“2025-12-22”版本间的差异

来自cslt Wiki
跳转至： 导航、 搜索
 
（7位用户的7个中间修订版本未显示）
第56行： 第56行：
 
||
 
||
 
* ICASSP rebuttal
 
* ICASSP rebuttal
−
* For MT-HuBERT exp,  I'll share the current experimental progress after status report
+
* MT-HuBERT exp
 
||
 
||
 
*
 
*
第67行： 第67行：
 
|Xiaolou Li
 
|Xiaolou Li
 
||
 
||
−
*  
+
* PhD Application
 
||
 
||
 
*  
 
*  
第89行： 第89行：
 
|Pengqi Li
 
|Pengqi Li
 
||
 
||
−
*
+
* Quantitatively measure the reliability of 7 Saliency map methods for speaker verification. (FGGB, SSAB, Corr-RISE, SFAM, three CAMs)
 +
**Conclusion: Corr-RISE and LayerCAM are relatively reliable methods.
 
||
 
||
−
*
+
* Start drafting the report for paper
 
||
 
||
 
*
 
*
第111行： 第112行：
 
|Xiaoxue Luo
 
|Xiaoxue Luo
 
||
 
||
−
*  
+
* paper sharing
 +
* 2-mix experiment of attractor-based SS
 
||
 
||
 
*
 
*
第137行： 第139行：
 
|Jiaying Wang
 
|Jiaying Wang
 
||
 
||
−
*
+
* spk model code finished and training
 
||
 
||
 
*
 
*
第178行： 第180行：
 
|Yang Wei
 
|Yang Wei
 
||
 
||
−
*  
+
* MD experiment for SSL frontend baseline, and some other result analysis for ICCASP rebuttal.[https://z1et6d3xtb.feishu.cn/docx/QGKwdgXoIoWW40xafrTcVUcNnoh?from=from_copylink]
 
||
 
||
 
*
 
*
第213行： 第215行：
 
|Bochao Hu
 
|Bochao Hu
 
||
 
||
−
*  
+
* the latest epoch's per: 33.6%, it will be a better result after training
 +
* preparing for running a whole pipeline
 
||
 
||
 
*
 
*

2025年12月22日 (一) 10:52的最后版本

People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • AI textbook high school grad-2 (6/12)
  • AIGE 2025 seminar
Lantian Li
  • Complete the 2nd review of AI Book (High School).
  • Go on an MoE business trip.
  • Submit the Phase I proposal for the Huang Danian Project.
  • MLA book final check.
Ying Shi
  • ICASSP rebuttal
  • thesis
Zhenghai You
  • Complete the Hubert iterative adaptation Exp, and the result is slightly worse than baseline in clean enroll(~0.9). Trying to test on noisy enroll
Junming Yuan
  • ICASSP rebuttal
  • MT-HuBERT exp
Xiaolou Li
  • PhD Application
Zehua Liu
  • Iterative Decoding Exp for graduation thesis
Pengqi Li
  • Quantitatively measure the reliability of 7 Saliency map methods for speaker verification. (FGGB, SSAB, Corr-RISE, SFAM, three CAMs)
    • Conclusion: Corr-RISE and LayerCAM are relatively reliable methods.
  • Start drafting the report for paper
Tianhao Wang
  • finish ChainSep paper
Xiaoxue Luo
  • paper sharing
  • 2-mix experiment of attractor-based SS
Junhui Chen
  • LLM
    • merge metrics detection module into swarm pipline (done)
    • check some code recently written by ZhangYu
  • finished HUAWEI demo web developing
  • write analysis code for swarm
  • do homework and prepare for some final exams
Jiaying Wang
  • spk model code finished and training
Yu Zhang
  • GPU Util: https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh
  • LLM
    • Completed the implementation and metric collection for the reversal curse experiment; results are still under analysis.
    • Completed the Qwen-based metric collection implementation to support stronger base model performance.
  • help @chenjunhui for Huawei gradio demo deploy
  • Obtain some quantitative results from the Reversal Curse latent learning experiments.
  • Need to work on some homework and prepare for the final exams.
Wenqiang Du
  • Rebuttal the reviewers' comments about high school AI course's draft(done)
  • Some work about project
    • Huawei and AIbabel
Yang Wei
  • MD experiment for SSL frontend baseline, and some other result analysis for ICCASP rebuttal.[1]
Yue Gu
  • write my Phd thesis
Lily
  • AIGE Annual Forum（December 19th）
  • ICASSP author respond
  • Thesis
Bochao Hu
  • the latest epoch's per: 33.6%, it will be a better result after training
  • preparing for running a whole pipeline
Hongcheng Zhang
  • did research for ASU-LLM tasks
取自“http://cslt.org/mediawiki/index.php?title=2025-12-22&oldid=43729”