Autoregressive Boltzmann Generators

2026-06-25 · Source: Takara TLDR - Daily AI Papers · Field: Science & Research — Physical Sciences & Chemistry, Mathematics & Computational Sciences · Depth: Expert, quick

Summary

Autoregressive Boltzmann Generators (ArBG) introduce a novel framework for efficient sampling of molecular systems at thermodynamic equilibrium, addressing limitations of existing Boltzmann Generators (BGs) that rely on normalizing flows (NFs). NFs suffer from either limited expressivity due to strict invertibility constraints or computationally expensive likelihoods. ArBG circumvents these topological constraints and enables sequential inference-time interventions, achieving enhanced scalability by using architectures effective in Large Language Models. Empirical demonstrations show ArBG significantly improves over flow-based models across all benchmarks, particularly in larger peptide systems such as the 10-residue Chignolin. Furthermore, the authors introduce Robin, a 132 million parameter transferable model trained with the ArBG framework, which reduces the zero-shot energy error, E-W$_2$, on 8-residue systems by over 60% compared to the previous state-of-the-art.

Key takeaway

For research scientists focused on molecular system simulations, Autoregressive Boltzmann Generators (ArBG) present a significant advancement over traditional normalizing flow-based methods. You should consider ArBG for its enhanced scalability and ability to handle larger peptide systems, potentially reducing computational costs and improving accuracy. Specifically, the Robin model, with its 132 million parameters, offers a proven solution for reducing zero-shot energy error on 8-residue systems by over 60%.

Key insights

ArBG offers a new autoregressive approach for molecular sampling, outperforming flow-based Boltzmann Generators, especially for larger systems.

Principles

• Autoregressive models can overcome flow-based topological constraints.
• LLM architectures enhance scalability in molecular sampling.
• Sequential inference-time interventions improve system analysis.

Method

ArBG uses an autoregressive framework to generate uncorrelated equilibrium samples, departing from normalizing flows and enabling sequential inference-time interventions.

In practice

• Apply ArBG for efficient molecular system sampling.
• Use Robin model for 8-residue system energy error reduction.
• Explore ArBG for larger peptide systems like Chignolin.

Topics

• Autoregressive Models
• Boltzmann Generators
• Molecular Dynamics
• Peptide Systems
• Robin Model
• Statistical Physics

Code references

• danyalrehman/autobg

Best for: AI Scientist, Research Scientist

Related on AIssential

• Force-free molecular dynamics through autoregressive equivariant networks — TrajCast uses equivariant MPNNs to directly predict molecular states, enabling faster, data-efficient MD simulations with strong generalization.
• (HB-ARFM) History-Bootstrapped Flow Matching for Inverse Boiling Reconstruction — HB-ARFM reconstructs spatiotemporal fields from partial observations by bootstrapping initial states with history and autoregressively propagating predictions.
• GPT-Based Fast Simulation of CLAS12 Detector Hits via Conditional Autoregressive Generation — A GPT-style transformer rapidly simulates CLAS12 detector hits, maintaining physics fidelity with significant speedup over traditional methods.
• CARD: Coarse-to-fine Autoregressive Modeling with Radix-based Decomposition for Transferable Free Energy Estimation — CARD offers a generative framework for free energy estimation, combining radix-based decomposition with autoregressive modeling for enhanced generalization.
• GPT-5 lowers the cost of cell-free protein synthesis — GPT-5 connected to a robotic lab significantly reduced cell-free protein synthesis costs through autonomous closed-loop optimization.
• Sample Efficient Generative Molecular Optimization with Joint Self-Improvement — Joint Self-Improvement optimizes molecular design by integrating generative and predictive models to overcome sample scarcity and distribution shift.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.