“2026-06-29”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
 
(2位用户的2个中间修订版本未显示)
第6行: 第6行:
 
|Dong Wang
 
|Dong Wang
 
||
 
||
*
+
* Double check AI handbook (primary/middle school version)
 +
* Work on high-education version
 +
 
 +
 
 
||
 
||
 
*
 
*
第122行: 第125行:
 
|Xiaoxue Luo
 
|Xiaoxue Luo
 
||
 
||
*  
+
* revise the USS-TDA paper
 +
* writing the rebuttal version and supplementing the experimental results for the ChainSep paper with tianhao
 
||
 
||
 
*
 
*
第146行: 第150行:
 
|Hongcheng Zhang
 
|Hongcheng Zhang
 
||
 
||
*   
+
The AED model didnt benefit from the TTA data much
 +
*  learn the music seperatation code of huawei
 
||
 
||
 
*
 
*

2026年6月29日 (一) 11:02的最后版本

People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • Double check AI handbook (primary/middle school version)
  • Work on high-education version


Lantian Li
Wenqiang Du
  • music sep
  • AIbabel
    • New methord for Fewshot, neeed to control FA.
Yang Wei
  • Huilan
    • Search and verify turn detection and barge-in method for voice chat
  • Huawei
    • add reverb for real badcase (speech leak to song).
Yue Gu
Lily
Pengqi Li
Junming Yuan
  • prepare Mid-Term defence
Yu Zhang
  • GPU Util: [1]
  • Read some papers and surveys on self-evolving MAS, organized the development trajectories of different subfields, and considered possible directions I could work on. (Will share some papers on Friday)
Junhui Chen
  • Recent work sharing (soon)
Xiaoxue Luo
  • revise the USS-TDA paper
  • writing the rebuttal version and supplementing the experimental results for the ChainSep paper with tianhao
Bochao Hu
  • on new benchmark(TOPICS), Qwen7B-VL is better than trans/ctc
  • on multi eval: Qwen7B-VL: 32.84%, + ctc head: 31.31%, better than 25baseline: 31.55%
  • revise paper(the final version)
Hongcheng Zhang
  • The AED model didnt benefit from the TTA data much
  • learn the music seperatation code of huawei
Weiman Sun
Shuailong Li