Autoregressive Boltzmann Generators

2026-06-25 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computational Chemistry & Molecular Modeling · Depth: Expert, quick

Summary

Autoregressive Boltzmann Generators (ArBG) introduce a novel autoregressive modeling framework designed to enhance the efficient sampling of molecular systems at thermodynamic equilibrium. This approach addresses limitations of existing Boltzmann Generators (BGs) that rely on normalizing flows (NFs), which often suffer from restricted expressivity or high computational costs for likelihoods. ArBG circumvents the topological constraints inherent in flow-based models, allows for sequential inference-time interventions, and improves scalability by adopting architectures effective in Large Language Models. Empirical evaluations demonstrate ArBG's significant performance improvements over flow-based models, particularly in larger peptide systems such as the 10-residue Chignolin. Furthermore, the paper introduces Robin, a 132 million parameter transferable model built with the ArBG framework, which reduces the zero-shot energy error, E-W$_2$, on 8-residue systems by over 60% compared to previous benchmarks.

Key takeaway

For research scientists focused on molecular dynamics and equilibrium sampling, consider adopting Autoregressive Boltzmann Generators (ArBG) to overcome the expressivity and computational limitations of normalizing flow-based methods. Your simulations, especially with larger peptide systems like Chignolin, could see significant performance improvements. Furthermore, explore integrating the 132 million parameter Robin model, trained with ArBG, to potentially reduce zero-shot energy error on 8-residue systems by over 60%.

Key insights

Autoregressive Boltzmann Generators (ArBG) enhance molecular system sampling efficiency by replacing normalizing flows with autoregressive models.

Principles

• Autoregressive models overcome flow-based topological constraints.
• LLM architectures enhance scalability in molecular modeling.
• Sequential inference-time interventions are enabled by ArBG.

Method

Autoregressive Boltzmann Generators (ArBG) employ an autoregressive modeling framework, departing from flow-based paradigms, to generate uncorrelated equilibrium samples for molecular systems.

In practice

• Apply ArBG for efficient sampling of large peptide systems.
• Utilize the Robin model for 8-residue system energy error reduction.
• Explore ArBG for enhanced scalability in molecular simulations.

Topics

• Molecular Dynamics
• Boltzmann Generators
• Autoregressive Models
• Normalizing Flows
• Peptide Systems
• Robin Model

Code references

• danyalrehman/autobg

Best for: AI Scientist, Research Scientist

Related on AIssential

• Autoregressive Boltzmann Generators — ArBG offers a new autoregressive approach for molecular sampling, outperforming flow-based Boltzmann Generators, especially for larger systems.
• Reward Weighted Classifier-Free Guidance as Policy Improvement in Autoregressive Models — RCFG enables autoregressive models to optimize new reward functions at test time without re-training.
• Foundations of Molecular Generation with GP-MoLFormer on AMD Instinct MI300X Accelerators — GP-MoLFormer enables efficient, large-scale molecular generation using sequence-based language models on AMD GPUs.
• Future Forcing: Future-aware Training-free KV Cache Policy for Autoregressive Video Generation — Future Forcing uses stable pre-RoPE query distributions to enable training-free, future-aware KV cache management for AR video generation.
• Energy-Weighted Flow Matching: Unlocking Continuous Normalizing Flows for Efficient and Scalable Boltzmann Sampling — Energy-Weighted Flow Matching enables continuous normalizing flows to model Boltzmann distributions using only energy evaluations, not target samples.
• Automating the Expert Eye: A System-Agnostic Deep Learning Framework for Rare Event Discovery in Imbalanced Force Spectroscopy — Deep learning automates rare event discovery in SMFS, overcoming extreme data imbalance with high accuracy and interpretability.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.