Scholar
Luca Soldaini
Google Scholar ID: 3KPvwcgAAAAJ
Allen Institute for AI
Citations & Impact
Citations
4,323
H-index
29
i10-index
53
Publications
20
Co-authors
8
list available
Publications
Cited
0
Resume (English only)
Academic Achievements
- Olmo project received two Best Paper Awards at ACL 2024
- Led the release of fully open Olmo 2 models (7B, 13B, 32B)
- Launched Tülu 3 post-training pipeline and Molmo multimodal models
- Developed olmOCR, a high-performance toolkit for PDF text extraction
- Created predictive techniques and benchmarks to characterize LLM behavior during pretraining
Research Experience
- Co-leads the data team for Ai2’s Olmo project with Kyle Lo
- Develops adaptation recipes for LLMs, including the Tülu 3 post-training pipeline (supporting models up to 405B parameters)
- Contributed to the Molmo family of open multimodal AI models
- Co-developed tools for analyzing and improving LLM pipelines: AboutMe, WIMBD, WebOrganizer, and olmOCR
- Investigated LLM-retrieval system interfaces; co-proposed FollowIR with Orion Weller, later extended to multilingual settings
- Collaborated on OpenSciLLM, an end-to-end demo for literature-grounded scientific synthesis using LLMs
Miscellany
- Enjoys brewing espresso and going on runs
- Dreams about utopian mass transit systems
- Curates a growing collection of laptop stickers
- Spends time with his handsome cat
- Believes raccoons are the best
