The registration is now open for authors, speakers and PC (if you didn't get the link in an email from the organizers, please let us know). If you got an email from someone else, do not provide any payment data in any other way - it is very likely a scam!
Background
Large-scale semantic processing and strong computer assistance of mathematics and science is our inevitable future. New combinations of AI and reasoning methods and tools deployed over large mathematical and scientific corpora will be instrumental to this task. The AITP conference is the forum for discussing how to get there as soon as possible, and the force driving the progress towards that.
Topics
- AI and big-data methods in theorem proving and mathematics
- Collaboration between automated and interactive theorem proving, in particular their AI/ML aspects
- Common-sense reasoning and reasoning in science
- Alignment and joint processing of formal, semi-formal, and informal libraries, Formal Abstracts
- Methods for large-scale computer understanding of mathematics and science
- Combinations of linguistic/learning-based and semantic/reasoning methods
- Formal verification of AI and machine learning algorithms, explainable AI
Sessions
There will be several focused sessions on AI for ATP, ITP, mathematics, relations to general AI (AGI), Formal Abstracts, linguistic processing of mathematics/science, modern AI and big-data methods, and several sessions with contributed talks. The focused sessions will be based on invited talks and discussion oriented. AITP'24 is planned as an in-person conference.
Confirmed Participants/Speakers (TBC)
| João Araújo | Universidade Nova de Lisboa | |
| Michael R. Douglas | Stony Brook University | |
| Mario Carneiro | Chalmers University | |
| Thibault Gauthier | Czech Technical University in Prague | |
| Georges Gonthier | INRIA | |
| Thomas C. Hales | University of Pittsburgh | |
| Sean Holden | University of Cambridge | |
| Jan Jakubuv | Czech Technical University in Prague | |
| Mikoláš Janota | University of Lisbon | |
| Moa Johansson | Chalmers University | |
| Peter Koepke | University of Bonn | |
| Konstantin Korovin | The University of Manchester | |
| Michael Kinyon | University of Denver | |
| Miroslav Olsak | University of Cambridge | |
| Aarne Ranta | Chalmers University | |
| Michael Rawson | University of Southampton, UK | |
| Stephan Schulz | DHBW Stuttgart | |
| Martin Suda | Czech Technical University in Prague | |
| Josef Urban | Czech Technical University in Prague | |
| Adam Vandervorst | Qoba.ai | |
| Robert Veroff | University of New Mexico | |
| Petr Vojtěchovský | University of Denver | |
| Sean Welleck | Carnegie Mellon University | |
| Zsolt Zombori | Alfréd Rényi Institute of Mathematics |
Invited talks (TBC)
| Jonas Bayer and Mantas Bakšys | Kimina Prover -- Tackling Competition Level Mathematics through Reinforcement Learning |
| Mario Carneiro | Equational Theories Project |
| Moa Johansson | Learning Efficient Recursive Numeral Systems via Reinforcement Learning |
Dates
- Submission deadline:
May 5Extended: May 12, 2025 - Author notification: June 20, 2025
- Conference registration: TBA
- Camera-ready versions: TBA
- Conference: August 31 - September 5, 2025
Contributed talks
| J.D. Phillips. Applying AI to automated theorem proving in loop theory |
| Nil Geisweiller. Estimating the Probability of a Conjecture to be a Theorem with PLN for Inference Control |
| Thibault Gauthier. A Strategy for Lowering the Upper Bound of R(5,5) |
| Aarne Ranta. Symbolic Informalization: Fluent, Productive, Multilingual |
| Konstantinos Kogkalidis, Orestis Melkonian and Jean-Philippe Bernardy. Learning Structure-Aware Representations of Dependent Types |
| Jonathan Julian Huerta Y Munive. Reaping the fruits of the Isabelle/RL project |
| Paweł Balawender. Simpler proof assistants via bounded arithmetic |
| Auguste Poiroux, Antoine Bosselut and Viktor Kuncak. RLMEval: Evaluating Autoformalization on Research-Level Mathematics |
| Thibault Gauthier and Josef Urban. Learning Conjecturing from Scratch |
| Zarathustra Goertzel. MeTTaMath: Integrating Formal Verification into an AGI Cognitive Architecture via the MeTTa language |
| Chad Brown, Cezary Kaliszyk, Martin Suda and Josef Urban. Hammering Higher Order Set Theory |
| Isaac Li. Towards Lightweight and LLM-Free Semantic Search for mathlib4 |
| Adrian De Lon, Josef Urban, Peter Koepke, Atle Hahn and Mario Carneiro. Le Miz s’approche: Informalization and Autoformalization with Mizar and Naproche |
| Christoph Wernhard and Zsolt Zombori. Exploring Metamath Proof Structures: Progress Report |
| Jan Jakubuv, Mikoláš Janota, Jelle Piepenbrock and Josef Urban. Machine Learning for Quantifier Selection in cvc5 |
| David Cerna. Hypothesis Space Processing for Efficient Rule Learning Through Inductive Logic Programming |
| Adam Dingle. Natural-Language Proofs with Higher-Order Logic |
| Guy Axelrod, Moa Johansson, Devdatt Dubhashi, Nicholas Smallbone, Andrea Silvi and Sandro Stucki. Learning to Generate Abstractions for an Equational Solver |
| Martin Suda. Deepire II = RL(GNN+2RvNN) |
| Chad Brown, Karel Chvalovský, Mikoláš Janota, Miroslav Olšák and Stefan Ratschan. SMT and Functional Equation Solving over the Reals: Challenges from the IMO |
| Michael Rawson. First-Order Equational Reasoning via E-graphs and λ-terms |
| David Fuenmayor and Christoph Benzmüller. HOL as a Lingua Franca for Argumentative Reasoning Agents |
| Aishik Ghosh. Towards AI-assisted neutrino theory design |
| Chad Brown and Mikoláš Janota. Can Pigeonhole Principle Definitions Be Learned? |
| Jan Hůla. GNNs as Parametrized Primal-dual Algorithms |
| František Koutenský, Petr Hyner and Jan Hůla. Generalization of LLMs in SAT Reasoning via Structured Scratchpad Interaction |
| Yousef Alhessi, Sólrún Halla Einarsdóttir, George Granberry, Emily First, Moa Johansson, Sorin Lerner and Nicholas Smallbone. Lemmanaid: Neuro-Symbolic Lemma Conjecturing |
| Qiqi Gu, Jan Hůla and Mikoláš Janota. Keeping LLMs in Check by Automated Reasoning |
| Marek Dančo, Petra Hozzová and Mikoláš Janota. Project Proposal: Machine Learning for Model-Based Quantifier Instantiation |
| Risako Ando, Koji Mineshima and Mitsuhiro Okada. Can Large Language Models Support Proving Theorems Involving Multiply Nested Mathematical Induction? --- A Preliminary Report --- |
| Simon Frieder, Sam Bealing and Arsenii Nikolaiev. PROOFLESS: Final-Answer Datasets Do Not Assess\\Mathematical Reasoning Reliably |
| Josef Urban. LLMs as Proof Reconstructors for ATP Hammers? (Project Proposal) |
| Bernardo Atalaia, Mikoláš Janota and João Araújo. Project Proposal: Autoformalization for Algebras |
| Ziyu Zhou and Ziwei Li. LLMs Can Learn Theorem Libraries Through Dialogue to Become Effective Autoformalizers |
| Stephan Schulz. Theorem Provers and the Future AI Math Ecosystem |
| Bartosz Piotrowski, Witold Drzewakowski, Konrad Staniszewski and Piotr Miłoś. Do LLMs know when they are wrong? |
Program Committee (TBC)
| Michael R. Douglas (co-chair) | Stony Brook University |
| Ulrich Furbach | University of Koblenz |
| Thibault Gauthier | Czech Technical University in Prague |
| Thomas C. Hales (co-chair) | University of Pittsburgh |
| Sean Holden | University of Cambridge |
| Mikoláš Janota | University of Lisbon |
| Moa Johansson | Chalmers University |
| Cezary Kaliszyk (co-chair) | University of Melbourne |
| Michael Kinyon | University of Denver |
| Konstantin Korovin | The University of Manchester |
| Mirek Olsak | University of Cambridge |
| Bartosz Piotrowski | IDEAS NCBR |
| Michael Rawson | University of Southampton, UK |
| Stephan Schulz (co-chair) | DHBW Stuttgart |
| Sho Sonoda | RIKEN AIP |
| Martin Suda | Czech Technical University in Prague |
| Josef Urban (co-chair) | Czech Technical University in Prague |
| Sean Welleck | Carnegie Mellon University |
| Zsolt Zombori | Alfréd Rényi Institute of Mathematics |
Organizers
| Georges Gonthier | INRIA |
| Cezary Kaliszyk | University of Melbourne |
| Josef Urban | Czech Technical University in Prague |
AITP Program
August 31
| 19:30 | dinner |
September 1
| 9:00-10:15 Chair: Josef Urban |
Welcome Moa Johansson Learning Efficient Recursive Numeral Systems via Reinforcement Learning (50m) Yousef Alhessi, Sólrún Halla Einarsdóttir, George Granberry, Emily First, Moa Johansson, Sorin Lerner and Nicholas Smallbone Lemmanaid: Neuro-Symbolic Lemma Conjecturing (25m) |
| 10:15-10:45 | coffee break |
| 10:45-12:05 Chair: Moa Johansson |
Thibault Gauthier and Josef Urban Learning Conjecturing from Scratch (30m) Risako Ando, Koji Mineshima and Mitsuhiro Okada Can Large Language Models Support Proving Theorems Involving Multiply Nested Mathematical Induction? (20m) David Fuenmayor and Christoph Benzmüller HOL as a Lingua Franca for Argumentative Reasoning Agents (30m) |
| 12:05-13:30 | lunch |
| 16:30-17:00 | coffee break |
| 17:00-19:00 Chair: Mario Carneiro |
Nil Geisweiller Estimating the Probability of a Conjecture to be a Theorem with PLN for Inference Control (40m) Zarathustra Goertzel MeTTaMath: Integrating Formal Verification into an AGI Cognitive Architecture via the MeTTa language (20m) Christoph Wernhard and Zsolt Zombori Exploring Metamath Proof Structures: Progress Report (30m) Chad Brown, Cezary Kaliszyk, Martin Suda and Josef Urban Hammering Higher Order Set Theory (30m) |
| 19:00 | dinner |
September 2
| 9:00-10:15 Chair: Joao Araujo |
J.D. Phillips Bol-Moufang rings and things, yet again (45m) Guy Axelrod, Moa Johansson, Devdatt Dubhashi, Nicholas Smallbone, Andrea Silvi and Sandro Stucki Learning to Generate Abstractions for an Equational Solver (30m) |
| 10:15-10:45 | coffee break |
| 10:45-12:05 Chair: Sean Holden |
Michael Rawson First-Order Equational Reasoning via E-graphs and λ-terms (20m) David Cerna Hypothesis Space Processing for Efficient Rule Learning Through Inductive Logic Programming (30m) Jan Hůla GNNs as Parametrized Primal-dual Algorithms (30m) |
| 12:05-13:30 | lunch |
| 16:30-17:00 | coffee break |
| 17:00-19:00 Chair: Georges Gonthier |
Aarne Ranta Symbolic Informalization: Fluent, Productive, Multilingual (30m) Auguste Poiroux, Antoine Bosselut and Viktor Kuncak RLMEval: Evaluating Autoformalization on Research-Level Mathematics (30m) Adrian De Lon, Josef Urban, Peter Koepke, Atle Hahn and Mario Carneiro Le Miz s’approche: Informalization and Autoformalization with Mizar and Naproche (20m) Isaac Li Towards Lightweight and LLM-Free Semantic Search for mathlib4 (30m) Paul-André Melliès The Malinca Project (10m) |
| 19:00 | dinner |
September 3
| 9:00-10:15 Chair: Mirek Olsak |
Mantas Bakšys and Jonas Bayer Kimina Prover — Tackling Competition Level Mathematics through Reinforcement Learning (50m) Simon Frieder, Sam Bealing and Arsenii Nikolaiev PROOFLESS: Final-Answer Datasets Do Not Assess Mathematical Reasoning Reliably (25m) |
| 10:15-10:45 | coffee break |
| 10:45-12:15 Chair: Stephan Schulz |
Chad Brown, Karel Chvalovský, Mikoláš Janota, Miroslav Olšák and Stefan Ratschan SMT and Functional Equation Solving over the Reals: Challenges from the IMO (25m) Jan Jakubuv, Mikoláš Janota, Jelle Piepenbrock and Josef Urban Machine Learning for Quantifier Selection in cvc5 (25m) Marek Dančo, Petra Hozzová and Mikoláš Janota Project Proposal: Machine Learning for Model-Based Quantifier Instantiation (20m) František Koutenský, Petr Hyner and Jan Hůla Generalization of LLMs in SAT Reasoning via Structured Scratchpad Interaction (20m) |
| 12:30-14:00 | lunch |
| 17:00-17:30 | coffee break |
| 17:30-19:00 Chair: TBA |
Stephan Schulz Theorem Provers and the Future AI Math Ecosystem (25m) Panel Discussion (65m) |
| 19:00 | dinner |
September 4
| 9:00-10:20 Chair: Mikolas Janota |
Mario Carneiro Equational Theories Project (50m) Martin Suda Deepire II = RL(GNN+2RvNN) (25m) |
| 10:15-10:45 | coffee break |
| 10:45-12:05 Chair: Peter Koepke |
Jonathan Julian Huerta Y Munive Reaping the fruits of the Isabelle/RL project (30m) Paweł Balawender Simpler proof assistants via bounded arithmetic (20m) Aishik Ghosh Towards AI-assisted neutrino theory design (30m) |
| 12:10-13:30 | lunch |
| 16:30-17:00 | coffee break |
| 17:00-19:00 Chair: Aarne Ranta |
Adam Dingle Natural-Language Proofs with Higher-Order Logic (25m) Atle Hahn Autoformalization with Naproche-ZF (15m) Bernardo Atalaia, Mikoláš Janota and João Araújo Project Proposal: Autoformalization for Algebras (20m) Ziyu Zhou and Ziwei Li LLMs Can Learn Theorem Libraries Through Dialogue to Become Effective Autoformalizers (20m) Qiqi Gu, Jan Hůla and Mikoláš Janota Keeping LLMs in Check by Automated Reasoning (20m) Chad Brown and Mikoláš Janota Can Pigeonhole Principle Definitions Be Learned? (20m) |
| 19:00 | dinner |
September 5
| 9:00-10:15 Chair: TBA |
Thibault Gauthier A Strategy for Lowering the Upper Bound of R(5,5) (40m) Joao Araujo TBA (35m) |
| 10:15-10:45 | coffee break |
| 10:45-12:05 Chair: TBA | Josef Urban TBA/Discussions |
| 12:30-14:00 | lunch and departure from the center |
September 6 (optional for participants staying in Aussois)
| 10:00-16:30 | Excursion/Discussions in groups |
Pictures from the previous conferences
Location, Prices and Further Local Information
The conference will take place from August 31 to September 5, 2025, in
the
CNRS Paul-Langevin Conference Center
located in
the mountain village of Aussois in Savoy. Dominated by the "Dent
Parrachée", one of the highest peaks of La Vanoise, Aussois is located
on a sunny plateau at 1500 m altitude, offering a magnificent panorama
of the surrounding mountains and a direct access to the downhill ski
slopes or cross country slopes in winter.
The total price for accommodation and food for the five days will be around
650 EUR.
Arrival/Departure:
The first meal is dinner on August 31st and the last meal is lunch on September 5th. Aussois is less than 2h from the airports of Lyon, Geneve, Chambery, Annecy, Grenoble and Turin. There are trains and buses to Modane from these airports. Aussois is 8km from the Modane TGV station with direct trains from/to Paris.
We will organize a bus to Aussois from Modane at around 19:10 pm on Sunday, August 31st. (Note that the Modane station is now again reachable by trains.) The bus will wait for the TGV train from Paris arriving to Modane at 18:50 (starting at 14:48 in Paris) and the train from Milan arriving at 19:05 (starting at 16:10 in Milan). If you plan to travel to Aussois on your own, there are taxis and alternative buses from Modane (see here - the Aussois Office de Tourisme stop is close to the conference center).
We have not yet set the time for the departure of the taxis after lunch on September 5th. This will be optimized based on your departure flights. The center will likely close after our last session on Friday. If you want to stay for the weekend, there are other hotels in and around Aussois. If you have more questions/notes, put them into the registration form and/or look into the FAQ.