Nima Anari

Papers

Total Citations

H-Index

About

Nima Anari is a rising researcher at the forefront of generative AI and robot learning, whose work tackles fundamental bottlenecks in both fields. His primary research areas span efficient sampling for diffusion models and preference-based reward learning for robotics. Anari’s most impactful contribution addresses the critical speed-quality tradeoff in diffusion models. In his 2023 paper, "Parallel Sampling of Diffusion Models," he proposes a novel method to generate high-quality samples in far fewer sequential steps, offering a paradigm shift from the standard 1000-step denoising process. This work, already garnering 9 citations, promises to make powerful generative models practical for real-time applications. In robotics, Anari is pioneering methods to learn from human feedback more efficiently. His 2024 work on "Batch Active Learning of Reward Functions from Human Preferences" (7 citations) introduces a strategy to minimize costly human labeling by intelligently selecting the most informative queries. Additionally, his 2021 paper on "Learning Multimodal Rewards from Rankings" tackles the realistic scenario where human preferences are not uniform, enabling robots to learn from diverse or conflicting expert feedback. Through these contributions, Anari is shaping a future where AI systems are both faster to run and more aligned with nuanced human values.