“2026-05-18”版本间的差异

来自cslt Wiki
跳转至： 导航、 搜索
 
（3位用户的3个中间修订版本未显示）
第6行： 第6行：
 
|Dong Wang
 
|Dong Wang
 
||
 
||
−
*
+
 
 +
* Revise AI textbook for primary school (3/4)
 +
 
 +
 
 +
 
 
||
 
||
 
*
 
*
第111行： 第115行：
 
|Yu Zhang
 
|Yu Zhang
 
||
 
||
 +
* GPU Util: [https://z1et6d3xtb.feishu.cn/wiki/XX4NwX3tJiBDcgkMi0hcFUtInHh]
 
* Check MLA book code and translate (sec 1-6)
 
* Check MLA book code and translate (sec 1-6)
 
||
 
||
第145行： 第150行：
 
|Bochao Hu
 
|Bochao Hu
 
||
 
||
−
*  
+
* topic-based vsr-llm:
 +
** build topic-category-keyword dataset structure
 +
** to verify peripheral info is useful for other llm(ablation): FT qwen3-VL-8B, qwen3.5-9B(in training)
 +
* cncvs3: FT avhubert + bi-trans/ctc in cncvs+cnvsrc
 +
** single cer: 44%(24baseline: 41%), multi cer: 48%(24baseline: 52%)
 
||
 
||
 
*
 
*
第156行： 第165行：
 
|Hongcheng Zhang
 
|Hongcheng Zhang
 
||
 
||
−
*   
+
*  Fine-tuning the FineLAP didn't work,trying to adopt data augement & change train loss
 +
*  read some paper about Contrastive Learning
 
||
 
||
 
*
 
*

2026年5月18日 (一) 11:00的最后版本

People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • Revise AI textbook for primary school (3/4)


Lantian Li
Wenqiang Du
  • Huawen separation project
    • speech sep：Huawei chip don't support Memba module computation.so replace it with LSTM .
    • audio sep: vocal and accoompainment separation.
Yang Wei
  • On leave.
Ying Shi
Yue Gu
  • continue revise the Phd thesis
Lily
  • AIGE2026
  • German delegation
  • Routine AIGE tasks
Pengqi Li
  • Completed the draft of the SV-XAI paper
  • Arranged the framework of the phd thesis.
Junming Yuan
  • collating and read some ICASSP papers
  • AI Primary School English PPT for AI radiance(5/44)
Yu Zhang
  • GPU Util: [1]
  • Check MLA book code and translate (sec 1-6)
Junhui Chen
  • Check MLA book code and translate
Xiaoxue Luo
  • USS-tda experiments[2]
  • paper writing(method section done)
Bochao Hu
  • topic-based vsr-llm:
    • build topic-category-keyword dataset structure
    • to verify peripheral info is useful for other llm(ablation): FT qwen3-VL-8B, qwen3.5-9B(in training)
  • cncvs3: FT avhubert + bi-trans/ctc in cncvs+cnvsrc
    • single cer: 44%(24baseline: 41%), multi cer: 48%(24baseline: 52%)
Hongcheng Zhang
  • Fine-tuning the FineLAP didn't work,trying to adopt data augement & change train loss
  • read some paper about Contrastive Learning
Weiman Sun
  • Read some papers about Acoustic Synthesis and Editing.
  • Completed my graduation thesis
Ge Gao
  • prepare for graduation thesis defense
  • organize audio separation data
Shuailong Li
  • Finished the thesis defense draft
  • Look at the TDA code
取自“http://index.cslt.org/mediawiki/index.php?title=2026-05-18&oldid=44161”