Added calanzone2024locolm

_news/locolm-accepted-r2fm.md

@@ -0,0 +1,7 @@
+---
+title: "Loco LM @ R2FM"
+collection: news
+permalink: /news/loco-lm-r2fm
+date: 2024-03-06
+---
+How to make LLMs more logically consistent? Check <a href="https://openreview.net/forum?id=q3SGbfj19d"><b>our work</b></a> at the <a href="https://iclr-r2fm.github.io/"><b>ICLR 2024 Workshop on on Reliable and Responsible Foundation Models</b></a>.

_publications/calanzone2024locolm.md

@@ -0,0 +1,21 @@
+---
+collection: publications
+ref: "calanzone2024locolm"
+permalink: "publications/calanzone2024locolm"
+title:  "Galerkin meets Laplace: Fast uncertainty estimation in neural PDEs"
+date:   2024-03-05 10:00
+tags: nesy probml llm
+image: "/images/papers/calanzone2024locolm/locolm.png"
+authors: "Diego Calanzone, Antonio Vergari, Stefano Teso"
+paperurl: "https://openreview.net/forum?id=q3SGbfj19d"
+pdf: "https://openreview.net/pdf?id=q3SGbfj19d"
+venue: "R2FM Workshop @ ICLR 2024"
+excerpt: "We introduce a training objective based on principled probabilistic reasoning that teaches a LLM to be logically consistent with a set of external facts and rules, allowing to extrapolate to unseen but semantically similar factual knowledge."
+abstract: "Large language models (LLMs) are a promising venue for natural language understanding and generation tasks. However, current LLMs are far from reliable: they are prone to generate non-factual information and, more crucially, to contradict themselves when prompted to reason about beliefs of the world. These problems are currently addressed with large scale fine-tuning or by delegating consistent reasoning to external tools. In this work, we strive for a middle ground and introduce a training objective based on principled probabilistic reasoning that teaches a LLM to be consistent with external knowledge in the form of a set of facts and rules. Fine-tuning with our loss on a limited set of facts enables our LLMs to be more logically consistent than previous baselines and allows them to extrapolate to unseen but semantically similar factual knowledge more systematically."
+bibtex: "@inproceedings{calanzone2024locolm,
+  title={Galerkin meets Laplace: Fast uncertainty estimation in neural PDEs},
+  author={Diego Calanzone, Antonio Vergari, Stefano Teso},
+  booktitle={ICLR 2024 Workshop on Reliable and Responsible Foundation Models},
+  year={2024}
+}"
+---