“2025-12-29”版本间的差异

来自cslt Wiki
跳转至： 导航、 搜索
第212行： 第212行：
 
||
 
||
 
* video to phoneme per: 27%, phoneme to sentence wer: 44%
 
* video to phoneme per: 27%, phoneme to sentence wer: 44%
−
* try to explore the correct mistakes ability of llm: adding noised data(in training)
+
* try to explore the correct mistakes ability of llm: rebulid training data(in training)
 
* study for final exam
 
* study for final exam
 
||
 
||

2025年12月29日 (一) 10:56的版本

People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • All the textbook for primary and middle schools done
  • English version of Middle school handbook almost done (reference needed)
Lantian Li
  • Final review of my MLA book (1/10)
  • Project support for HUAWEI and FYT
  • MoE daily work
Ying Shi
  • Thesis
  • Huawei project
Zhenghai You
  • IRA TSE with SSL[1]
  • training in noisy enroll & Improve the fusion method & Trying more SSL models
Junming Yuan
  • Aug-MT-HuBERT exp(in training)
  • write journal paper(4/7)
Xiaolou Li
Zehua Liu
  • Train Topic and Context Iterative Decode exp for Graduation Thesis
  • Reading some paper for Intership
Pengqi Li
Tianhao Wang
  • supplemental experiments for ChainSep
    • ablation study for iterative separation (only use text)
    • finding some better samples for visualization
Xiaoxue Luo
  • 2-mix experiment of attractor-based SS(Done)
  • 2-5mix USS model for Huawei project(in training)
Junhui Chen
  • LLM
    • discuss the reversal experiment results with @ZhangYu
    • code for swarm metrics analysis (needs further discussion)
  • study for final exam
  • study for final exam
Jiaying Wang
  • spk model recall@k test finished：2mix recall@2=0.9860, 3mix recall@3=0.9176, 4mix recall@4=0.8236
  • prepare spk perceptual data for separation training
Yu Zhang
  • study for final exam...
Wenqiang Du
  • Recording high school AI courses and make related ppt(10/14)
Yang Wei
  • MD fine-tune experiment on AISHELL2 + UY/CH-CHILD (cross validation) [2]
Yue Gu
  • reading some papers about the passive acoustic monitoring（PAM）to understand the current research status and future prospects of this field.
  • writing my Phd thesis
Lily
Bochao Hu
  • video to phoneme per: 27%, phoneme to sentence wer: 44%
  • try to explore the correct mistakes ability of llm: rebulid training data(in training)
  • study for final exam
Hongcheng Zhang
  • study for final exam
  • read some papers
取自“http://www.cslt.org/mediawiki/index.php?title=2025-12-29&oldid=43749”