Chalmers University of Technology

Gothenburg CHAIR Workshop 2026

A CHAIR-funded workshop on AI for mathematics, automated reasoning, formalisation, autoformalisation, and related research directions.

Overview

This workshop is part of the 2026 CHAIR Research Themes program at Chalmers. The underlying theme, Cognition, Reasoning and AI in Collaboration, supports research activities spanning formalisation of mathematics, neuro-symbolic methods, reinforcement learning with feedback from provers, and cognitive aspects of reasoning AI.

The workshop aims to strengthen ties between Chalmers and related international research groups working in AI, reasoning, and formal mathematics.

Host institutionChalmers University of Technology, Gothenburg, Sweden
FundingCHAIR Research Themes 2026
Dates26–29 May 2026
FormatTalks, demos, discussions, and networking

Program

The schedule is tentative and may be updated. Slide links can be added to individual talks as they become available by replacing the placeholder link in the relevant program item.

Tuesday, 26 May

15:00
Optional coffee
MV lunch/coffee room, second floor
15:30–16:30
Colloquium with local mathematicians
Thierry Coquand, Devdatt Dubhashi, and Josef Urban — short talks. Room: Euler, Skeppsgränd 3.

Wednesday, 27 May

09:00–09:15
Arrival and welcome
09:15–09:45
Strong Within Benchmarks, Weak Across Them: Evaluating Neural Guidance in Vampire
Martin Suda
AbstractA neural clause-selection guidance approach in Vampire was recently shown to substantially improve the success rate of the prover's default strategy on TPTP. This talk studies the impact across ITP-derived benchmark sets and theorem proving strategies.
09:45–10:00
Comparing Learned Models for Neural Guidance in Vampire short
Karel Chvalovský
10:00
Coffee
10:30–11:15
Twee and Twitch
Nick Smallbone
11:15–11:45
To QED and Beyond: Adventures in Vibe Coding
Zar Goertzel
12:00–13:00
Lunch
13:00–13:35
Maths, The Very Confident Intern: How I Use LLMs in Mathematical Research
Dennis Eriksson
AbstractLLMs are fast, fluent, often useful, sometimes wrong, and always in need of supervision. This talk describes practical uses of LLMs in mathematical research, experimental mathematics, proof structuring, and writing.
13:35–14:05
Automated Root Certification in Lean
Dattabhasvant Ghadiyaram
AbstractA Lean library using effective transversality for certified root finding in multivariate polynomial systems, including dyadic intervals, interval arithmetic, the Krawczyk method, custom syntax, and a tactic for certification.
14:05–14:30
AI-assisted formalization of typing for Lean’s kernel
Mario Carneiro
14:30–15:00
Coffee
15:00–15:45
An experiment in autoformalization
Thierry Coquand
AbstractAn experiment on autoformalising a proof of Gödel's second incompleteness theorem, using interaction with Claude and a system of primitive recursive arithmetic.
15:45–16:00
Latest agentic tricks tentative
George Granberry
16:00–16:25
Connecting classical ATPs to constructive ITPs
Michael Rawson
AbstractA talk loosely based on “When Vampire met Agda”, about connecting classical ATPs such as Vampire to constructive ITPs such as Agda.

Thursday, 28 May

09:00–09:30
Beyond Theorem Proving
Yousef Alhessi
09:30–10:00
Model Diversity in Theorem Proving and Recursive Tasks
Emily First
10:00
Coffee
10:30–11:00
Vibe-coding ITP from scratch
Mirek Olšák
11:00–11:30
Learning-Guided Higher-Order Automated Reasoning for Isabelle/HOL
Jan Jakubův
AbstractHigher-order extensions of ENIGMA and Deepire 2.0 are evaluated on a large Isabelle/HOL benchmark exported via Sledgehammer in several TPTP formats, showing complementary contributions from higher-order reasoning.
11:30–12:00
Premise Selection for Higher-Order ATP Problems
Keneni Tesema
12:00–13:00
Lunch
Afternoon
Excursion/hike with discussions outdoors

Friday, 29 May

10:00–11:30
Open session
Room open for meetings, demos, discussions, and improvised talks.
11:30–12:00
Porting Mizar to Informath
Soline du Crest
12:00–12:15
Informath demo tentative
Aarne Ranta
12:15–12:40
Autoformalization with Naproche-ZF
Atle Hahn
AbstractAn update on the Naproche-ZF Autoformalization project with Adrian and Josef, including post-processing steps such as proof distillation, proof elaboration, and HTML rendering.

Travel and administration

Chalmers strongly prefers to book travel tickets directly for invited participants. Attendees who plan to come are asked to send their preferred flight options as soon as possible to Andrea Brodén, with Moa Johansson copied.

Accommodation arrangements will be coordinated after the participant list is finalized.