When your LM decoder overpowers your VLM
Wow Priya must be a very strong girl
Ravi Kiran S
@vikataravi
Associate Prof, IIIT-H, India | Vision@BharatGen | iHub-Data | Alum: UW, IISc | Enjoy interdisciplinary problems involving multimedia | Meme fan | 🇮🇳
- 🚀 Segment Anything (SAM)... for Documents... is here! 📄✨ 💬 Prompt your document: 🔹 "Segment the radar plot" 🔹 "Mask all personal information" 🔹 "Give me hierarchical layout" 🎉 To be presented at #ECCV2026 ! Stay tuned ! 🔥 #ECCV @CVIT_IIIT_Hyd @iiit_hyderabad
- Sometimes... x really is 5. 😄📐👀 Teasing our first #ECCV2026 paper: 🚀 DoCoG: Mask-based Multi-Type Grounded Chain-of-Thought for Document QA Precise mask grounding for all ! 📝 Text 📊 Charts 🔀 Flowcharts More soon! #ECCV @CVIT_IIIT_Hyd @iiit_hyderabad
- Our unique 2⃣D representation enables occlusion-aware 3⃣D control for text-to-image generation ! 🧊 seethrough3d.github.io Come to poster 3⃣7⃣0⃣ in poster session 4 today (4.45pm) and find out how @CVPR #CVPR2026 @iiit_hyderabad @CVIT_IIIT_Hyd @val_iiscExcited to share SeeThrough3D: Occlusion Aware 3D Control in Text-to-Image Generation, accepted to #CVPR2026 🎉🎉. See🧵 below👇.
- Come to Poster 2⃣1⃣0⃣ in Poster Session 4, 4.45pm at #CVPR2026 today. Find out how to add a smart multi-level highlighter to your VLM ! 🏛️ m3grounder.github.io @CVIT_IIIT_Hyd @BharatGen_Com @iiit_hyderabad @CVPRPaper 3/3 at #CVPR2026 📄 M3Grounder from @BharatGen_Com and @iiit_hyderabad Most #DocumentAI systems can answer questions… …but they rarely show where the answer came from. 👉 What if we gave document AI systems a smart highlighter? Thread 🧵👇
