update

Author: YIGE24

_bibliography/papers.bib

@@ -161,20 +161,6 @@ @inproceedings{shao2025holitom
      selected = {true}
 }
 
-@inproceedings{zhang2025poison,
-  bibtex_show = {true},
-         abbr = {NeurIPS},
-        title = {Poison as Cure: Visual Noise for Mitigating Object Hallucinations in LVMs},
-       author = {Zhang, Kejia and Tao, Keda and Tang, Jiasheng and Wang, Huan},
-    booktitle = {NeurIPS},
-         year = {2025},
-        arxiv = {2501.19164},
-      website = {https://kejiazhang-robust.github.io/poison-cure-lvm/},
-         code = {https://github.com/KejiaZhang-Robust/VAP},
-      preview = {vap.png},
-     selected = {true}
-}
-
 @inproceedings{li2025freqexit,
   bibtex_show = {true},
          abbr = {NeurIPS},

_data/coauthors.yml

@@ -69,8 +69,6 @@
     url: https://hanzhangshen03.github.io
 
 "zhang":
-  - firstname: ["Kejia"]
-    url: https://kejiazhang-robust.github.io/homepage
   - firstname: ["Yitian"]
     url: https://bespontaneous.github.io/homepage/
 

_news/20250926-NeurIPS-4Papers.md

@@ -5,10 +5,8 @@ inline: true
 related_posts: false
 ---
 
-`[NeurIPS'25]` **Four papers** accepted to NeurIPS 2025 in the field of efficient and reliable AI. Congrats to our students and collaborators! Three papers from our lab:
+`[NeurIPS'25]` **Four papers** accepted to NeurIPS 2025 in the field of efficient and reliable AI. Congrats to our students and collaborators! Two papers from our lab:
 
 - **HoliTom**: A top-performing video LLM token compression method that maintains 99.1% performance while reducing FLOPs to just 6.9%—and it's training-free! [[arxiv](https://arxiv.org/abs/2505.21334)] [[code](https://github.com/cokeshao/HoliTom)] [[webpage](https://cokeshao.github.io/HoliTom_Web/)]
 
-- **Poison as Cure**: Is adversarial visual noise always harmful like "poison"? We show it can actually cure the hallucination problem in VLMs. [[arxiv](https://arxiv.org/abs/2501.19164)] [[code](https://github.com/KejiaZhang-Robust/VAP)] [[webpage](https://kejiazhang-robust.github.io/poison-cure-lvm/)]
-
 - **FreqExit**: A dynamic inference framework for visual autoregressive (VAR) models via early exit with novel frequency-aware guidance. [[openreview](https://openreview.net/pdf?id=DUlZTgLkeh)] [[code](https://github.com/NeuraLiying/FreqExit)] [[webpage](https://neuraliying.github.io/FreqExit/)]

_pages/Home.md

@@ -30,7 +30,7 @@ The <span style="color: var(--global-theme-color); font-weight: 500">ENCODE Lab<
 
 **Research Focus:** <br>
 Our research centers on **Efficient AI**, spanning multiple domains in vision and language:
-- **LLMs/MLLMs**: [[DyCoke](https://github.com/KD-TAO/DyCoke), [HoliTom](https://github.com/cokeshao/HoliTom), [OmniZip](https://arxiv.org/abs/2511.14582), [MergeMix](https://arxiv.org/abs/2510.23479), [StreamingTOM](https://arxiv.org/abs/2510.18269), [RLKV](https://arxiv.org/abs/2510.08525), [VidKV](https://arxiv.org/abs/2503.16257), [ResSVD](https://arxiv.org/abs/2505.20112), [SparseSSM](https://arxiv.org/abs/2506.09613), [VAP](https://kejiazhang-robust.github.io/poison-cure-lvm/), [ReasonMap](https://fscdc.github.io/Reason-Map/), [RewardMap](https://arxiv.org/abs/2510.02240), [Survey](https://github.com/cokeshao/Awesome-Multimodal-Token-Compression)]
+- **LLMs/MLLMs**: [[DyCoke](https://github.com/KD-TAO/DyCoke), [HoliTom](https://github.com/cokeshao/HoliTom), [OmniZip](https://arxiv.org/abs/2511.14582), [MergeMix](https://arxiv.org/abs/2510.23479), [StreamingTOM](https://arxiv.org/abs/2510.18269), [RLKV](https://arxiv.org/abs/2510.08525), [VidKV](https://arxiv.org/abs/2503.16257), [ResSVD](https://arxiv.org/abs/2505.20112), [SparseSSM](https://arxiv.org/abs/2506.09613), [ReasonMap](https://fscdc.github.io/Reason-Map/), [RewardMap](https://arxiv.org/abs/2510.02240), [Survey](https://github.com/cokeshao/Awesome-Multimodal-Token-Compression)]
 - **Image Generation** (diffusion/AR models): [[OBS-Diff](https://arxiv.org/abs/2510.06751), [VideoPruning](https://arxiv.org/abs/2411.18375), [ARPG](https://hp-l33.github.io/projects/arpg), [FreqExit](https://neuraliying.github.io/FreqExit/)]
 - **3D Reconstruction**: [[Niagara](https://ai-kunkun.github.io/Niagara_page/)]
 - **Snapshot Compressive Imaging**: [[QuantizedSCI](https://github.com/mcao92/QuantizedSCI), [MobileSCI](https://github.com/mcao92/MobileSCI)]

_pages/people.md

@@ -404,18 +404,6 @@ nav_order: 2
     <p class="profile-description">BS Cambridge <br> SWE Intern at Meta</p>
     </div>
   </div>
-  <div class="profile-item">
-    <img class="profile-image" src="../assets/img/people/kejia_zhang.png" alt="Profile"/>
-    <div class="profile-content">
-
-      <h3 class="profile-name">Kejia Zhang</h3>
-
-      <div class="profile-social">
-      <a href="https://kejiazhang-robust.github.io/homepage" title="Homepage"><i class="fa-solid fa-house"></i></a>
-    </div>
-    <p class="profile-description">MS Xiamen Univ</p>
-    </div>
-  </div>
   <div class="profile-item">
     <img class="profile-image" src="../assets/img/people/lingcheng_kong.jpg" alt="Profile"/>
     <div class="profile-content">