Priyadarsi Mishra

About Me

Hello! I am Priyadarsi Mishra, a senior undergraduate student at Texas A&M University studying Computer Science with minors in Statistics and Mathematics. I am fortunate to be advised by Tomer Galanti.

My research interests lie broadly in multimodal reasoning, world models, and efficient AI systems. I am particularly interested in understanding how models learn representations of temporal and physical structure, how those representations support reasoning and prediction, and how model and system design can make these capabilities more efficient and scalable. I also have a complementary interest in reliable agentic reasoning, including how language models use feedback, verification, and interaction while maintaining robust reasoning.

My recent work spans representation learning, LLM-guided program search, multimodal systems, world models, and agentic reasoning. I am always happy to discuss research ideas or potential collaborations via email.

Research Interests

• Multimodal Reasoning & World Models: video understanding, predictive representations, physical and temporal reasoning & representation geometry
• Efficient AI Systems: efficient model architectures, speculative and parallel generation & adaptive computation
• Reliable Agentic Reasoning: verification, feedback, self-improving agents, multi-agent interaction & robustness to unreliable or sycophantic behavior

News

• [May 2026 - Aug. 2026] Software Engineer Intern @ Ads.
• [Apr. 2026] LLM Priors for ERM over Programs was accepted to ICML 2026.
• [Jan. 2026] On the Alignment Between Supervised and Self-Supervised Contrastive Learning was accepted to ICLR 2026.
• [Sept. 2025 - Jan. 2026] Generative AI Research Intern @ BMW Group.
• [May 2025 - Aug. 2025] Software Engineer Intern @ Ads.

Publications

Experience

• Software Engineer Intern, Ads, May 2026 - Aug. 2026.
• Generative AI Research Intern, BMW Group, Sept. 2025 - Jan. 2026.
• Software Engineer Intern, Ads, May 2025 - Aug. 2025.