arXiv subjects

Zhenlin Zhao

Publications and source records attributed to Zhenlin Zhao.

4 recordsLinked to original sources

arXiv2026

Self-Evolving Skills via Surrogate-Guided Solve-and-Reproduce

Agent skills are portable packages of instructions and resources an agent consults at deployment. Self-evolving them fails in two ways today. First, skills evolved from scratch underperform human-curated ones and, on a weak model, using no skill at all. Second, an evolution-time pass records one lucky trajectory that a fresh stochastic agent often fails to reproduce at deployment. We present reSolve, a per-task, oracle-in-the-loop framework built on three components. It decouples interactive solving from a self-contained deliverable that is independently re-executed in a fresh container, a protocol we call solve-and-reproduce. It enhances the sparse reward signal with a surrogate verifier that cannot access hidden tests or reference answers. It then runs verifier-guided beam search over a solution-construction graph. Within a fixed harness, a cheap model self-evolves skills that reach $74.9\%$ mean-of-3, $+14.8$ points over the $60.1\%$ human-curated baseline, exceeding the strongest official curated-skill result ($67.3\%$, GPT-5.5/OpenHands). We also report observed failure cases and domain-level results, including performance on the 14 Natural Science tasks, to clarify when the approach does and does not help.

cs.AI↗

arXiv2026

RamanPFN: learning from Raman spectral structure with a tabular foundation model

Raman spectroscopy enables non-destructive, label-free molecular characterization across materials science, biomedicine and process monitoring. Predictive Raman datasets often contain few labelled spectra and thousands of ordered wavenumbers, with informative variation within bands and across distant spectral regions. Latent-variable chemometrics accommodates collinear small-sample data but can obscure fine peak morphology, whereas deep spectral networks resolve this structure only after task-specific training. TabPFN avoids task-specific parameter fitting through pretrained in-context inference, but processes very wide inputs as feature-subsampled views that do not preserve joint visibility of related bands. We present RamanPFN, a spectral representation framework that encodes these dependencies before TabPFN inference. Global Compositional Unmixing constructs non-negative coordinates over the complete spectrum so that distant bands with shared latent variation occupy a common predictive axis. Local Vibrational Subspace Encoding represents contiguous wavenumber regions with multiple orthogonal modes that retain independent changes in peak shape, intensity and position. The representations are evaluated separately and combined at the prediction level. Evaluation covered 150 tasks from 74 public Raman datasets. RamanPFN reduced root-mean-square error by 19.6% on average across 129 regression targets relative to direct TabPFN inference and further reduced the remaining classification error by 9.0% across 21 classification tasks. These results establish explicit spectral representation as an effective interface between high-dimensional Raman measurements and reusable tabular inference.

cs.LG↗

arXiv2026

Chem World: A Large-Scale Benchmark and Physics-Informed Framework for Trustworthy Chemical Property Prediction

Chemical property prediction plays a critical role in accelerating scientific discovery in chemistry, materials science, and drug development. However, existing benchmarks often suffer from limited task diversity, fragmented datasets, and inconsistent evaluation protocols, making it challenging to systematically assess the reliability and generalization of AI models. In this work, we introduce Chem World, a comprehensive benchmark for chemical property prediction that integrates 17 diverse chemical datasets with over 800,000 molecular samples, covering various properties including density, electrical conductivity, solubility, and other molecular characteristics. Chem World provides a unified platform for evaluating AI models across multiple property prediction tasks. Furthermore, we propose Mixture-PINN, a physics-informed neural network based prediction framework that incorporates chemical prior knowledge into data-driven learning, improving the accuracy, robustness, and reliability of chemical property prediction. Extensive experiments on Chem World demonstrate the effectiveness of our approach compared with existing methods. By combining large-scale standardized evaluation with physics-informed learning, Chem World establishes a foundation for developing trustworthy AI systems for computational chemistry and advancing AI-driven scientific discovery.

cs.LG↗

arXiv2026

Deterministic Minimum-Leakage Continuous-Variable Quantum Key Distribution with Phase-Conjugated Twin Beams

Minimum-leakage continuous-variable quantum key distribution suppresses Eve's Holevo information by engineering the signal ensemble during state preparation. The existing symmetric realization relies on Alice-side heralding. Here we propose an alternative two-mode symmetric realization that is deterministic without heralding. The minimum-leakage condition of the new protocol requires transmitting two phase-conjugated twin beams (PCTB). We show that the PCTB protocol is related to the heralding protocol through a common entanglement-based source but corresponds to a different prepare-and-measure decomposition. The two protocols attain the same secret key rate per transmitted optical mode in the very-large-squeezing limit. For finite squeezing, however, the PCTB protocol requires approximately 3 dB less squeezing to achieve the same performance. We further optimize mode-symmetric Gaussian two-mode entangling-cloner attacks and find that correlated ancillary modes provide only a limited advantage over independent attacks under the minimum-leakage condition. These results establish phase-conjugated twin beams as a deterministic and experimentally appealing route to symmetric minimum-leakage CV-QKD.

quant-ph↗