Rohin Manvi

Hello! I’m a final-year Computer Science student at Stanford University with a broad interest in the development and responsible application of generative foundation models, particularly large language models. Lately, I’ve been exploring ways to enable these models to “think” longer or harder when necessary, leveraging inference-time techniques like search, self-refinement, and new paradigms such as diffusion-based language models.

I’ve had the privilege of working with Prof. Stefano Ermon at Stanford, focusing on efficient inference techniques for LLMs, geospatial tasks, and bias analysis. Previously, I applied my machine learning skills in industry at Meta and Lacework, and I’ve also investigated using LLMs for decision-making in autonomous systems.

If you’re interested in these topics—or just want to say hello—I’d love to hear from you!

selected publications

Adaptive Inference-Time Compute: LLMs Can Predict if They Can Do Better, Even Mid-Generation

Rohin Manvi, Anikait Singh, and Stefano Ermon

2024

arXiv Bib

@misc{manvi2024adaptiveinferencetimecomputellms,
  title = {Adaptive Inference-Time Compute: LLMs Can Predict if They Can Do Better, Even Mid-Generation},
  author = {Manvi, Rohin and Singh, Anikait and Ermon, Stefano},
  year = {2024},
  eprint = {2410.02725},
  archiveprefix = {arXiv},
  primaryclass = {cs.CL},
  url = {https://arxiv.org/abs/2410.02725},
}

Large Language Models are Geographically Biased

Rohin Manvi, Samar Khanna, Marshall Burke, David B. Lobell, and Stefano Ermon

In Forty-first International Conference on Machine Learning, 2024

arXiv Bib Code

@inproceedings{manvi2024large,
  title = {Large Language Models are Geographically Biased},
  author = {Manvi, Rohin and Khanna, Samar and Burke, Marshall and Lobell, David B. and Ermon, Stefano},
  booktitle = {Forty-first International Conference on Machine Learning},
  year = {2024},
  url = {https://arxiv.org/abs/2402.02680},
  journal = {Forty-first International Conference on Machine Learning, 2024},
}

GeoLLM: Extracting Geospatial Knowledge from Large Language Models

Rohin Manvi, Samar Khanna, Gengchen Mai, Marshall Burke, David B. Lobell, and Stefano Ermon

In The Twelfth International Conference on Learning Representations, 2024

arXiv Bib Code Website

@inproceedings{manvi2024geollm,
  title = {Geo{LLM}: Extracting Geospatial Knowledge from Large Language Models},
  author = {Manvi, Rohin and Khanna, Samar and Mai, Gengchen and Burke, Marshall and Lobell, David B. and Ermon, Stefano},
  booktitle = {The Twelfth International Conference on Learning Representations},
  year = {2024},
  url = {https://arxiv.org/abs/2310.06213},
  journal = {The Twelfth International Conference on Learning Representations, 2024},
}