QMBL: the Quantum Many-Body Leaderboard. A live record of published ground-state energies for lattice Hamiltonians

In the vast landscape of condensed matter physics, scientists study how matter behaves when billions of atoms interact with one another. These interactions give rise to complex states, such as superconductors or magnets, which are defined by their lowest possible energy state, known as the ground state. Determining this exact energy is like finding the deepest valley in a mountain range; it tells researchers exactly how stable a material is and what phases it can enter. However, calculating this energy for large groups of atoms is incredibly difficult, often requiring powerful computers and clever mathematical shortcuts. For years, different research groups have used various methods to estimate these values, but they have often spoken different languages, using different units or conventions that made it hard to compare their results directly. Without a shared standard, the scientific community struggles to reach a consensus on which methods are truly the most accurate or which problems remain unsolved.

To solve this problem of scattered information, Tristan Zerweck and Juan Carrasquilla from the Institute for Theoretical Physics at ETH Zurich have created a new, living record called the Quantum Many-Body Leaderboard. This digital platform acts as a central hub where the community can see every published attempt to calculate the ground-state energy for specific arrangements of atoms. Instead of leaving these findings buried in thousands of separate academic papers, the leaderboard gathers them into a single, interactive view. The project currently holds 1,434 energy values covering 395 different scenarios across six major families of physical models. These scenarios vary by the size of the system, the shape of the grid the atoms sit on, and the strength of the forces between them. By bringing these numbers together, the researchers have turned a fragmented collection of data into a clear map of scientific progress.

The creation of this leaderboard relied on a significant shift in how data is collected. While gathering such information manually would have been a tedious and slow task just a few years ago, the authors utilized artificial intelligence agents to do the heavy lifting. These digital assistants read through scientific papers, translated the different ways researchers reported their numbers into a common language, and checked every single value against its original source to ensure accuracy. This process allowed them to build a comprehensive database that includes not just the final numbers, but also details about the methods used, the estimated cost of the computer time required, and the uncertainty associated with each measurement. The result is a resource where a scientist can instantly see the current best estimate for a specific problem and understand how it compares to previous attempts.

A key feature of this work is its focus on transparency and open science. The leaderboard distinguishes between results that are mathematically exact and those that are estimates, or variational bounds, which are educated guesses that aim to be as close to the truth as possible. When an exact solution exists, it is highlighted as the record to beat. When no exact solution is known, the lowest energy found so far becomes the new benchmark. The project specifically identifies 153 instances where no exact solution is known, marking these as open problems for the community to solve. These unsolved cases serve as the targets for a companion benchmark, inviting researchers to push the boundaries of what is computationally possible. The data is entirely open source, allowing anyone to explore the details, contribute new results, or correct errors, ensuring that the record remains a dynamic and evolving reflection of the field's collective effort.

The utility of this leaderboard extends beyond simple data storage; it provides a way to evaluate the efficiency of different computational methods. The authors have included estimates of the computing power required to achieve each result, plotting energy accuracy against the number of parameters used in the calculations. This allows researchers to see which methods offer the best balance between precision and cost. For example, the data shows how different types of neural networks and mathematical models perform when trying to solve the same physical problem. By visualizing these trade-offs, the leaderboard helps the community understand which tools are most effective and where the limits of current technology lie. It transforms a list of numbers into a story of scientific competition and collaboration, where every new entry represents a step toward a deeper understanding of the quantum world.

Ultimately, this project represents a new standard for how scientific results are shared and validated in the field of quantum physics. By standardizing how results are reported and making them easily accessible, the Quantum Many-Body Leaderboard removes the barriers that once kept valuable data isolated. It encourages a culture where errors are openly flagged and corrected, and where the best available knowledge is always just a click away. As the database grows, it will continue to serve as the definitive reference for ground-state energies, guiding future research and helping scientists agree on the fundamental properties of matter. The work stands as a testament to the power of combining human curiosity with modern computational tools to create a shared foundation for discovery.