What is common in: 🔸ImageNet (2012), 🔸GANs (2014), 🔸Attention is all you need (2017) 🔸Language models are few-shot learners (GPT 3) (2020) 🔸Chain of Thought Reasoning (2022) All these were some of the most impactful papers in AI, which debuted in NeurIPS, the launchpad for revolutionary AI breakthroughs. Now, we're thrilled to join this prestigious lineage with not one, but TWO papers in a single year! 🤯 Our accepted works: 1️. Quality-aware DPO for diffusion models (Kalaido.ai) - Adaptive Foundational Model track 2️. SBSC for Mathematics Olympiad (#ProjectRamanujan) - MATH-AI track This double acceptance cements our position at the forefront of AI research. We're pushing boundaries in both visual AI and mathematical problem-solving! Join us in celebrating this massive achievement as we continue to shape the future of AI! Stay tuned for more details on our contributions to the AI community! Kunal Singh, Pradeep Moturi, Mukund Khanna, Ankan Biswas, SIVA KISHORE GOLLAPALLI, Suraj Amonkar, Sayandeep Bhowmick #NeurIPS2024 #AIResearch #BreakthroughScience #FutureTech #AI #MachineLearning
🔥𝗘𝘅𝗰𝗶𝘁𝗶𝗻𝗴 𝗡𝗲𝘄𝘀!🔥 Fractal AI research team is presenting not one but two breakthrough #AI papers at #NeurIPS2024, pushing the limits in image generation and mathematical reasoning! (https://lnkd.in/dChDsKM7) 𝗪𝗵𝗮𝘁 𝗶𝘀 𝗡𝗲𝘂𝗿𝗜𝗣𝗦? NeurIPS (𝗡𝗲𝘂𝗿𝗮𝗹 𝗜𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 𝗦𝘆𝘀𝘁𝗲𝗺𝘀) is the pre-eminent #AI conference of our times, where revolutionary papers like #ImageNet (2012), #GANs (2014), #Attention is All You Need (2017), and Chain of Thought Reasoning (2022) were first presented. AI papers presented at NeurIPS have changed the world! 𝗣𝗮𝗽𝗲𝗿 𝟭: 𝗘𝗳𝗳𝗲𝗰𝘁𝗶𝘃𝗲 𝗧𝗲𝘅𝘁-𝘁𝗼-𝗜𝗺𝗮𝗴𝗲 𝗔𝗹𝗶𝗴𝗻𝗺𝗲𝗻𝘁 𝘄𝗶𝘁𝗵 𝗤𝘂𝗮𝗹𝗶𝘁𝘆 𝗔𝘄𝗮𝗿𝗲 𝗣𝗮𝗶𝗿 𝗥𝗮𝗻𝗸𝗶𝗻𝗴 Earlier this year, we launched Kalaido.ai (see this post https://lnkd.in/dAQ4EqqU). Underneath this tech we created a new method "Text-to-Image alignment with quality aware pair ranking" - This method enhances the accuracy and efficiency of text-to-image alignment by ranking pairs based on a novel quality metric, with wide-reaching applications in image generation and visual media. 𝗣𝗮𝗽𝗲𝗿 𝟮: 𝗦𝗕𝗦𝗖: 𝗦𝘁𝗲𝗽-𝗯𝘆-𝗦𝘁𝗲𝗽 𝗖𝗼𝗱𝗶𝗻𝗴 𝗳𝗼𝗿 𝗜𝗺𝗽𝗿𝗼𝘃𝗶𝗻𝗴 𝗠𝗮𝘁𝗵𝗲𝗺𝗮𝘁𝗶𝗰𝗮𝗹 𝗢𝗹𝘆𝗺𝗽𝗶𝗮𝗱 𝗣𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 A few weeks before the launch of OpenAI's o1 model, I had posted about our reasoning system called Project Ramanjum (See the post https://lnkd.in/dHwP_EHx) Underneath this tech, we built a multi-turn-based, step-by-step coding approach to improve problem-solving skills in mathematical Olympiads, with potential for AI-driven educational tools. AI research often goes unseen, hidden behind the tools and applications that change the way we live, work, and create. But it’s this deep research—the complex designs and innovative frameworks—that powers the impact AI has on the world. And that’s what we’re excited to share at NeurIPS 2024: the potential to influence the future of image generation and AI-driven tools for education. We’re proud to be contributing to the rapidly evolving AI research ecosystem. https://lnkd.in/dChDsKM7. #NeurIPS2024 #AIResearch #AIInnovation #IndiaInAI #MachineLearning Suraj Amonkar Kunal Singh Mukund Khanna Pradeep Moturi Ankan Biswas Sayandeep Bhowmick SIVA KISHORE GOLLAPALLI