cslt Wiki - 新页面 [zh-cn]

论文列表

2026-02-13T15:33:44Z

Cslt：

*[1] Haoran Sun, Chen Chen, Lantian Li, Dong Wang, CycleFlow: Purify Information Factors by Cycle Loss, Odyssey 2022 (Best Student Paper)
*[2] Haoran Sun, Dong Wang, Lantian Li, Chen Chen, Thomas Fang Zheng, Random Cycle Loss and Its Application to Voice Conversion, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol.45, no.8, August, 2023.
*[3] Chen Chen, Dong Wang, Thomas Fang Zheng, CN-CVS: A Mandarin Audio-Visual Dataset for Large Vocabulary Continuous Visual to Speech Synthesis, ICASSP 2023.
*[4] Chen Chen, Xiaolou Li, Zehua Liu, Lantian Li, Dong Wang, Quantitative Analysis of Audio-Visual Tasks: An Information-Theoretic Perspective, ISCSLP 2024.
*[5] Liu Z, Li X, Chen C, et al. CNVSRC 2024: The Second Chinese Continuous Visual Speech Recognition Challenge[J]. Interspeech 2025
*[6] Chen Chen, Zehua Liu, Xiaolou Li, Lantian Li, Dong Wang, CNVSRC 2023: The First Chinese Continuous Visual Speech Recognition Challenge, Interspeech 2024
*[7] Ying Shi, Lantian Li, Shi Yin, Dong Wang, Jiqing Han, Serialized Output Training by Learned Dominance, Interspeech 2024
*[8] Yuan J, Shi Y, Wang D, et al. MT-HuBERT: Self-Supervised Mix-Training for Few-Shot Keyword Spotting in Mixed Speech[J]. ICASSP 2026.
*[9] Junming Yuan, Ying Shi, Lantian Li, Dong Wang, Askar Hamdulla, Few-Shot Keyword Spotting from Mixed Speech, Interspeech 2024
*[10] Shi, Y., Wang, D.*, Li, L., Han, J., Yin, S. (2023) Spot Keywords From Very Noisy and Mixed Speech. Proc. INTERSPEECH 2023, 1488-1492
*[11] Lin W, Chen J, Wang T, et al. Neural Scoring: A Refreshed End-to-End Approach for Speaker Verification in Complex Conditions[J]. IEEE Signal Processing Letters, 2025.
*[12] Ying Shi, Lantian Li, Dong Wang, Jiqing Han, Keyword Guided Target Speech Recognition, IEEE Signal Processing Letters, 2024
*[13] Zehua Liu, Xiaolou Li, Li Guo, Lantian Li, Dong Wang, Leveraging Large Language Models in Visual Speech Recognition: Model Scaling, Context-Aware Decoding, and Iterative Polishing, APSIPA 2025.
*[14] 陈琛，语音转换任务中的信息解耦优化方法研究，硕士论文，清华大学，2024.
*[15] Liu Z, Li X, Chen C, et al. AlignVSR: Audio-visual cross-modal alignment for visual speech recognition[C]//Proceedings of the 2025 11th International Conference on Communication and Information Processing. 2025: 161-165.
*[16] Cai Y, Li L, Abel A, et al. Maximum Gaussianality training for deep speaker vector normalization[J]. Pattern Recognition, 2024, 145: 109977
*[17] Lantian Li, Ruiqi Liu, Jiawen Kang, Yue Fa, Hao Cui, Yunqi Cai, Ravichander Vipperla, Thomas Fang Zheng and Dong Wang. "CN-Celeb: multi-genre speaker recognition", Speech Communication, 2022.
*[18] Cai, Y., Li, J. & Wang, D. Fast and generalizable micromagnetic simulation with deep neural nets. Nat Mach Intell 6, 1330–1343 (2024).

2026-04-27

2026-01-12T10:44:37Z

Lishuailong：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
*
||
*
||
*
|-

|-
|Lantian Li
||
*
||
*
||
*
|-

|-
|Wenqiang Du
||
*
||
*
||
*
|-

|-
|Yang Wei
||
*
||
*
||
*
|-

|-
|Ying Shi
||
*
||
*
||
*
|-

|-
|Yue Gu
||
*
||
*
||
*
|-

|-
|Lily
||
*
||
*
||
*
|-

|-
|Pengqi Li
||
*
||
*
||
*
|-

|-
|Junming Yuan
||
*
||
*
||
*
|-

|-
|Yu Zhang
||
*
||
*
||
*
|-

|-
|Junhui Chen
||
*
||
*
||
*
|-

|-
|Jiaying Wang
||
*
||
*
||
*
|-

|-
|Bochao Hu
||
*
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*
||
*
||
*
|-

|-
|Weiman Sun
||
*
||
*
||
*
|-
|Ge Gao
||
*
||
*
||
*
|-
|-
|Shuailong Li
||
*
||
*
||
*
|-

2026-04-20

2026-01-12T10:44:26Z

Lishuailong：

2026-04-13

2026-01-12T10:44:17Z

Lily：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
*
||
*
||
*
|-

|-
|Lantian Li
||
* NDRC daily work
* MLA book (3/4)
||
*
||
*
|-

|-
|Wenqiang Du
||
* Baseline testing of multimodal models(ing)
||
*
||
*
|-

|-
|Yang Wei
||
* Train audio separation model for 3 class (speech, song, bird). Dealing with low volume output problem.
* Test streaming AVSR demo (CER: mix: 72%, offline_separation: 20%, streaming_separation: 42%).
||
*
||
*
|-

|-
|Ying Shi
||
* revise my thesis
||
*
||
*
|-

|-
|Yue Gu
||
* write my Phd thesis
||
*
||
*
|-

|-
|Lily
||
* AI handbook check (HK version)
* AIGE related tasks
||
*
||
*
|-

|-
|Pengqi Li
||
* Paper Draft Completion & Revision Plan
||
*
||
*
|-

|-
|Junming Yuan
||
* Preparing the materials for attending ICASSP
* ZH paper draft (need refine)
||
*
||
*
|-

|-
|Yu Zhang
||
* GPU Util: [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh]
* Chain level experiments:
** After introducing the Metric Reward, the weights of correct edges converge faster compared to training with pure reinforcement learning alone.
** The worse the situation when the Metric Reward is introduced (i.e., the lower the weights of critical edges), the more significant the difference compared to not using the Metric Reward.
||
*
||
*
|-

|-
|Junhui Chen
||
* To strengthen the robustness of the conclusions, conducting additional experiments:
** Introduce a new baseline (AgentPrune).
** Add experiments on a new dataset (GSM8K).
** Reproduce the results on other LLM base models.
* Paper writing
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* attractor visualization analysis [https://z1et6d3xtb.feishu.cn/docx/BAoRdM2jQo19krxwH0pcowCsnih]
* The accuracy of attractor counting is lower than expected, may be due to the mixed scenes are complex（2-5mix), retrain the 2-3mix model
||
*
||
*
|-

|-
|Bochao Hu
||
* meet the all requirements and hand over vts pipeline to Sun Chang, waiting for his test
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*test MLLM for aibabel's project
||
*
||
*
|-

|-
|Weiman Sun
||
* supplement our audioset dataset for specific classes
* test large multimodal models
||
*
||
*
|-
|Ge Gao
||
*reproduce spatialnet for speech separation
*write my graduation thesis
||
*
||
*
|-
|-
|Shuailong Li
||
*read some papers
**USE（Sepformer and BSRNN and TDN）
||
*
||
*
|-

2026-04-06

2026-01-12T10:44:10Z

Lishuailong：

2026-03-30

2026-01-12T10:44:01Z

Lilt：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
* AI book University version (1/2)
* English slides cleaning up
* Material preparation for 5.11 world digitial education conference
* Several meetings

||
*
||
*
|-

|-
|Lantian Li
||
* FgW daily work
||
*
||
*
|-

|-
|Wenqiang Du
||
* Complete test Qwen3-omni model
* Trip to Pingyao
||
*
||
*
|-

|-
|Yang Wei
||
* Data cleaning of concert songs
* Trip to Pingyao
||
*
||
*
|-

|-
|Ying Shi
||
* Revise my thesis (almost done)
||
*
||
*
|-

|-
|Yue Gu
||
* write my Phd thesis, reivew some papers
||
*
||
*
|-

|-
|Lily
||
* Journal paper final version submitted
* Material preparation for 5.11 conference
* AIGE2026 conference related tasks
* AIGE Lecturer system testing ，ready for released
||
*
||
*
|-

|-
|Pengqi Li
||
* finished paper draft.[https://z1et6d3xtb.feishu.cn/docx/XTdQdsVRYogVGRxY28Mc5Q9AnCe]
* Traveled to Taiyuan to physical examination(again and pass).
||
*
||
*
|-

|-
|Junming Yuan
||
* MT-HuBERT Journal paper EN(done)
* ICASSP PPT refinement(done)
||
*
||
*
|-

|-
|Yu Zhang
||
* Single Node Behavior experiments append, Currently running Chain level Best Order optimize (final experiment)
* Refactored the overall narrative structure of the paper.
||
*
||
*
|-

|-
|Junhui Chen
||
*
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* test the performance of USS with TDA(transformer decoder-based attractor)
** USS-EDA: 2mix_sdri=8.501, 3mix_sdri=7.372
** USS-TDA: 2mix_sdri=9.616, 3mix_sdri=8.603
* prepare some test and analysis
||
*
||
*
|-

|-
|Bochao Hu
||
* Gongan vts: stoi has improved by using e2e vts model, continue to filter test dataset
* read papers
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*reproduce Dynamic Acoustic Mixer（40%）&email for the author
*inhome data test for qwen3-omni on AED perform not good，shifting a proper way to eval it
*prepare paper share
||
*
||
*
|-

|-
|Weiman Sun
||
* Preparing for the Mid-term report of My Graduation Project:
** Writing the Mid-term Report
** Making adjustments based on the project proposal
||
*
||
*
|-
|Ge Gao
||
*Download concert data for speech separation
*focus on my graduation project
||
*
||
*
|-
|-
|Shuailong Li
||
* Writing the Mid-term Report
*Complete the mid-term defense
*Write a technical report on edge-side models
*finished reviewing the code.
||
*
||
*
|-

2026-03-23

2026-01-12T10:43:48Z

Sunweiman：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
* AIGE book (University version 1/4)
* Keep on NMI paper proof editing

||
*
||
*
|-

|-
|Lantian Li
||
* MoE daily work
||
*
||
*
|-

|-
|Wenqiang Du
||
*AIbabel
** Deployment funasr-nano
** find the reason for the low KWS ACC(current threshold setting strategy is wrong）
**Local deployment the Qwen3 and Testing Qwen's audio-video understanding capabilities（90%）
* Audio separation
** add causal convolution (80/100 epoch)
** add slft-attention (50/100 epoch)
* AIGE
** AI 100 question (60/100)
** High school AI couse done（14/14）
||
*
|-

|-
|Yang Wei
||
* Audio separation stream inference. (sliding window; in progress)[https://z1et6d3xtb.feishu.cn/docx/MUdRd9v0AoenS8xgRrtcR67qnCe]
* Support MDPI rebuttal.
||
*
||
*
|-

|-
|Ying Shi
||
* Revise my thesis according to Prof. Han's suggestions.
* Learn some new skills for my future job
||
*
||
*
|-

|-
|Yue Gu
||
* write my thesis
||
*
||
*
|-

|-
|Lily
||
* Revise thesis
* Journal paper rebuttal
||
*
||
*
|-

|-
|Pengqi Li
||
* Compiled the existing experimental results into a full paper draft.
* Traveled to Taiyuan to attend the interview and take the required physical examination.
||
*
||
*
|-

|-
|Junming Yuan
||
* Preparing the materials for attending ICASSP 2026 (40%)
* ICASSP Oral PPT (done)
* MT-HuBERT Journal paper refinement(5/6)
||
*
||
*
|-

|-
|Yu Zhang
||
* GPU Util [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh]
* Iterative chain-based LLM experiments:
** PKS steadily decreases as useful information accumulates (reflecting reduced uncertainty)
** When no new useful information is introduced, increasing the number of reasoning iterations alone no longer reduces PKS.
||
*
||
*
|-

|-
|Junhui Chen
||
* Reorganized the narrative structure and identified additional experiments to include （objectives and design of the Chain structure and Single Node settings）
||
*
||
*
|-

|-
|Jiaying Wang
||
* thesis modification
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* USS with TDA(transformer decoder-based attractor)
** write code and train the model(in training)
* prepare my resume for Huawei internship
||
*
||
*
|-

|-
|Bochao Hu
||
* GongAn vts
** problem: stoi is difficult to satisfy and test dataset size they want is too large.
** result: cnvsrc-multi is added to train vts, stoi metric becomes better, but it's not good enough.
** solution: more data is used to train vts(CN-CVS+CN-CVS2-P1+CN-CV3)
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*reproduce the TAC（Timestamped Audio Captioning）'s Dynamic Acoustic Mixer（30%）
||
*
||
*
|-

|-
|Weiman Sun
||
* read the paper shared by hongcheng
* fine-tune the multi-talker llm
||
*
||
*
|-
|Ge Gao
||
*convert the MTASS framework to a streaming architecture
*prepare my graduation project
||
*
||
*
|-

2026-03-16

2026-01-12T10:43:40Z

Sunweiman：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
* NMI micromagnetic paper, reformulate theoretical foundation and rewrite the paper.
* AIGE PPT (English version) check, high-school part done.

||
*
||
*
|-

|-
|Lantian Li
||
* MoE daily work
* Examine master's theses
* Write NSFC proposal
||
*
||
*
|-

|-
|Wenqiang Du
||
* New project collaboration
* Some AI course reviews
||
*
||
*
|-

|-
|Yang Wei
||
* Deal with streaming audio visual speech separation.
||
*
||
*
|-

|-
|Ying Shi
||
*
||
*
||
*
|-

|-
|Yue Gu
||
* still write my Phd paper
||
*
||
*
|-

|-
|Lily
||
* Thesis Writing
* CMT System Setup for AIGE 2026
||
*
||
*
|-

|-
|Pengqi Li
||
* Interview preparation
* Experimentation[https://z1et6d3xtb.feishu.cn/docx/U2ahdnEdbolPluxExxWcnmI3nog]
||
*
||
*
|-

|-
|Junming Yuan
||
*
||
*
||
*
|-

|-
|Yu Zhang
||
* GPU Util: [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh]
* LLM: Swarm Lazy & Adv complex topo pruning finish (ECS for Adv Agents, PKS for Lazy Agents)
||
*
||
*
|-

|-
|Junhui Chen
||
* Organizing existing experiments and structuring the logic; identified some issues in the experiments and arguments (currently a bit messy) and continuing to refine.
||
*
||
*
|-

|-
|Jiaying Wang
||
* Thesis modifying
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* USS with EDA(encoder-decoder based attractor)
** train the model with filtered data, test result:2mix_sdri=8.474, 3mix_sdri=7.472(The test results are the same as those of SJTU）
* Huawei project
||
*
||
*
|-

|-
|Bochao Hu
||
* finetuning visual encoder in new domain, which cer is 0.2727
* baseline: VSR-pre-trained Encoder + Qwen_VL finetune only (VL-Decoder) is in training
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*inhome data test，the report will finish on Tuesday
*read some paper
||
*
||
*
|-

|-
|Weiman Sun
||
* debug the lora scripts of MTLLM
** about 20w data records
** takes about 34 hours to complete an epoch
* papers about speech instruction tune shared bay zhenghai and the TAC(Timestamped Audio Captioning) shared by hongcheng.
||
*
||
*
|-
|Ge Gao
||
*cheak some AI courses translation
*read some papers about audio seperation
||
*
||
*
|-

2026-03-09

2026-01-12T10:43:31Z

Hubochao：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
* Keep on two NMI papers

||
*
||
*
|-

|-
|Lantian Li
||
*
||
*
||
*
|-

|-
|Wenqiang Du
||
* Online KWS Model Debug for AIbabel
||
*
||
*
|-

|-
|Yang Wei
||
* Developing Audio visual speech separation streaming demo.
||
*
||
*
|-

|-
|Ying Shi
||
* thesis
||
*
||
*
|-

|-
|Yue Gu
||
*
||
*
||
*
|-

|-
|Lily
||
* Thesis writing
||
*
||
*
|-

|-
|Pengqi Li
||
*
||
*
||
*
|-

|-
|Junming Yuan
||
*
||
*
||
*
|-

|-
|Yu Zhang
||
* GPU Util: [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh]
* LLM: Still Tuning the complex topo loss design (to prune Lazy Agents), find some bugs in loss design with @chenjunhui
||
*
||
*
|-

|-
|Junhui Chen
||
* LLM
** check @ZhangYu's code and debug
** design a game to prove best swarm construction has highest parameter knowledge score (can be used in RL)
||
*
||
*
|-

|-
|Jiaying Wang
||
* Thesis modifying
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* attractor-based USS
** filter out a purer dataset for training(use a SED model to detect the anchor segment in audio clip that is most likely contain a sound event)
* Huawei project
** train the model using VGGSound and data provided by Huawei(in training)
*** 30epoch & 2mix: speech_sisdr = 11.785，music_sisdr = 11.7161. The previous training did have an overfitting problem.
||
*
||
*
|-

|-
|Bochao Hu
||
* CNVSRC.topics dataset pipeline completed and process 1000 data samples for training and testing
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*Evaluating the boundaries of visual capabilities for small and medium-sized models in Qwen3.5
||
*
||
*
|-

|-
|Weiman Sun
||
* papers about the representation analysis to multi-task and instruction-tune
* add the prompt-aware lora module to the MTLLM
||
*
||
*
|-

2026-03-02

2026-01-12T10:43:22Z

Zhanghc：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
* IS 2026 MD paper
* Review two papers on physics: (1) Automatic superconduct material discovery; (2) Large-scale parallel computing for micro magnetics
* Check for English materials of AI handbook

||
*
||
*
|-

|-
|Lantian Li
||
* MoE daily work
* TASLP & IS26 papers
* FYT & HUAWEI project matters
* Review of ongoing work (thesis, research direction, project progress)
||
*
||
*
|-

|-
|Wenqiang Du
||
* check AI course vedio of high school（6/14）
* Translate "AI100 Questions" into English with Gaoge
||
*
||
*
|-

|-
|Yang Wei
||
* MD experiment (3 heads for 3 annotators) for Interspeech.
* Develop streaming AVSR service, in progress.
||
*
||
*
|-

|-
|Ying Shi
||
* thesis(almost finished)
* huawei project
||
*
||
*
|-

|-
|Yue Gu
||
* write the Phd thesis
||
*
||
*
|-

|-
|Lily
||
* INTERSPEECH Submission
* Thesis Writing
||
*
||
*
|-

|-
|Pengqi Li
||
* Coding for multi-label BCE fine-tuning and still training.
||
*
||
*
|-

|-
|Junming Yuan
||
*
||
*
||
*
|-

|-
|Yu Zhang
||
* GPU Util [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh]
* Tunning Complex topo MMLU swarm with Lazy agents (almost done)
||
*
||
*
|-

|-
|Junhui Chen
||
* LLM：
** GAIA test: compare with MMLU test(more information from web, file, picture, etc). Result: metric loss still can detect more useful edges.
** Design experiments to prove metric loss can detect parameter knowledges (almost done)
||
*
||
*
|-

|-
|Jiaying Wang
||
* Thesis Writing
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* write code about USS with TDA(transformer decoder-based attractor)
* organize relevant information on Huawei projects for subsequent delivery to shuailong
||
*
||
*
|-

|-
|Bochao Hu
||
* topic-based VSR-LLM:
** summarize the topic words from our dataset
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*collect recent research about audio and visual omni LLM
||
*
||
*
|-

|-
|Weiman Sun
||
*improved the graduation project
||
*
||
*
|-

2026-02-23

2026-01-12T10:43:14Z

Lilt：

2026-02-16

2026-01-12T10:42:58Z

Lilt：

2026-02-09

2026-01-12T10:42:40Z

Guyue：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
*
||
*
||
*
|-

|-
|Lantian Li
||
*
||
*
||
*
|-

|-
|Wenqiang Du
||
*
||
*
||
*
|-

|-
|Yang Wei
||
*
||
*
||
*
|-

|-
|Ying Shi
||
*
||
*
||
*
|-

|-
|Yue Gu
||
* write my thesis and make the photo album
||
*
||
*
|-

|-
|Lily
||
*
||
*
||
*
|-

|-
|Pengqi Li
||
*
||
*
||
*
|-

|-
|Junming Yuan
||
*
||
*
||
*
|-

|-
|Yu Zhang
||
*
||
*
||
*
|-

|-
|Junhui Chen
||
*
||
*
||
*
|-

|-
|Jiaying Wang
||
*
||
*
||
*
|-

|-
|Bochao Hu
||
*
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*
||
*
||
*
|-

|-
|Weiman Sun
||
*
||
*
||
*
|-

2026-02-02

2026-01-12T10:42:24Z

Shiying：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
* English version for Middle School AI Textbook done.
* AI textbook for old people
* Initial draft for 2nd MD paper.
||
*
||
*
|-

|-
|Lantian Li
||
* MoE daily work (busy week)
* FYT project delivery
||
* Go on the final review of my MLA book (6/10)
||
*
|-

|-
|Wenqiang Du
||
* Collaboration project acceptance (Tencent)
* Year end summary(AIbabel)
||
*
||
*
|-

|-
|Yang Wei
||
* Train MD model with Chinese pretrained front end with cross attention architecture. (Not test yet)
* Trying to figure out actual meaning of existed annotation data with help from xianyu [https://gcndy0j2mmt7.feishu.cn/wiki/LOZKwIQiTiQZdKkJlzEcuOl6n6c]
||
*
||
*
|-

|-
|Ying Shi
||
* Thesis
||
*
||
*
|-

|-
|Yue Gu
||
* write my Phd thesis and collect the picture
||
*
||
*
|-

|-
|Lily
||
* Organized course materials production （小初高分册+ Middle school English version）
* Participated in reviewing the AI handbooks
* AIGE 2026 meeting-related tasks
||
*
||
*
|-

|-
|Pengqi Li
||
* Currently teaching at the AI Winter Camp. I will be fully back to my research in Friday.
||
*
||
*
|-

|-
|Junming Yuan
||
* checking EN middle school AI handbook(M5-M7)
* learn-not-to-listen MT-HuBERT:
** setting a threshold can improve performance on clean-speech tasks(ASR, MT-HuBERT:9.22 --> LNTL-MT-HuBERT: 8.43)
** there is some performance degradation on mixed-speech tasks(SD, MT-HuBERT:3.93 --> LNTL-MT-HuBERT: 4.31).
* draft of the journal paper(almost done)
||
*
||
*
|-

|-
|Yu Zhang
||
* GPU Util: [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh?from=from_copylink]
* LLM: Finish complex topo code writing, and build swarm with copying node inside it (experiment still running)
||
*
||
*
|-

|-
|Junhui Chen
||
* LLM: build swarm for MMLU exp with different LLM base model in FinalDecision: exchange qwen2.5 to llama3-8b
||
*
||
*
|-

|-
|Jiaying Wang
||
* spk order experiment finished, 4mix result are under test.
* loudness & content order verification done.
* spk order verification code preparing(2/3).
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* 2-5mix multi_head separation model for Huawei project
** Modifications：
*** remove data that the mixed audio equals the target audio
*** add sing data to the speech category
*** add MSE_loss to the original SISDR_loss to control the energy amplitude of output audio
** Result: compared to the previous model, the performance is even worse
** Reasons：
*** add sing data to the speech branch is not feasible
*** there are some issues with the istft function in the original paper code
||
*
||
*
|-

|-
|Bochao Hu
||
* stage2 P2S: finetune with NLLB model using syn data, result is nearly same as before
* read papers
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*prepare for weekly report
*read some new paper
*enrich my train data
||
*
||
*
|-

|-
|Weiman Sun
||
* relevant papers
* hook the first token after the prompt to observe the model's understanding of the task at different depths.
||
*
||
*
|-

2026-01-26

2026-01-05T10:34:12Z

Lily：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
* AI middle school hand book / English version almost completed.
||
*
||
*
|-

|-
|Lantian Li
||
* Recheck AI textbooks for senior middle school.
* MoE daily work (busy week)
||
* Go on the final review of my MLA book (6/10)
||
*
|-

|-
|Wenqiang Du
||
* Project work about AIbabel
** Upgraded online model
** New demand technology survey
||
*
||
*
|-

|-
|Yang Wei
||
* Analyze uy/ch_child data annotation and exp result.
* Test the av speech separation model with a dataset with more speakers.
||
*
||
*
|-

|-
|Ying Shi
||
* Thesis
* Huawei project
||
*
||
*
|-

|-
|Yue Gu
||
* write my thesis
||
*
||
*
|-

|-
|Lily
||
* Organized course materials production （小初高分册）
* Participated in reviewing the AI handbooks
* Miyun 3-high school AI report
||
*
||
*
|-

|-
|Pengqi Li
||
* Experimental analysis revealed the unreliability of the insertion and deletion metrics adapted from face verification.[https://z1et6d3xtb.feishu.cn/docx/BZxUdDC1YoOSXwxQCrycW9Ymnqf]
* Proposed and implemented a novel framework for assessing insertion and deletion.(still test)
||
*
||
*
|-

|-
|Junming Yuan
||
* learn-not-to-listen MT-HuBERT
** PR(PER%), LNTL-MT-HuBERT: 7.24%, Clean-HuBERT: 5.41%, WavLM: 4.84%
** ASR(WER%), LNTL-MT-HuBERT: 8.43%, Clean-HuBERT: 6.42%, WavLM: 6.21%
** SD(DER%), LNTL-MT-HuBERT: 4.31%, Cocktail-HuBERT: 4.07%, MT-HuBERT: 3.93%
* revise ICASSP final paper
||
*
||
*
|-

|-
|Yu Zhang
||
* GPU Util: [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh]
* LLM
** Completed the code for Swarm MMLU metric-based optimization together with @chenjunhui
** Still working on modifying the Swarm topology code (for more complex swarm topology)
||
*
||
*
|-

|-
|Junhui Chen
||
* (caught a cold)
* MMLU
** difference in the ECS distribution between good nodes and bad nodes.
** metrics loss helps select high-quality nodes and eliminate low-quality ones. [https://z1et6d3xtb.feishu.cn/wiki/WIbewlbcwihGoCkChpZcwZpEnnh?from=from_copylink]

||
*
||
*
|-

|-
|Jiaying Wang
||
* spk order separation：2-mix，3-mix training finished，testing； 4-mix are training
* loudness order verification done
* content order verification 1/3, will be finished these days
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* prepare 2mix speech separation results for ASR test
* 2-5mix multi_head separation model for Huawei project
** data: remove data that the target audio is mixed audio(input equals output, model does not learn)
** loss: add MSE_loss to the original SISDR_loss to control the energy amplitude of output audio
||
*
||
*
|-

|-
|Bochao Hu
||
* read papers
* prepare the weekly report
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*made zero-shot on clotho and Audiocaps
**result：ROUGE-L（0.2931，0.3998），CIDEr（0.0986，0.2312），SPIDEr（0.0800，0.1762），（clotho，Audiocaps）
||
*
||
*
|-

|-
|Weiman Sun
||
*Ran a completely opposite ASR task on the same data
**Consistent with last week's result
*Analyze different models
||
*
||
*
|-

2026-01-19

2026-01-05T10:33:43Z

Lilt：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
* Recheck AI textbooks for primary and middle schools.
||
*
||
*
|-

|-
|Lantian Li
||
* Final review of my MLA book (6/10, hold...)
* MoE daily work (busy week)
* My Prof. Title defense
||
*
||
*
|-

|-
|Wenqiang Du
||
* Statistics the accounts for companies, our laboratory and AIGE.
* Check AI textbooks
||
*
||
*
|-

|-
|Yang Wei
||
* Analyze MD experiment with 3 annotator labels [https://z1et6d3xtb.feishu.cn/docx/YGHjdXhHIouqZZxZuYmcnOJnnJb?from=from_copylink]
* Develop streaming audio visual speech separation, still with latency
||
*
||
*
|-

|-
|Ying Shi
||
* Thesis
* Huawei multi-talker ASR project(Mandarin 2-mix test with FunASR-Nano-800M)
** ground-truth WER 3.89%
** separated WER 12.11%
||
*
||
*
|-

|-
|Yue Gu
||
* write my Phd thesis
* the seminar
||
*
||
*
|-

|-
|Lily
||
* Participated in reviewing the AI handbooks (primary/middle/high school)
* Compiled course materials for the AI Handbook (Senior Edition)
||
*
||
*
|-

|-
|Pengqi Li
||
* Writing the Experiments section.
** Focus on hard-to-explain results by formulating hypotheses and conducting deeper analysis.
||
*
||
*
|-

|-
|Junming Yuan
||
* Aug-MT-HuBERT(failed)
** Continued pre-training for 600K steps. there is still no improvement observed on clean-speech tasks.
*** Aug-MT-HuBERT has more substitution errors in clean ASR adaptation.
* Incorporate a "learn-not-to-listen" mechanism into MT-HuBERT and retrain the backbone. (in progress)
** At 425K steps, PR(PER%): 8.18%, ASR(WER%): 9.32%, SD(DER%): 5.05%
* middle school AI textbook checking(picture, table)
||
*
||
*
|-

|-
|Yu Zhang
||
* GPU Util: [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh]
* LLM
** Rewrite the graph connectivity of Swarm MMLU (the original topology does not align with our expectations).
** Add context segmentation logic to the ECS computation, so that we can trace which preceding node a given text segment originates from.
||
*
||
*
|-

|-
|Junhui Chen
||
* LLM
** Get swarm crosswords experiment with metric detection working
*** because crosswords has more different role characters than MMLU, so it is necessary
** Check and find some bugs in code of @ZhangYu
** Implement multi-process of LLM instances in experiments
*** p.s.: try multi-threading, fails because a single Python interpreter only initializes one CUDA context, causing resource contention between different instances.
||
*
||
*
|-

|-
|Jiaying Wang
||
* loudness verification 2/3
* spk experiment 3-mix testing, 2-mix & 4-mix training
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* 2-5mix multi_head separation model for Huawei project [https://z1et6d3xtb.feishu.cn/docx/JtLidXiOVoQUedx3h5IcQOwAnFh]
* prepare 2mix speech separation results for ASR test
||
*
||
*
|-

|-
|Bochao Hu
||
* read some papers
* refactor P2S code, using ms-swift and adding n-best seq to input.
||
*
||
*
|-

|-
|Hongcheng Zhang
||
*Complete the training of short audio descriptions for AudioSet-SL in WavCaps
**METEOR: 0.3373， ROUGE-L: 0.3121， BLEU-4 : 0.0769， CIDEr : 0.6207
||
*
||
*
|-

|-
|Weiman Sun
||
* Label audioset video dataset
* Analyze the representation of MT-LLM inference and review relevant literature
||
*
||
*
|-

2026-01-12

2026-01-05T10:33:10Z

Zhaomy：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
* Double check English version of high-scholl handbook.
* Check English version of middle-school handbook.
||
*
||
*
|-

|-
|Lantian Li
||
* Final review of my MLA book (6/10)
* MoE daily work
||
*
||
*
|-

|-
|Wenqiang Du
||

* Complete recording high school AI courses（14/14）
* Creat PPT for high school AI course (3/14)
||
*
||
*
|-

|-
|Yang Wei
||
* Test for audio visual speech separation + ASR (offline script), with 200 2mix speech video, CER: 20%.
* Chinese mispronunciation detection experiment, with Chinese Hubert as feature. (precision: 0.18, recall: 0.72)
||
*
||
*
|-

|-
|Ying Shi
||
* Thesis
* Some stuff about HUAWEI project
||
*
||
*
|-

|-
|Yue Gu
||
* Revise the Phd thesis structure
* Seminar. In an anonymous vote of 20 people, 75% chose to freely discuss.
||
*
||
*
|-

|-
|Lily
||
* Check English version of middle-scholl handbook
* Check English version of high-scholl handbook
* Organized course materials production （小初高分册）
||
*
||
*
|-

|-
|Pengqi Li
||
* Drafting the method part of paper.
* Identified bugs in the reproduction code. Re-ran experiments and confirmed that conclusions remain consistent.
* Assisting with the revision of the middle school handbook.
||
*
||
*
|-

|-
|Junming Yuan
||
* Aug-MT-HuBERT:
** Based on last week’s best configuration, Continued pre-training for 300K steps.
***No improvement was observed on clean-speech tasks.
** After inspection, found that the pre-training inherited a low lr from the previous model after 1.6M steps.
***After increase the lr and retraining for 200K steps, there was still no improvement on clean-speech tasks.
***200K steps, PR(PER): 8.14, ASR(WER): 8.93
* SS Adaptation:
** following the SA-WavLM strategy, further evaluated MT-HuBERT under low-resource settings (10% and 1% of the training data).
*** 10% data: Cocktail(11.29) > MT-HuBERT(11.07) > WavLM(10.81)
*** 1% data: Cocktail(8.56) > MT-HuBERT(8.43) > WavLM(8.11)
* Draft Paper writing(EN version almost done)
||
*
||
*
|-

|-
|Yu Zhang
||
* GPU Util: [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh]
* Finish final exam
* Writing code to analyze LLM Swarm metrics, mainly looking at how ECS/PKS correlate with the optimized edge probability.
||
*
||
*
|-

|-
|Junhui Chen
||
* finish final exam
* debug code about swarm minicrossword test
||
*
||
*
|-

|-
|Xiaolou Li
||
*
||
*
||
*
|-

|-
|Jiaying Wang
||
* spk model training(130/300):reconstruct data, now training data is aligned with separation data
** recall@k of 130 epoch: 2mix recall@2=0.9799, 3mix recall@3=0.9101, 4mix recall@4=0.8247

||
*
||
*
|-

|-
|Tianhao Wang
||
* revised ChainSep paper
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* 2-5mix multi_head separation model for Huawei project
** write code for the multi-speaker and multi-sound events separation task，complete data preparation and feature extraction
** adjust the model structure to three heads(speech, music and others), the model is still in training, and current val_sisdr is 11.38
||
*
||
*
|-

|-

|Bochao Hu
||
* finish final exam
* debug and train vsr E2E model, still in training
* resently results [https://z1et6d3xtb.feishu.cn/wiki/MQS9wt7gYikboBkJ09ScovGxnmg?from=from_copylink]
||
*
||
*
|-

|-
|Hongcheng Zhang
||
* finish final exam
* debug for asu-llm code and train the task audio caption with WavCaps(1/20 epoches)
||
*
||
*
|-

Work-status-2025

2026-01-05T10:16:42Z

Zhaomy：以“2025-12-29 2025-12-22 2025-12-15 2025-12-08 2025-12-01 2025-11-24 2025-11-17 2025-11-10 2025-11-03 2025-10-27 202...”为内容创建页面

[[2025-12-29]]

[[2025-12-22]]

[[2025-12-15]]

[[2025-12-08]]

[[2025-12-01]]

[[2025-11-24]]

[[2025-11-17]]

[[2025-11-10]]

[[2025-11-03]]

[[2025-10-27]]

[[2025-10-20]]

[[2025-10-13]]

[[2025-10-06]]

[[2025-09-29]]

[[2025-09-22]]

[[2025-09-15]]

[[2025-09-08]]

[[2025-09-01]]

[[2025-08-25]]

[[2025-08-18]]

[[2025-08-11]]

[[2025-08-04]]

[[2025-07-28]]

[[2025-07-21]]

[[2025-07-14]]

[[2025-07-07]]

[[2025-06-30]]

[[2025-06-23]]

[[2025-06-16]]

[[2025-06-09]]

[[2025-06-02]]

[[2025-05-26]]

[[2025-05-19]]

[[2025-05-12]]

[[2025-05-05]]

[[2025-04-28]]

[[2025-04-21]]

[[2025-04-14]]

[[2025-04-07]]

[[2025-03-31]]

[[2025-03-24]]

[[2025-03-17]]

[[2025-03-10]]

[[2025-03-03]]

[[2025-02-24]]

[[2025-02-17]]

[[2025-02-10]]

[[2025-02-03]]

[[2025-01-27]]

[[2025-01-20]]

[[2025-01-13]]

[[2025-01-06]]

2025-01-26

2026-01-05T10:16:02Z

Duwenqiang：以“{| class="wikitable" !People !! This Week !! Next Week !! Task Tracking (DeadLine) |- |- |Dong Wang || || * || * |- |- |Lantian Li || |...”为内容创建页面

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||

||
*
||
*
|-

|-
|Lantian Li
||

||
*
||
*
|-

|-
|Ying Shi
||

||
*
||
*
|-

|-

|Junming Yuan
||

||
*
||
*
|-

|-
|Xiaolou Li
||

||
*
||
*
|-

|-
|Pengqi Li
||

||
*
||
*
|-

|-
|Tianhao Wang
||

||
*
||
*
|-

|-
|Xiaoxue Luo
||

||
*
||
*
|-

|-
|Junhui Chen
||

||

||
*
|-

|-
|Jiaying Wang
||

||
*
||
*
|-

|-
|Yu Zhang
||

||

||
*
|-

|-
|Wenqiang Du
||

||
*
||
*
|-

|-
|Yang Wei
||
||
*
||
*
|-

|-
|Yue Gu
||
||
*
||
*
|-

|-
|Lily
||
||
*
||
*
|-

|-
|Bochao Hu
||
||
*
||
*
|-

|-
|Hongcheng Zhang
||

||
*
||
*
|-

2025-01-19

2026-01-05T10:15:53Z

2025-01-12

2026-01-05T10:15:46Z

2026-01-05

2026-01-05T10:14:56Z

Lilt：

{| class="wikitable"
!People !! This Week !! Next Week !! Task Tracking (DeadLine)
|-

|-
|Dong Wang
||
* English version of the AI handbook - middle/high school almost done

||
*
||
*
|-

|-
|Lantian Li
||
* Final review of my MLA book (3/10)
* MoE daily work
||
*
||
*
|-

|-
|Wenqiang Du
||
* Check AI course PPT(4/14)
* Completed AI100 question task and will be uploaded to the platform this month
||
*
||
*
|-

|-
|Yang Wei
||
* Some additional result analysis and some revisions on manuscript according to the reviewer' comments. [https://z1et6d3xtb.feishu.cn/docx/Woz0dFwsaoWZuOxZ2KGcm9MxnKe?from=from_copylink][https://z1et6d3xtb.feishu.cn/docx/VIZadovoZohLscxQUYNcu59HnGd?from=from_copylink]
||
*
||
*
|-

|-
|Ying Shi
||
*
||
*
||
*
|-

|-
|Pengqi Li
||
* Draft Paper: The framework and core content are complete.
* AI Handbook: Checking and verifying references.
* Winter Camp Preparation: Preparing materials for the Tsinghua University Middle School AI Winter Camp.
||
*
||
*
|-

|-
|Junming Yuan
||
* Aug-MT-HuBERT:
** Fixed the pre-training to 100K steps, compared 3 noise ratios（50%, 80%, and 100%）and 2 different SNR sampling strategies([-3, 3] uniform / [0, 10] Gauss).
*** Best setting, PR performance(PER): 8.12 -> 8.08, ASR performance(WER): 9.22 -> 9.07
*** There is still around 3% gap compared to HuBERT.
* SS adaptation(SI-SDRi):
** Cocktail-HuBERT(14.76) > MT-HuBERT(13.93) > WavLM(13.83) > ConvTasNet(10.39)
* Draft Paper( CN version almost done)
||
*
||
*
|-

|-
|Yu Zhang
||
* GPU Util [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh?from=from_copylink]
* working on the big assignment and preparing for the final exam.
||
*
||
*
|-

|-
|Junhui Chen
||
* prepare for final exam
||
*
||
*
|-

|-
|Xiaolou Li
||
* Prepared applications for 4 CDT programmes
||
*
||
*
|-

|-
|Jiaying Wang
||
* spk order separation training
||
*
||
*
|-

|-
|Tianhao Wang
||
* additional experiments for ChainSep [https://z1et6d3xtb.feishu.cn/docx/RG5Qd9ashoGhw5xc9M2cAVcHnLc]
||
*
||
*
|-

|-
|Xiaoxue Luo
||
* 2-5mix attractor-based USS model for Huawei project
** SI-SDR : 2mix: 4.725， 3mix: 0.866， 4mix: -1.155， 5mix: -2.709
* design a better solution for Huawei project
** multi-head separation (two groups: speakers and events)
||
*
||
*
|-

|-
|Yue Gu
||
* write Phd defense slides
||
*
||
*
|-

|-
|Lily
||
*
||
*
||
*
|-

|-
|Bochao Hu
||
* P2S wer: 45%.data:20% GT 20% real seq 60% synthetic seq. llm can't correct with high per(> 30%)
||
*
||
*
|-

|-
|Hongcheng Zhang
||
* finish the main code of ASU-LLM (CLAP + HuBERT). The core architecture consists of a dual-stream encoder combining CLAP and Audio Hubert, integrated with LLaMA
||
*
||
*
|-