arXiv subjects

Dengpan Dong

Publications and source records attributed to Dengpan Dong.

4 recordsLinked to original sources

arXiv2026

VERDICT: Agreement Beats Pixel-Space Verification in Real-Document OCSR

Optical Chemical Structure Recognition (OCSR) converts 2D molecular depictions in the published literature into SMILES, and is increasingly important for constructing large-scale chemical training datasets. Automation at that scale requires identifying unreliable predictions in the absence of ground truth. Three families of label-free signals were compared on $263$ ACS journal depictions with verified ground truth: model confidence, re-rendering similarity, and agreement among recognizers. Pixel-space re-rendering performed little better than chance (AUROC $0.547$, $95\%$ CI $[0.465,0.629]$), and an oracle-tuned threshold on it reduced correct labels per image from $0.745$ to $0.205$. Agreement among four architecturally distinct recognizers instead reached an AUROC of $0.916$ ($[0.880,0.952]$). The two-of-four rule accepted $81.7\%$ of images at $88.8\%$ precision, the three-of-four rule $52.1\%$ at $98.5\%$. The same pattern held on CLEF-IP, UOB, and USPTO. This distinction is obscured on synthetic benchmarks, where re-rendered predictions naturally resemble their inputs. A substance filter removed $2{,}193$ false agreements on wildcards and R-group fragments, after which the three-of-four rule rejected all $68$ generic depictions. VERDICT was then applied to PMC Open Access, producing $6{,}146$ structure labels for $4{,}833$ molecules; chemist adjudication of $400$ released labels in two independent samples yielded precisions of $0.995$ for the three-of-four tier and $0.958$ for the two-of-four tier. VERDICT therefore enables validated labels for multimodal molecular databases linking structure images, machine-readable representations, and source-publication information. In SES AI's Molecular Universe platform, VERDICT further serves as an image-based interface for searching and retrieving molecular records.

cs.CV↗

arXiv2026

Real Data Closes Synthetic-to-Real Gap in Optical Chemical Structure Recognition

Millions of chemical structures appear in patents and papers only as drawings, and using that information at scale requires reading the drawings. OCSR appears nearly solved on synthetic images yet remains difficult on real documents: the starting recognizer, Qwen2.5-VL-7B, exceeds 91% accuracy on synthetic renders but falls below 16% on three real-world benchmarks (ACS, CLEF-IP, USPTO). To identify the main source of improvement, 21 recognizers were fine-tuned on mixtures of synthetically rendered structures and labeled real depictions from patents, journal figures, and hand-drawn collections, varying the vision language model (VLM) base, the fraction of real training data, and the vision-tower adaptation strategy. Labeled real training images make the largest difference. For Qwen2.5-VL, ACS exact match rises from 0.15 with no real data to 0.37 at 9.5% and 0.46 at 50.2%; a controlled experiment across three base models reproduces the trend. A vision-tower LoRA, in contrast, does nothing for Qwen (+0.00, paired p=1.00), substantially helps InternVL3-8B (+22.8 to +34.6 pt), and modestly helps GLM-4.1V-9B (+1.0 to +9.6 pt), so its value depends on the base model. The best configuration reaches 0.96 exact match on clean renders and 0.49, 0.65, 0.84, and 0.76 on ACS, CLEF-IP, UOB, and USPTO, respectively. Gaps between base models are largest without real data (0.21), shrink to 0.06 at 70% real data, and reorder the ranking; base model and real-data mixture must therefore be selected together. Small-scale experiments on handwritten image-to-LaTeX recognition and chart-to-table conversion show that base-model rankings also vary beyond chemistry. More generally, model and adaptation choices for visual structure recognition should be evaluated on the target task.

cs.LG↗

arXiv2020

First-Principles Experimental Demonstration of Ferroelectricity in a Thermotropic Nematic Liquid Crystal: Spontaneous Polar Domains and Striking Electro-Optics

We report the experimental determination of the structure and response to applied electric field of the lower-temperature nematic phase of the previously reported calamitic compound 4-[(4-nitrophenoxy)carbonyl]phenyl2,4-dimethoxybenzoate (RM734). We exploit its electro-optics to visualize the appearance, in the absence of applied field, of a permanent electric polarization density, manifested as a spontaneously broken symmetry in distinct domains of opposite polar orientation. Polarization reversal is mediated by field-induced domain wall movement, making this phase ferroelectric, a 3D uniaxial nematic having a spontaneous, reorientable, polarization locally parallel to the director. This polarization density saturates at a low temperature value of ~ 6 microcoulombs/cm-sqd, the largest ever measured for an organic material or for any fluid. This polarization is comparable to that of solid state ferroelectrics, and is close to the average value obtained by assuming perfect, polar alignment of molecular long axes in the nematic. We find a host of spectacular optical and hydrodynamic effects driven by ultra-low applied field (E~1V/cm), produced by the coupling of the large polarization to nematic birefringence and flow. Electrostatic self-interaction of the polarization charge renders the transition from the nematic phase mean-field-like and weakly first-order, and controls the director field structure of the ferroelectric phase. Atomistic molecular dynamics simulation reveals short-range polar molecular interactions that favor ferroelectric ordering, including a tendency for head-to-tail association into polar, chain-like assemblies having polar lateral correlations. These results indicate a significant potential for transformative new nematic science and technology based on the enhanced understanding, development, and exploitation of molecular electrostatic interaction.

cond-mat.soft↗

arXiv2019

Liquid-Like Interfaces Mediate Structural Phase Transitions in Lead Halide Perovskites

Microscopic pathways of structural phase transitions are difficult to probe because they occur over multiple, disparate time and length scales. Using $in$ $situ$ nanoscale cathodoluminescence microscopy, we visualize the thermally-driven transition to the perovskite phase in hundreds of non-perovskite phase nanowires, resolving the initial nanoscale nucleation and subsequent mesoscale growth and quantifying the activation energy for phase propagation. In combination with molecular dynamics computer simulations, we reveal that the transformation does not follow a simple martensitic mechanism, and proceeds via ion diffusion through a liquid-like interface between the two structures. While cations are disordered in this liquid-like region, the halide ions retain substantial spatial correlations. We find that the anisotropic crystal structure translates to faster nucleation of the perovskite phase at nanowire ends and faster growth along the long nanowire axis. These results represent a significant step towards manipulating structural phases at the nanoscale for designer materials properties.

cond-mat.mes-hall↗