新闻与公告
- 08/2026. Congratulations to Catherine, Jaywon, Jefferson and Zilin for having their papers HypoExplore, RA-RFT, and Beyond Referring Expressions accepted to Findings and the Main EMNLP 2026 Conference.
- 07/2026. Jefferson 关于 Softmax Advantage Group Estimation(SoftmaxGRPO)的工作被 COLM 2026 接收。
- 04/2026. Zilin 完成了博士学位论文答辩,即将加入 Meta Superintelligence Labs 担任新职位~!
- 03/2026. Moayed 的 ELIT 被 CVPR 2026 接收,他将于今年夏天晚些时候在科罗拉多州丹佛进行展示。
- 01/2026. 三篇论文被 ICLR 2026 接收。Zilin、Jaywon 和 Jefferson 的 ProxyThinker 获得接收。通过与 Meta Superintelligence Labs 的合作,我们还将展示 MetaEmbed;此外,由 Hanjie Chen 的团队主导,我们还有 SportR,这是一个面向体育领域视觉推理的新基准。
- 09/2025. Moayed 关于 Decomposable Flow Matching (DFM) 的工作被 NeurIPS 2025 接收 [DFM]
- 09/2025. Jaywon、Jefferson 和 Moayed 关于使用条件 Fréchet 距离评估文本到图像合成的工作被 WACV 2026 接收 [cFreD]
- 07/2025. Catherine 的 SynGround 论文被 British Machine Vision Conference (BMVC) 接收,该会议今年将在英国谢菲尔德举行。
- 06/2025. ICCV 2025:Moayed 将展示 AV-Link,Jefferson 将展示 Panel-of-Peers,今年的 ICCV 在夏威夷檀香山举行!
- 03/2025. Zilin 与 Czech Technical University 合作完成的长上下文图像重排序工作被 CVPR 2025 接收 [LoCoRe]
近期论文
全部论文 →
SoftmaxGRPO: Learning to Reason using Softmax Advantage Group Estimation
HypoExplore: Agentic Discovery with Active Hypothesis Exploration
Learning to Reason by Analogy via Retrieval-Augmented Reinforcement Fine-Tuning
Beyond Referring Expressions: Scenario Comprehension Visual Grounding
One Model, Many Budgets: Elastic Latent Interfaces for Diffusion Transformers
ProxyThinker: Test-Time Guidance through Small Visual Reasoners
MetaEmbed: Scaling Multimodal Retrieval at Test-Time with Flexible Late Interaction
SportR: A Benchmark for Multimodal Large Language Model Reasoning in Sports
现任及历任赞助方