---
title: "APO: Unsupervised Atomic Policy Optimization for 3D Structure Prediction of Atomic Systems"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.28553v1"
permalink: "https://feed.astralor.com/items/itm_dee9fea3d377d91fc3200671"
published: "2026-07-30T17:21:58.000Z"
score: 77
channel: "ai"
tags: ["APO","FlowDPO","Flow Matching","Unsupervised Alignment","Structure Prediction"]
---

# APO: Unsupervised Atomic Policy Optimization for 3D Structure Prediction of Atomic Systems

## AI commentary

> The most notable aspect is replacing expensive experimental labels with physical consistency, setting a new benchmark for unsupervised structure prediction. For teams in materials or antibody computation, APO directly addresses data scarcity; for alignment researchers, the dual-reward design offers a valuable reference.

## At a glance

Predicting the 3D structures of atomic systems is fundamental to materials science and drug discovery. Existing flow-matching models like FlowDPO rely on supervised preference learning to align with ground-truth coordinates, but experimental labels are prohibitively expensive. This work proposes APO, a fully unsupervised alignment framework that removes the need for reference structures. Using a dual-reward mechanism—structural mode reinforcement and thermodynamic stability—APO self-corrects within sampled groups, outperforming fully supervised baselines on crystal and antibody benchmarks while improving inference efficiency.

## Key points

- APO requires no labeled coordinates and aligns via a dual-reward mechanism.
- Outperforms fully supervised baselines on crystal and antibody structure benchmarks, achieving new SOTA.
- Effectively straightens probability paths, significantly improving inference efficiency.
- Demonstrates that intrinsic physical consistency is a superior alignment signal over noisy supervised coordinates.

## Full text

Predicting the 3D structures of atomic systems is fundamental to advancing material science and drug discovery. While flow-matching models (, FlowDPO) have recently shown promise in this domain, their performance relies heavily on alignment with ground-truth coordinates via supervised preference learning. However, obtaining experimental labels for novel crystal phases or de novo proteins is prohibitively expensive, creating a bottleneck for structural modeling in data-scarce regimes. In this work, we propose (Atomic Policy Optimization), a fully unsupervised alignment framework that eliminates the need for ground-truth reference structures. APO adapts group-relative policy optimization to 3D atomic environments, utilizing a novel dual-reward mechanism: (i) a that reinforces the policy's dominant latent structural modes through eigen-decomposition of sample similarities, and (ii) a that enforces thermodynamic stability. Our framework enables the model to ``self-correct'' by identifying physically plausible configurations within sampled groups. Extensive benchmarks on crystal and antibody structure prediction demonstrate that APO consistently outperforms fully supervised baselines, achieving a new state-of-the-art in match rates and structural fidelity. Furthermore, we show that APO effectively straightens probability paths, significantly improving inference efficiency. Our results suggest that intrinsic physical consistency can serve as a superior guide for alignment compared to noisy, supervised coordinate matching.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_dee9fea3d377d91fc3200671