trivial updates

Author: jla-gardner

paper/paper.bib

@@ -39,15 +39,18 @@ @article{Batzner-22-05
 }
 
 @inproceedings{Batatia-22-10,
-  title     = {{{MACE}}: {{Higher Order Equivariant Message Passing Neural Networks}} for {{Fast}} and {{Accurate Force Fields}}},
-  booktitle = {Advances in {{Neural Information Processing Systems}}},
-  author    = {Batatia, Ilyes and Kovacs, David Peter and Simm, Gregor N. C. and Ortner, Christoph and Csanyi, Gabor},
-  year      = {2022},
-  month     = oct,
-  urldate   = {2025-06-22}
+  author    = {Batatia, Ilyes and Kovacs, David P and Simm, Gregor and Ortner, Christoph and Csanyi, Gabor},
+  booktitle = {Advances in Neural Information Processing Systems},
+  editor    = {S. Koyejo and S. Mohamed and A. Agarwal and D. Belgrave and K. Cho and A. Oh},
+  pages     = {11423--11436},
+  title     = {MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields},
+  url       = {https://proceedings.neurips.cc/paper_files/paper/2022/file/4a36c3c51af11ed9f34615b81edb5bbc-Paper-Conference.pdf},
+  volume    = {35},
+  year      = {2022}
 }
 
 
+
 @article{HjorthLarsen-17-06,
   title   = {The Atomic Simulation Environment---a {{Python}} Library for Working with Atoms},
   author  = {Hjorth Larsen, Ask and J{\o}rgen Mortensen, Jens and Blomqvist, Jakob and Castelli, Ivano E and Christensen, Rune and Du{\l}ak, Marcin and Friis, Jesper and Groves, Michael N and Hammer, Bj{\o}rk and Hargus, Cory and Hermes, Eric D and Jennings, Paul C and Bjerre Jensen, Peter and Kermode, James and Kitchin, John R and Leonhard Kolsbjerg, Esben and Kubal, Joseph and Kaasbjerg, Kristen and Lysgaard, Steen and Bergmann Maronsson, J{\'o}n and Maxson, Tristan and Olsen, Thomas and Pastewka, Lars and Peterson, Andrew and Rostgaard, Carsten and Schi{\o}tz, Jakob and Sch{\"u}tt, Ole and Strange, Mikkel and Thygesen, Kristian S and Vegge, Tejs and Vilhelmsen, Lasse and Walter, Michael and Zeng, Zhenhua and Jacobsen, Karsten W},
@@ -70,26 +73,21 @@ @article{Thompson-22-02
   urldate = {2022-11-08}
 }
 
-
-@misc{Simeon-23-06,
-  title         = {{{TensorNet}}: {{Cartesian Tensor Representations}} for {{Efficient Learning}} of {{Molecular Potentials}}},
-  author        = {Simeon, Guillem and {de Fabritiis}, Gianni},
-  year          = {2023},
-  number        = {arXiv:2306.06482},
-  eprint        = {2306.06482},
-  primaryclass  = {physics},
-  archiveprefix = {arXiv}
+@inproceedings{Simeon-23-06,
+  title     = {TensorNet: Cartesian Tensor Representations for Efficient Learning of Molecular Potentials},
+  author    = {Simeon, Guillem and {de Fabritiis}, Gianni},
+  booktitle = {Thirty-seventh Conference on Neural Information Processing Systems},
+  year      = {2023},
+  url       = {https://openreview.net/forum?id=BEHlPdBZ2e}
 }
 
-@misc{Schutt-21-06,
-  title         = {Equivariant Message Passing for the Prediction of Tensorial Properties and Molecular Spectra},
-  author        = {Sch{\"u}tt, Kristof T. and Unke, Oliver T. and Gastegger, Michael},
-  year          = {2021},
-  number        = {arXiv:2102.03150},
-  eprint        = {2102.03150},
-  primaryclass  = {physics},
-  doi           = {10.48550/arXiv.2102.03150},
-  archiveprefix = {arXiv}
+@inproceedings{Schutt-21-06,
+  title     = {Equivariant Message Passing for the Prediction of Tensorial Properties and Molecular Spectra},
+  author    = {Sch{\"u}tt, Kristof T. and Unke, Oliver T. and Gastegger, Michael},
+  booktitle = {Proceedings of Machine Learning Research},
+  year      = {2021},
+  note      = {Pre-print at: \url{https://arxiv.org/abs/2102.03150}},
+  doi       = {10.48550/arXiv.2102.03150}
 }
 
 @article{Wang-18-07,
@@ -352,17 +350,15 @@ @software{fairchem
   year    = {2025}
 }
 
-@misc{Kong-25-04,
-  title         = {{{MatterTune}}: {{An Integrated}}, {{User-Friendly Platform}} for {{Fine-Tuning Atomistic Foundation Models}} to {{Accelerate Materials Simulation}} and {{Discovery}}},
-  author        = {Kong, Lingyu and Shoghi, Nima and Hu, Guoxiang and Li, Pan and Fung, Victor},
-  year          = {2025},
-  month         = apr,
-  number        = {arXiv:2504.10655},
-  eprint        = {2504.10655},
-  primaryclass  = {cond-mat},
-  doi           = {10.48550/arXiv.2504.10655},
-  urldate       = {2025-04-16},
-  archiveprefix = {arXiv}
+@article{Kong-25-08,
+  title   = {{{MatterTune}}: An Integrated, User-Friendly Platform for Fine-Tuning Atomistic Foundation Models to Accelerate Materials Simulation and Discovery},
+  author  = {Kong, Lingyu and Shoghi, Nima and Hu, Guoxiang and Li, Pan and Fung, Victor},
+  year    = 2025,
+  journal = {Digital Discovery},
+  volume  = {4},
+  number  = {8},
+  pages   = {2253--2262},
+  doi     = {10.1039/D5DD00154D}
 }
 
 @misc{Radova-25-02,

paper/paper.md

@@ -51,7 +51,7 @@ The `graph-pes` package provides a **unified interface and framework** for defin
 
 A number of existing packages offer training and validation pipelines for particular ML-PES architectures, including `schnetpack` [@schutt2019schnetpack; @schutt2023schnetpack], `deepmd-kit` [@Wang-18-07; @Zeng-23-08], `nequip` [@Batzner-22-05], `mace-torch` [@Batatia-22-10], `torchmd-net` [@TorchMDNet], and `fairchem` [@fairchem]. 
 These frameworks focus on their associated model families and do not share a common interface for training.
-While `MatterTune` [@Kong-25-04] offers a unified interface for foundation model fine-tuning, it does not easily support training arbitrary models from scratch. 
+While `MatterTune` [@Kong-25-08] offers a unified interface for foundation model fine-tuning, it does not easily support training arbitrary models from scratch. 
 In contrast to these, `graph-pes` is a general, model-agnostic framework, designed to enable exact side-by-side comparisons, easy implementation of arbitrary new architectures, and standardized training and evaluation workflows.
 
 # Features and implementation
@@ -86,7 +86,7 @@ As well as training from scratch, we also support the fine-tuning of existing mo
 Under the hood, `graph-pes-train` builds upon the `PyTorch Lightning` [@Lightning] training loop, allowing the user to configure a variety of common training features and callbacks.
 We also support the use of arbitrary, user-defined components, including custom loss functions, model architectures, optimisers, and datasets.
 
-Because all models conform to the same interface, all training features can be used with any model architecture. Similarly, all downstream model uses can be written in an architecture-agnostic manner, allowing for MD, relaxations, and other scripts to be written once, and then used with any MLIP architecture, _e.g._ for extended validation beyond simple error metrics [@Morrow-23-03].
+Because all models conform to the same interface, all training features can be used with any model architecture. Similarly, all downstream model uses can be written in an architecture-agnostic manner, allowing for MD, relaxations, and other scripts to be written once, and then used with any MLIP architecture, _e.g._, for extended validation beyond simple error metrics [@Morrow-23-03].
 
 ## Easy access to foundation models