Weixi Feng
@weixi_feng
Followers
686
Following
631
Media
17
Statuses
138
Final-year CS Ph.D. @UCSB. Ex-intern @NVIDIA/Adobe/Amazon. Working on visual generation and VLMs. Author of StructureDiff., LayoutGPT, BlobGEN-Vid.
Santa Barbara, CA
Joined October 2022
RT @WenhuChen: Happy to announce that we have six papers accepted to ACL 2025. Five of them are main and one is findings. These papers cove….
0
8
0
Our ablation study shows: 1) Masked 3D self-attention greatly helps convergence and object consistency. 2) Continuously improved controllability and prompt-video alignment as we scale up our video dataset. 5/5
0
0
2
Combined with GPT-4o, BlobGEN-Vid outperforms proprietary models like Gen-3 and Kling in multiple aspects of video compositionality. BlobGEN-Vid can also generate multi-view indoor scenes as a single video and surpasses BlobGEN-3D on object consistency. 4/5
1
0
1
BlobGEN-Vid inserts a Masked Spatial Cross-Attention to inject blob tokens to each frame, and a Masked 3D Self-Attention that only allows information flow between the same object in two different frames. It works on both U-Net (VC2) and DiT-based (CogVideoX-5B) models. 3/5
1
0
2
Our blob video representations include per-frame blob parameters and a sparse set of descriptions for each object that aim to characterize semantic change across time. 2/5
1
0
2
I am confused by @arxiv release nowadays. The email seems to imply that all submissions undergo checks that take 2 business days on average. But the schedule and cutoff days on the official website implies same day announcement if submitted before 2pm. If my submission is “on
0
0
2
0
1K
0
RT @WendaXu2: Using racial labels to describe misconduct is harmful and inappropriate. @NeurIPSConf must not condone speech that targets sp….
0
51
0
RT @JiachenLi11: I will be graduating early next year and on the industrial job market! My expertise lies in post-training diffusion models….
0
3
0
RT @m2saxon: 🚨😱Obligatory job market announcement post‼️🤯. I'm searching for faculty positions/postdocs in multimodal/multilingual NLP and….
0
60
0
RT @xwang_lk: 🚀 Since its invention, the mouse has been our way to control computers. But what if it didn’t have to be? 🤔.Thrilled to intro….
0
63
0
RT @AlphaDesignAI: 🚀Introducing ChipAgents: the World's First AI Agent for Chip Design and Verification. Get ready to supercharge your wor….
0
27
0
RT @WilliamWangNLP: New career update: 🚀 Today, we’re thrilled to launch ChipAgents, our most ambitious project. The agentic AI chip design….
0
16
0
RT @XihuiLiu: Excited to share our work, GenArtist: Multimodal LLM as an Agent for Unified Image Generation and Editing, which has been acc….
0
41
0
RT @JiachenLi11: 🚀🚀I'm delighted to share that our paper "Reward Guided Latent Consistency Distillation" has been accepted by @TmlrOrg #TML….
0
5
0
RT @JiachenLi11: Super excited that our paper gets accepted to #neurips2024 @NeurIPSConf ! .Many thanks to my awesome collaborator @weixi_f….
0
4
0
RT @xwang_lk: Our Discffusion paper is officially accepted to TMLR! First time submitting to TMLR and love the constructive reviewing proce….
0
3
0