The Matter Lab

Over the last year, the Acceleration Consortium and its members have helped create a materially better future, with positive impacts both at U of T and beyond. Explore our 2025-25 achievements in our impact report: shorturl.at/hVBsI

The Matter Lab reposted

In 2023, the AC received a $200M TIPS grant (@sshrc-crsh.canada.ca) to turn 🇨🇦 research strengths into world-leading capabilities. 3+ years on, the AC is a global leader in accelerated discovery, using AI & automation to up research productivity. Explore our impact report: shorturl.at/hVBsI

🎉Congratulations to our the Matter Lab team <The|Matter>, comprised of Yuma Nakamura, Yoshihiro Ogura, and Jenyu Chang, on receiving the Excellent Research Award (優秀研究賞) in the NEDO Quantum Computing Challenge QoL category!

🚀The team earned recognition in the award tier and secured continued priority access to IBM and Quantinuum quantum hardware to support the next phase of their research.

With this recognition in Season 1 (2025–2026), and continued hardware access, the team is moving forward into Season 2 (2026–2027), aiming for the top prize. We look forward to seeing what they accomplish next!

Yoshihiro helped shape the final presentation’s narrative, while Jenyu contributed substantially to the implementation.

The Matter Lab reposted

“The Acceleration Consortium is proving that a faster, more efficient model of discovery is not just possible, but already underway:" AC Director @aspuru.bsky.social See how we’re making it happen in our 2025-26 Impact Report: shorturl.at/hVBsI 🚀

Congratulations to @aspuru.bsky.social on being named a 2026 ScholarGPS® Highly Ranked Scholar™! This recognition places him among the top 0.05% of scholars worldwide within his ranking categories, reflecting his publication record, research impact, and the quality of his scholarly contributions.

At the Matter Lab, we are proud to celebrate this achievement and grateful for Alán’s leadership, mentorship, and support. 🎉Congratulations, Alán, and thank you to the lab members and collaborators whose contributions help make this work possible! ➡️ scholargps.com/scholars/240...

doi.orgDiscovery of Nanomolar CDK2 Inhibitors Using Boltz-2 | ChemRxivThe computational design of small-molecule ligands for pharmaceutical targets has historically been hindered by a strict trade-off between screening throughput and predictive accuracy. While next-gene...

What does a cofolding model do with chemistry it has never seen? We answer this in our new preprint: A prospective CDK2 inhibitor discovery campaign guided by Boltz-2. Preprint: doi.org/10.26434/che... Code: github.com/papidocha/CD... Data: doi.org/10.5281/zeno...

• 12 of 18 compounds from the first synthesis round inhibited CDK2 at or below 10 µM • Nanomolar inhibitors after a single structure-guided optimization cycle

• A chemotype distinct from all 10,402 known CDK inhibitors in ChEMBL • Boltz-2 outperformed a docking baseline across all 57 synthesized compounds Congratulations to Sven Papidocha, Yoshihiro Ogura, Jiahe Ren, Mikhail Saykin, Stuart Green, @variniabernales.bsky.social, and @aspuru.bsky.social!

Instead of screening a supplier catalogue, the team built a virtual library of 74,000 compounds, none of them for sale, all reachable in a single synthetic step. Boltz-2 then navigated that space prospectively, and the resulting predictions were translated directly into a synthesis plan.

sciencenews.dkAI-system kan omsætte forskeres spørgsmål til kvantesimuleringerEn ny AI-agent kan selv vælge metode og program, skrive og rette koden og køre en kvantesimulering, når brugeren beskriver opgaven med ord i stedet for computerkode. I gentagne test var...

🎉We’re excited to see El Agente Cuántico featured in ScienceNews.dk, highlighting work from The Matter Lab at the University of Toronto, led by our Co-Director @variniabernales.bsky.social, with our interdisciplinary team. ➡️ sciencenews.dk/da/ai-system...

🦾Six specialist AI agents, coordinated by a central agent, support tasks ranging from quantum circuits to energy transport in photosynthetic systems. The goal is to make advanced simulations more accessible and give researchers more time to develop ideas and evaluate results.

🔑 As Varinia mentioned, successful automation still requires careful validation and human oversight. Looking ahead, the team aims to connect simulations with laboratory experiments, bringing computation and experiment into a shared research cycle. Thank you to ScienceNews for featuring our work!

🚀Quantum simulation often requires researchers to navigate complex software, learn different programming interfaces, and troubleshoot calculations.

El Agente Cuántico helps lower these barriers: researchers describe their scientific question, and the system selects computational tools, writes and runs code, and corrects errors along the way.

arxiv.orgSelf-Building LabsSelf-driving laboratories (SDLs) accelerate scientific discovery by automating experimental execution and optimization. AI agents can increasingly propose new experiments, but existing SDLs can only r...

🤔 What if AI could not only automate scientific experiments, but also design and build the laboratories needed to perform them? We are excited to introduce Self-Building Labs (SBL), a new paradigm for autonomous scientific laboratories! 🌐 self-building-labs.github.io 📜 arxiv.org/abs/2610.23826

Alán Aspuru-Guzik@aspuru.bsky.social

Tired: Self-Driving Lab Wired: Self-Building Lab arxiv.org/abs/2610.23826

Our vision is to move beyond laboratories that simply execute predefined experiments toward scientific infrastructure that can build new capabilities from scratch, continuously adapt, and evolve alongside scientific discovery. We also envision leveraging SBL to build safer laboratories.

👏 This work was co-led by @jiarubai.bsky.social and @allanzhao.bsky.social, along with Han Hao, Miroslav Bogdanovic, Yang Cao, @variniabernales.bsky.social, and @aspuru.bsky.social. Huge congratulations to the entire team for this collaborative effort!

Alán Aspuru-Guzik@aspuru.bsky.social

Tired: Self-Driving Lab Wired: Self-Building Lab arxiv.org/abs/2610.23826

In our latest work, we demonstrate how SBL can: 🔹 Automatically reconstruct digital twins of laboratory equipment and environments. 🔹 Develop and validate new robotic manipulation skills through simulation and physical feedback.

🔹 Execute real-world quantum-dot synthesis experiments, coordinating liquid handling, robotic manipulation, plate sealing, and optical measurements. 🔹 Scale from individual robotic workstations to interconnected laboratories with multiple robots and instruments.

Alán Aspuru-Guzik@aspuru.bsky.social

Tired: Self-Driving Lab Wired: Self-Building Lab arxiv.org/abs/2610.23826

🚀 Self-Building Labs (SBLs) takes a step further: enabling laboratories to autonomously design, configure, and operate based on high-level scientific goals.

Whether starting from an empty laboratory or an existing self-driving lab, SBL aims to autonomously develop the capabilities needed to achieve scientific objectives, from laboratory layouts and digital twins to robotic skills and experimental workflows.

Alán Aspuru-Guzik@aspuru.bsky.social

Tired: Self-Driving Lab Wired: Self-Building Lab arxiv.org/abs/2610.23826

Nature Computational Science

While self-driving laboratories (SDLs) have significantly advanced scientific automation, their capabilities are often limited by predefined equipment, robotic skills, and experimental workflows.

The Matter Lab reposted

@natcomputsci.nature.com

📢Out now! @andreasburger.bsky.social, @thematterlab.bsky.social and colleagues develop a direct approach for accurately and efficiently predicting molecular Hessians. www.nature.com/articles/s43... 🔓 #chemsky

nature.comHIP: Hessian Interatomic Potentials without derivatives - Nature Computational ScienceThe authors introduce an extension of machine learning interatomic potentials that predicts molecular Hessians directly, improving accuracy and speed compared in transition-state searches, geometry op...

The Matter Lab reposted

AC & @thematterlab.bsky.social researchers co-create a new AI research collaborator, El Agente Quntur, supporting researchers using quantum computational chemistry. acceleration.utoronto.ca/news/ac-u-of...

cen.acs.orgAlán Aspuru-Guzik blurs the boundaries between chemistry, AI, and roboticsThe 2024 Trailblazers guest editor also gives chemists the tools to accelerate their science

We are all proud of our director, @aspuru.bsky.social. C&EN’s Trailblazers profile captures his prioneering work across chemistry, AI and robotics, but also the mentor who has welcomed hundreds of scientists to The Matter Lab. cen.acs.org/physical-che...

www.science.org/content/arti...

A few weeks ago, more than 80 current and former graduate students and postdocs came together to celebrate him and inaugurate the Matter Lab International Conference at the Universidad Nacional Autónoma de México (UNAM). What an incredible community 👨‍🔬❤️

The Matter Lab reposted

Marcel Swart

@marcelswart.bsky.social

science.orgHow did Rosalind Franklin miss the helix in her iconic DNA image? She didn’tNew analysis refutes James Watson’s claim he was first to recognize significance of famed photograph 51

The Matter Lab reposted

Alán Aspuru-Guzik

@aspuru.bsky.social

Read this @cenmag articule about @variniabernales.bsky.social bernales.org it is a privilege leading the Matter Lab with her.

bernales.orgVarinia Bernales — AI for Science | University of TorontoVarinia Bernales, Assistant Professor at the University of Toronto, building autonomous agents for quantum chemistry and self-driving laboratories.

C&EN (Chemical & Engineering News)@cenmag.bsky.social

Varinia Bernales develops safer chemicals while making science more equitable This computational chemist advocates for women in science. #HispanicHeritageMonth #chemsky 🧪

cen.acs.orgVarinia Bernales develops chemicals that are safer for humans and the environmentThis computational chemist is also a strong advocate for women in science

The Matter Lab reposted

UofT Department of Chemistry

@chemuoft.bsky.social

Chemistry Stories: Matter Lab researchers unveil new AI research collaborator for quantum chemistry. uoft.me/AgenteQuntur

The Matter Lab reposted

The AC is training today’s talent with the skills in AI, automation and ethics needed to advance accelerated discovery. From self-driving lab 101 workshops to a $1M+ national training initiative, learn more about how the AC is building capacity in our latest impact report: shorturl.at/hVBsI

🚀 New work from the group- DIA: Denoising Intermediate Advantage for Diffusion Policy Optimization Paper: arxiv.org/pdf/2609.12245 Blog: github.com/dia-rl/dia Code: dia-rl.github.io

🚀 We also find that DIA can reach successful states more efficiently and move further from the pretrained behavior distribution, highlighting the importance of assigning credit to the intermediate decisions that construct an action.

Kudos to the authors: Arjun Sohal, @allanzhao.bsky.social, Miroslav Bogdanovic and @aspuru.bsky.social 🎉

🚀 New work from the group- DIA: Denoising Intermediate Advantage for Diffusion Policy Optimization Paper: arxiv.org/pdf/2609.12245 Blog: github.com/dia-rl/dia Code: dia-rl.github.io

These inner-loop learning signals are combined with the standard environment-level PPO advantage, giving the policy more informative feedback during optimization.

🔥 Across benchmarks, DIA improves over state-of-the-art diffusion-policy fine-tuning methods for both state- and image-based policies.

🚀 New work from the group- DIA: Denoising Intermediate Advantage for Diffusion Policy Optimization Paper: arxiv.org/pdf/2609.12245 Blog: github.com/dia-rl/dia Code: dia-rl.github.io

🔑 Key insights: DIA introduces credit assignment within the denoising process itself. We showed that better credit assignment inside the generative process can lead to better policy optimization.

The method learns a value function over partially denoised actions and uses it to construct intermediate advantages for individual denoising decisions.

🚀 New work from the group- DIA: Denoising Intermediate Advantage for Diffusion Policy Optimization Paper: arxiv.org/pdf/2609.12245 Blog: github.com/dia-rl/dia Code: dia-rl.github.io

🤔 Diffusion policies generate actions through multiple denoising steps, but standard RL fine-tuning methods often assign the same environment-level credit across the full denoising trajectory.

The Matter Lab reposted

30+ staff scientists helped lead AC’s core self-driving labs (SDLs) over the last year. Together, these labs pushed accelerated discovery forward and supported R&D with partners. Explore their successes in our latest impact report: shorturl.at/hVBsI

Are you interested in learning about how agentic AI systems like El Agente operate? Curious on how to incorporate them into your research workflow? Join our live webinar as we show El Agente in action 🤖 🗓️ 24 Nov 2026, 10 AM - 11AM EST ✅ Registration & qns: register.gotowebinar.com/register/828...

- Screen metal-organic frameworks containing an X node and a Y ligand for methane or other gas adsorption - Identify plausible transition states for the reaction of A and B leading to product C

Using El Agente Quntur, Gráfico and Estructural, the team will select several questions and demonstrate how El Agente approaches them, based on scientific relevance and what can reasonably be addressed during the webinar. Looking forward to see you there!

Registrants are invited to submit questions through the form above. Examples: - Investigate the IR, UV, and Raman spectra of a series of organic molecules - Evaluate small-molecule binding to transition-metal complexes relevant to investigating a given catalytic activity

The Matter Lab reposted

UofT Department of Chemistry

Responsible solutions help define the AC's research excellence. Learn how the AC puts responsible discovery into action in our latest impact report: shorturl.at/hVBsI

The Matter Lab reposted

@chemuoft.bsky.social

Special Chemistry Seminar: Professor Zhanyun Wang, USDwill discuss “Rethinking Chemical Innovation: From Substances of Very High Concern to Safe and Sustainable by Design,” September 30, 2026 10:00 am Info at www.chemistry.utoronto.ca/events

The Matter Lab reposted