About me

My name is Amirhosein Javadi, and I am a Ph.D. student in ECE at UCSD, advised by Professor Tara Javidi. Prior to joining UCSD, I received a Bachelor of Science in Electrical Engineering from Sharif University of Technology in 2023. My research focuses on generative AI and multimodal learning, with an emphasis on world models, novel view synthesis, and video generation, as well as making foundation models more reliable and interpretable.

Publications

• Audio LLMs Know When They Can’t Hear You
  Authors: Amirhosein Javadi, Richa Dixit, Mehrdad Farajtabar, Minsik Cho, Devang Naik, Mohammad Samragh
  Under review at ICLR 2027

• CI-CBM: Class-Incremental Concept Bottleneck Model for Interpretable Continual Learning
  Authors: Amirhosein Javadi, Tuomas Oikarinen, Tara Javidi, Tsui-Wei Weng
  Transactions on Machine Learning Research (TMLR), 2026
  [Project Page] [Code]

• Active Sampling for Ultra-Low-Bit-Rate Video Compression via Conditional Controlled Diffusion
  Authors: Amirhosein Javadi, Shirin Saeedi Bidokhti, Tara Javidi
  Transactions on Machine Learning Research (TMLR), 2026
  [Project Page] [Code]

• A Single Image and Multimodality Is All You Need for Novel View Synthesis
  Authors: Amirhosein Javadi, Chi-Shiang Gau, Konstantinos D. Polyzos, Tara Javidi
  ICLR 2026 Workshop on Multimodal Intelligence
  [Project Page] [Code]

• Active World-Model with 4D-informed Retrieval for Exploration and Awareness
  Authors: Elaheh Vaezpour, Amirhosein Javadi, Tara Javidi
  ICLR 2026 Workshop on World Models

• BLO-SAM: Bi-Level Optimization Based Finetuning of the Segment Anything Model for Overfitting-Preventing Semantic Segmentation
  Authors: Li Zhang, Youwei Liang, Ruiyi Zhang, Amirhosein Javadi, Pengtao Xie
  ICML 2024

Experience

• Machine Learning Research Intern at Apple (Jun. 2026 – Sep. 2026)
  • Reliable Audio-Language Models. Developed a framework that predicts, before generation, whether an Audio LLM can reliably understand a spoken input. A lightweight predictor over frozen audio-encoder representations outperforms speech-quality, uncertainty, and transcript-based baselines, and generalizes across unseen acoustic conditions and Audio LLM families.
• Research Intern at Soter Technology, Inc. (Jun. 2025 – Sep. 2025)
  • Visual World Models. Developed camera-controlled novel-view video generation from single- or multi-view images and videos, where users specify the camera trajectory, improving performance on downstream monitoring and safety tasks in complex industrial environments.

News

• Sep. 2026: Our paper Audio LLMs Know When They Can’t Hear You is now on arXiv.
• 2026: CI-CBM and Active Sampling for Ultra-Low-Bit-Rate Video Compression accepted at TMLR.
• Jun. 2026: Started as a Machine Learning Research Intern at Apple.
• 2023: Received the Electrical and Computer Engineering Department Fellowship, University of California San Diego.