Geospatial data toolkit for LLMs (and humans!)
Pietro Bellodi
Projects
Drop-in tool for quickly benchmarking local LLMs on AIME
All-in-one tool for SAE training, interpretation and validation
Posts and experiments
- Can LLMs be used in reverse?
What happens if we don't stop an LLM and let it write the user message?
Research
- Assessing the (In)Ability of LLMs to Reason in Interval Temporal Logic
Synthetic benchmark for reasoning in interval temporal logic, tested on five open-weight LLMs. Published at TIME 2025