and Yang, Zengli , booktitle =. 2026 , doi = on Pith
- [29] Jain, Dhruv and Shukla, Harshit and Rajeev, Gautam and Kulkarni, Ashish and Khatri, Chandra and Agarwal, Shubham , year =. 2510.07978 , archivePrefix = on PitharXiv
- [30] Chen, Yiming and Yue, Xianghu and Zhang, Chen and Gao, Xiaoxue and Tan, Robby T. and Li, Haizhou , journal =. 2026 , address =. doi:10.1162/tacl.a.628 , url = on Pithdoi:10.1162/tacl.a.628
- [31] Generative Spoken Dialogue Language Modeling , author =. Transactions of the Association for Computational Linguistics , volume =. 2023 , address =. doi:10.1162/tacl_a_00545 , url = on Pithdoi:10.1162/tacl_a_00545
- [32] Veluri, Bandhav and Peloquin, Benjamin N. and Yu, Bokai and Gong, Hongyu and Gollakota, Shyamnath , booktitle =. Beyond Turn-Based Interfaces: Synchronous. 2024 , month = nov, address =. doi:10.18653/v1/2024.emnlp-main.1192 , url = on Pithdoi:10.18653/v1/2024.emnlp-main.1192
- [33] Turn-taking in Conversational Systems and Human-Robot Interaction: A Review , author =. Computer Speech & Language , volume =. 2021 , doi = on Pith
- [34] Aligning Spoken Dialogue Models from User Interactions , author =. Proceedings of the 42nd International Conference on Machine Learning , series =. 2025 , publisher = on Pith
- [35] Lin, Guan-Ting and Shivakumar, Prashanth Gurunath and Gourav, Aditya and Gu, Yile and Gandhe, Ankur and Lee, Hung-yi and Bulyko, Ivan. Align- SLM : Textless Spoken Language Models with Reinforcement Learning from AI Feedback. Proceedings of the 63rd Annual Meeting of the A sso... on Pithdoi:10.18653/v1/2025.acl-long.997
- [36] Schick, Timo and Dwivedi-Yu, Jane and Dessi, Roberto and Raileanu, Roberta and Lomeli, Maria and Hambro, Eric and Zettlemoyer, Luke and Cancedda, Nicola and Scialom, Thomas , booktitle =. Toolformer: Language Models Can Teach Themselves to Use Tools , url =. doi:10.52202/07528... on Pithdoi:10.52202/075280-2997
- [37] Yao, Shunyu and Zhao, Jeffrey and Yu, Dian and Du, Nan and Shafran, Izhak and Narasimhan, Karthik and Cao, Yuan , booktitle =. 2023 , url = on Pith
- [38] Ecapa-tdnn: Emphasized channel attention, propagation and aggregation in tdnn based speaker verification , author=. arXiv preprint arXiv:2005.07143 , year= on PithPith's reviewarXiv
- [39] Qin, Yujia and Liang, Shihao and Ye, Yining and Zhu, Kunlun and Yan, Lan and Lu, Yaxi and Lin, Yankai and Cong, Xin and Tang, Xiangru and Qian, Bill and Zhao, Sihan and Hong, Lauren and Tian, Runchu and Xie, Ruobing and Zhou, Jie and Gerstein, Mark and Li, Dahai and Liu, Zhiyu... on Pith
- [40] Patil, Shishir G and Mao, Huanzhi and Yan, Fanjia and Ji, Charlie Cheng-Jie and Suresh, Vishnu and Stoica, Ion and Gonzalez, Joseph E , booktitle =. The. 2025 , publisher = on Pith
- [41] Yao, Shunyu and Shinn, Noah and Razavi, Pedram and Narasimhan, Karthik , booktitle =. 2025 , url = on Pith
- [42] Shao, Zhihong and Wang, Peiyi and Zhu, Qihao and Xu, Runxin and Song, Junxiao and Bi, Xiao and Zhang, Haowei and Zhang, Mingchuan and Li, Y. K. and Wu, Y. and Guo, Daya , year =. 2402.03300 , archivePrefix = on PithPith's reviewarXiv
- [43] Liu, Zichen and Chen, Changyu and Li, Wenjun and Qi, Penghui and Pang, Tianyu and Du, Chao and Lee, Wee Sun and Lin, Min , booktitle =. Understanding. 2025 , url = on Pith
- [44] Yu, Qiying and Zhang, Zheng and Zhu, Ruofei and Yuan, Yufeng and Zuo, Xiaochen and Yue, Yu and Dai, Weinan and Fan, Tiantian and Liu, Gaohong and liu, juncai and Liu, LingJun and Liu, Xin and Lin, Haibin and Lin, Zhiqi and Ma, Bole and Sheng, Guangming and Tong, Yuxuan and Zha... on Pith
- [45] Proximal Policy Optimization Algorithms , author =. 2017 , eprint = on Pith
- [46] Scaling Relationship on Learning Mathematical Reasoning with Large Language Models , author =. 2023 , eprint = on Pith
- [47] Zelikman, Eric and Wu, Yuhuai and Mu, Jesse and Goodman, Noah D. , booktitle =. 2022 , publisher = on Pith
- [48] Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models , author =. Transactions on Machine Learning Research , issn =. 2024 , url = on Pith
- [49] Hu, Edward J. and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , booktitle =. 2022 , url = on Pith
- [50] Explicit Inductive Bias for Transfer Learning with Convolutional Networks , author =. Proceedings of the 35th International Conference on Machine Learning , series =. 2018 , publisher = on Pith
- [51] Decoupled Weight Decay Regularization , author =. International Conference on Learning Representations , year = on Pith
- [52] Yang, An and Li, Anfeng and Yang, Baosong and Zhang, Beichen and Hui, Binyuan and Zheng, Bo and Yu, Bowen and Gao, Chang and Huang, Chengen and Lv, Chenxu and others , year =. 2505.09388 , archivePrefix = on PithPith's reviewarXiv
- [53] Wu, Boyong and Yan, Chao and Hu, Chen and Yi, Cheng and Feng, Chengli and Tian, Fei and Shen, Feiyu and Yu, Gang and Zhang, Haoyang and Li, Jingbei and others , year =. 2507.16632 , archivePrefix = on PithPith's reviewarXiv
- [54] 2025 , howpublished = on Pith
- [55] Du, Zhihao and Chen, Qian and Zhang, Shiliang and Hu, Kai and Lu, Heng and Yang, Yexin and Hu, Hangrui and Zheng, Siqi and Gu, Yue and Ma, Ziyang and Gao, Zhifu and Yan, Zhijie , year =. 2407.05407 , archivePrefix = on PithPith's reviewarXiv
- [56] Du, Zhihao and Wang, Yuxuan and Chen, Qian and Shi, Xian and Lv, Xiang and Zhao, Tianyu and Gao, Zhifu and Yang, Yexin and Gao, Changfeng and Wang, Hui and Yu, Fan and Liu, Huadai and Sheng, Zhengyan and Gu, Yue and Deng, Chong and Wang, Wen and Zhang, Shiliang and Yan, Zhijie... on PithPith's reviewarXiv
- [57] Saeki, Takaaki and Xin, Detai and Nakata, Wataru and Koriyama, Tomoki and Takamichi, Shinnosuke and Saruwatari, Hiroshi , booktitle =. 2022 , doi = on Pith
- [58] Gao, Zhifu and Zhang, ShiLiang and McLoughlin, Ian and Yan, Zhijie , booktitle =. 2022 , doi = on Pith
- [59] Robust Speech Recognition via Large-Scale Weak Supervision , author =. Proceedings of the 40th International Conference on Machine Learning , series =. 2023 , publisher = on Pith
- [60] Rekesh, Dima and Koluguri, Nithin Rao and Kriman, Samuel and Majumdar, Somshubra and Noroozi, Vahid and Huang, He and Hrinchuk, Oleksii and Puvvada, Krishna and Kumar, Ankur and Balam, Jagadeesh and Ginsburg, Boris , booktitle =. 2023 , publisher = on Pith
- [61] 2024 , howpublished = on Pith
- [62] Decoupling Conversational Dynamics in Full-Duplex Spoken Models through Reinforcement Learning , author =. 2026 , eprint = on Pith
- [63] Yu, Wenyi and Wang, Siyin and Chiba, Terumi and Chen, Xianzhao and Tian, Xiaohai and Zhang, Jun and Lu, Lu and Zhang, Chao , year =. 2609.34247 , archivePrefix = on PithPith's reviewarXiv
- [64] Xu, Jin and Guo, Zhifang and He, Jinzheng and Hu, Hangrui and He, Ting and Bai, Shuai and Chen, Keqin and Wang, Jialin and Fan, Yang and Dang, Kai and Zhang, Bin and Wang, Xiong and Chu, Yunfei and Lin, Junyang , year =. 2503.20215 , archivePrefix = on PithPith's reviewarXiv
- [65] Towards Scalable Multi-Domain Conversational Agents: The Schema-Guided Dialogue Dataset , author =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2020 , doi = on Pith
- [66] Lee, Harrison and Gupta, Raghav and Rastogi, Abhinav and Cao, Yuan and Zhang, Bin and Wu, Yonghui , booktitle =. 2022 , doi = on Pith
- [67] Lu, Jiarui and Holleis, Thomas and Zhang, Yizhe and Aumayer, Bernhard and Nan, Feng and Bai, Haoping and Ma, Shuang and Ma, Shen and Li, Mengyu and Yin, Guoli and others , booktitle =. 2025 , month = apr, address =. doi:10.18653/v1/2025.findings-naacl.65 , url = on Pithdoi:10.18653/v1/2025.findings-naacl.65
- [68] Deng, Chong and Ji, Yunjie and Kong, Yuxiang and Li, Xiangang and Li, Xu and Zhang, Binbin and Zhu, Haina and Zhuo, Jianheng , year =. 2609.25195 , archivePrefix = on PithPith's reviewarXiv
- [69] Shen, Yingda and Wang, Yuxiang and Feng, Kunyu and Ni, Qinke and Li, Jiaqi and Hsu, Minghao and Zhang, Junan and Chen, Dekun and Bian, Yutong and Wu, Zhizheng , year =. 2610.11299 , archivePrefix = on PithPith's reviewarXiv
- [70] 2026 , howpublished = on Pith
- [71] Multimodal Duplex Interaction Agent , author =. 2026 , eprint = on Pith
Challenges and discussion
Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.
Come back by email
A short list of the week's newly reviewed papers. Nothing is sent until you confirm from your inbox, and every email has a link that stops it. No account.
No comments yet.
Sign in to comment