arXiv subjects

Zhihong Chen

Publications and source records attributed to Zhihong Chen.

At least 19 recordsLinked to original sources

arXiv2026

Process-Technology Co-optimization for 2D-FETs

We present the first experimental machine learning (ML)-enabled Process-Technology Co-Optimization (PTCO) framework for optimizing 2D transition metal dichalcogenide (TMD) FET fabrication directly from statistically meaningful experimental data rather than pure simulation data. We first introduce a transition voltage metric, VTrans, to quantify the gate voltage required for off-to-on switching and reveal its direct correlation with subthreshold swing (SS), highlighting an overlooked switching characteristic that governs both off-state and on-state performance. By integrating automated metric extraction, multi-objective recipe ranking, and predictive modeling, our framework uncovers hidden process-performance correlations and predicts the performance of unexplored fabrication recipes from limited experimental data. Experimental validation shows close agreement with ML predictions, thus demonstrating the framework's ability to efficiently guide gate stack optimization through iterative experimental feedback.

cond-mat.mes-hall↗

arXiv2026

Beyond Semantic IDs: Encoding Business-Value Ranking into Document Identifiers for Generative Retrieval

Generative Retrieval (GR) formulates retrieval as a sequence-to-sequence generation task, assigning each document a document identifier (DocID) and retrieving it through autoregressive decoding, making DocID design a critical factor in retrieval quality. However, existing schemes based on discrete representation learning suffer from inherent collision issues and create a mismatch between the DocID's encoding objective and the system's business optimization target. To address these limitations, we propose \textbf{Cluster-Ranked Identifier (CRID)}, which decouples DocID into \textit{semantic clustering} and \textit{business-value ranking}, yielding collision-free identifiers that support incremental updates via intra-cluster reranking. We further introduce an analytical framework that decomposes retrieval gains into \textit{personalized preference} and \textit{statistical prior} generalization, revealing how semantic cluster size governs the balance between the two components. Experiments on a Taobao e-commerce corpus of over 300M items show that CRID surpasses the strongest embedding-based retrieval baseline on top-K Hitrate, and delivers +1.06\% GMV in full-traffic deployment.

cs.IR↗

arXiv2026

OpenGPT-4o-Image: A Comprehensive Dataset for Advanced Image Generation and Editing

The performance of unified multimodal models for image generation and editing is fundamentally constrained by the quality and comprehensiveness of their training data. While existing datasets have covered basic tasks like style transfer and simple object manipulation, they often lack the systematic structure and challenging scenarios required for real-world applications. To address this bottleneck, we introduce OpenGPT-4o-Image, a large-scale dataset constructed using a novel methodology that combines hierarchical task taxonomy with automated data generation. Our taxonomy not only includes fundamental capabilities such as text rendering and style control but also introduces highly practical yet challenging categories like scientific imagery for chemistry illustrations and complex instruction editing requiring simultaneous execution of multiple operations. Through an automated pipeline leveraging structured resource pools and GPT-4o, we generate 80k high-quality instruction-image pairs with controlled diversity, covering 11 major domains and 51 subtasks. Extensive experiments show that fine-tuning leading models on our dataset achieves significant performance gains across multiple benchmarks, with improvements of up to 18\% on editing tasks (UniWorld-V1 on ImgEdit-Bench) and 13% on generation tasks (Harmon on GenEval). Our work demonstrates that systematic data construction is key to advancing multimodal AI capabilities.

cs.CV↗

arXiv2026

Spin Vector Control for Heisenberg-Inspired Probabilistic Computing

Probabilistic bits (p-bits) have emerged as a cornerstone of probabilistic computing, enabling energy-efficient hardware implementation for probabilistic inference and combinatorial optimization. A critical challenge in advancing this field beyond binary p-bits lies in realizing and manipulating vector spin information, essential for mapping complex energy-based models such as the Heisenberg Hamiltonian.Here, we demonstrate a spintronic platform capable of real-space vector summation by using dual ferromagnetic spin injections into a monolayer graphene channel. By electrically tuning the spin polarization through independently controlled injection currents, we achieve continuous control over the magnitude and direction of the resulting spin accumulation vector. Experimental observations, supported by theoretical vector summation models and spin-circuit simulations, reveal coherent vector interactions and angular tunability of the spin state. This approach enables direct implementation of vector-based spin logic and lays the groundwork for mapping classical Heisenberg models using stochastic low-barrier magnets. Our results establish a scalable pathway for realizing probabilistic spin circuits based on two-dimensional materials, offering new opportunities for low-power, non-Boolean computing architectures.

cond-mat.mes-hall↗

arXiv2026

TSGR: Taobao Search Generative Retrieval

Generative retrieval (GR) has demonstrated strong promise for industrial e-commerce search by training a single autoregressive model to directly generate the Semantic IDs (SIDs) of target items. However, existing GR systems are primarily optimized for semantic matching and remain insensitive to item business value: SID construction is value-unaware, and candidates are ranked without access to item side-info. Consequently, high-value items are often missed or deprioritized at the retrieval stage, limiting downstream business impact. This limitation is particularly critical in industrial settings such as Taobao Search, where business objectives are central to system design. To address this, we propose $\textbf{T}$aobao $\textbf{S}$earch $\textbf{G}$enerative $\textbf{R}$etrieval ($\textbf{TSGR}$), a unified generative retrieval framework that incorporates value awareness into both item representation and candidate ranking. 1) For item representation, TSGR introduces $\textbf{Query-aware Parallel SID (QP-SID)}$, which encodes query-conditioned value orderings into the SID construction by building parallel codebooks derived from query-item statistics, so that higher-value and query-relevant items are assigned better token indices. 2) For candidate ranking, we introduce a $\textbf{Value-aware Ranking Module (VRM)}$ that is built upon and jointly optimized with the GR, enabling a single model to seamlessly serve as both retriever and pre-ranker without a dedicated pre-ranking stage. A progressive training pipeline further aligns the model with semantic relevance, user preferences, and business objectives. Offline experiments show that TSGR achieves a 9.16\% improvement in HR@1000, and online A/B tests further validate its effectiveness, yielding gains of +0.43\% in IPV, +1.12\% in Transaction Count, and +1.64\% in GMV. TSGR has been fully deployed in production.

cs.IR↗

arXiv2026

Prompt Generation Technical Report

Generative retrieval has become an increasingly adopted paradigm for industrial search, recommendation, and advertising systems, delivering significant online gains. Most existing work combines user behavior sequences with large language models (LLMs) to model user preferences. In practice, feature engineering remains critical to model effectiveness, yet its complexity slows offline iteration and makes online deployment heavy and hard to reuse, all under tight online latency budgets. The root cause is a tight coupling between feature-processing logic and model architecture, where every feature change touches the training and serving code and resists reuse across scenarios. To break this coupling, we present Prompt Generation (PG), a high-level tokenizer and configuration-driven framework that decouples feature-processing logic from model architecture through two declarative JSON files, which serve as the single source of truth for both offline training and online serving, ensuring feature consistency across the two stages. Organizing features under four types with three composable processing components to assemble and compress heterogeneous features, PG delivers acceleration at three levels: (1)fast training iteration: feature experiments require only configuration changes, with built-in token compression for ultra-long sequences; (2)fast deployment: a new scenario only needs to conform to the PG schema and plug into a universal pipeline, with no scenario-specific engineering; (3)fast online inference: engine applies unified optimizations over the standardized configuration, reducing PG's overhead to a negligible level. PG has been deployed on Taobao Search with statistically significant online A/B uplifts of +0.47% in transaction count and +0.51% in GMV, and has been applied across multiple Taobao search and recommendation teams as the iteration framework for generative retrieval.

cs.IR↗

arXiv2026

Seed Layer Engineering for Effective Charge Transfer Doping of MoS$_2$ Transistors

Integrating two-dimensional semiconductors such as MoS$_2$ with dielectric materials remains a central challenge for their use in future logic technologies. While seed layers are typically introduced to promote dielectric nucleation and adhesion, we show that they also critically govern charge-transfer doping and, in turn, transistor performance. Back-gated monolayer MoS$_2$ transistors passivated on their top-surface with a Ta-seed/HfO$_x$ dielectric stack were fabricated and characterized electrically and physically using Raman, photoluminescence, and X-ray photoelectron spectroscopies. Threshold voltage and on-current varied strongly with Ta-seed thickness and deposition conditions, and these changes correlated with signatures observed across all spectroscopic probes. The results reveal that the seed layer both introduces disorder into the MoS$_2$ channel and modifies the interfacial charge environment controlling charge transfer between HfO$_x$ and MoS$_2$. Optical spectroscopy shows that on-current tracks seed-induced disorder, whereas X-ray photoelectron spectroscopy indicates that threshold voltage correlates with shifts in the local electrostatic environment associated with interfacial charge transfer. Better performance was obtained with ultrathin 0.2 nm Ta seed layers deposited under oxygen-poor conditions, which limit deposition-induced damage while facilitating charge transfer. These findings identify seed-layer engineering as a key strategy for controlling disorder and interfacial doping in MoS$_2$ devices and establish multimodal spectroscopy as a practical during-fabrication approach for process development and monitoring.

cond-mat.mtrl-sci↗

arXiv2026

Scaling Two-Dimensional Semiconductor Nanoribbons for High-Performance Electronics

As silicon transistors scale toward future technology nodes, three-dimensional architectures -- including gate-all-around (GAA) nanoribbon and complementary field-effect transistors (CFETs) -- require channel widths in the tens of nanometers to meet density targets. Monolayer transition metal dichalcogenides (TMDs), with their atomically thin bodies, are promising channel materials for these architectures, yet most TMD-based FETs remain limited to micrometer-scale widths. Here, we show that channel width scaling of monolayer MoS2 nanoribbon transistors not only preserves but also enhances device performance. Reducing the channel width from hundreds of nanometers to $\sim$30--40 nm increases the median on-current density by $\sim$42% and reduces the median subthreshold swing by $\sim$16%, with a champion device reaching 995 $μ$A $μ$m$^{-1}$ at a drain-to-source voltage of 1 V and an overdrive voltage of 2.5 V. We attribute these improvements to three mechanisms: minimal edge-induced disorder, enhanced gate electrostatics at ribbon edges, and more efficient side-contact injection, together reducing contact resistance from $\sim$860 $Ω$ $μ$m to $\sim$270 $Ω$ $μ$m. Extending the platform to n-type WS2 and p-type WSe2 FETs, we achieve WSe2 p-FET on-currents of 357 $μ$A $μ$m$^{-1}$. These findings suggest that monolayer TMD nanoribbon FETs are promising candidates for future ultra-scaled electronics.

cond-mat.mtrl-sci↗

arXiv2026

Experimental Demonstration of an On-Chip CMOS-Integrated 3T-1MTJ Probabilistic Bit -- A P-Bit

Ongoing semiconductor scaling challenges and the rise of neuromorphic computing have sparked interest in exploring novel computing schemes to achieve higher power efficiency and computational capabilities. Probabilistic computing is one candidate that endows low power consumption, capability of solving probability-encoded computational problems, and the ease of integration with existing CMOS technology. A basic building block of this scheme is the probabilistic bit (P-Bit), which utilizes a novel device such as a stochastic magnetic tunnel junction (sMTJ) to generate tunable randomness by nature. This work presents the first experimental demonstration of a fully CMOS-integrated sMTJ-based P-Bit, capable of generating rail-to-rail stochastic output with a mere collection of 3 transistors + 1 sMTJ. Furthermore, simulations also confirm this P-Bit's functionality in probabilistic logic circuits. The demonstration of such P-Bit paves the way towards realizing monolithic large-scale probabilistic computing architecture on CMOS chips.

cs.ET↗

arXiv2026

CheXmix: Unified Generative Pretraining for Vision Language Models in Medical Imaging

Recent medical multimodal foundation models are built as multimodal LLMs (MLLMs) by connecting a CLIP-pretrained vision encoder to an LLM using LLaVA-style finetuning. This two-stage, decoupled approach introduces a projection layer that can distort visual features. This is especially concerning in medical imaging where subtle cues are essential for accurate diagnoses. In contrast, early-fusion generative approaches such as Chameleon eliminate the projection bottleneck by processing image and text tokens within a single unified sequence, enabling joint representation learning that leverages the inductive priors of language models. We present CheXmix, a unified early-fusion generative model trained on a large corpus of chest X-rays paired with radiology reports. We expand on Chameleon's autoregressive framework by introducing a two-stage multimodal generative pretraining strategy that combines the representational strengths of masked autoencoders with MLLMs. The resulting models are highly flexible, supporting both discriminative and generative tasks at both coarse and fine-grained scales. Our approach outperforms well-established generative models across all masking ratios by 6.0% and surpasses CheXagent by 8.6% on AUROC at high image masking ratios on the CheXpert classification task. We further inpaint images over 51.0% better than text-only generative models and outperform CheXagent by 45% on the GREEN metric for radiology report generation. These results demonstrate that CheXmix captures fine-grained information across a broad spectrum of chest X-ray tasks. Our code is at: https://github.com/StanfordMIMI/CheXmix.

cs.CV↗

arXiv2026

RLHF in an SFT Way: From Optimal Solution to Reward-Weighted Alignment

Reinforcement Learning from Human Feedback (RLHF) is crucial for aligning Large Language Models (LLMs) with human values. However, RLHF has been continuously challenged by its high complexity in implementation and computation consumption, specifically for online sampling-based methods like Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO). Even with recent simplifications, such as Direct Preference Optimization (DPO) that designs an offline implicit reward learning objective relying on pre-collected preference datasets, the problems of over-fitting and training instability remain hindering the alignment process from the expected optimal performance. To address the existing challenges, we propose a novel simplification of RLHF from the perspective of variational inference, called Variational Alignment with Re-weighting (VAR). Specifically, by directly minimizing the distribution gap between the learning LLM policy and the optimal solution of RLHF, we transform the alignment objective into an offline reward-driven re-weighted supervised fine-tuning (SFT) form, which only requires minor adjustment on the SFT loss to obtain noticeable improvement on training stability and effectiveness. In comprehensive evaluation benchmarks, our objective empowers LLMs to outperform offline alignments, demonstrating superior performance in both helpfulness and harmlessness metrics (avg. $\uparrow7.16\%$ than DPO). Meanwhile, when compared to online sampling methods, our method is also comparable even better while significantly reducing computational overhead and accelerating convergence speed (over $5\times$ faster than GRPO), suggesting our approach as an efficient and effective solution in bridging the gap between efficiency and performance in LLM alignment.

cs.LG↗

arXiv2026

Evaluating data-driven background ensembles covariances from Graphcast: a case study for Hurricane Lee (2023)

Short-term background ensemble covariances (BEC) are crucial for ensemble-based data assimilation (DA). However, limited studies so far have examined the fidelity of the cost-effective data-driven model in producing the short-term BEC for hurricane data assimilation. In this study, we evaluate the background ensemble spread and correlations from GraphCast against those of GEFS for Hurricane Lee (2023) during both its intensification and non-intensification phases. Specifically, the BEC in the hurricane vortex, the hurricane environment, and the vortex-environment interactions are examined. Within the hurricane vortex, the background ensemble of GraphCast is less dispersive than GEFS. The two models agree well on the background ensemble correlations that are tied to the primary circulation but show a larger correlation difference associated with the secondary circulation, indicating the two models represent unbalanced and diabatic processes differently. In the hurricane environment, binned univariable correlations show linear relationships between the two models, with a weaker horizontal geopotential height correlation in GraphCast. GraphCast also shows a reduced spread and a flatter empirical orthogonal function spectrum of the 500 hPa geopotential height background ensemble, with more perturbation growth distributed to smaller-scale features such as shortwaves. For the vortex-environment interaction, the two models produce close background ensemble correlation patterns for Lee's track but differ more for intensity. Overall, GraphCast can produce broadly consistent short-term BEC for Hurricane Lee compared to those of GEFS. However, systematic difference exists in certain variables, scales, and processes, suggesting the need to further investigate its fidelity in a cycled hurricane DA context.

physics.ao-ph↗

arXiv2026

Merlin: A Computed Tomography Vision-Language Foundation Model and Dataset

The large volume of abdominal computed tomography (CT) scans coupled with the shortage of radiologists have intensified the need for automated medical image analysis tools. Previous state-of-the-art approaches for automated analysis leverage vision-language models (VLMs) that jointly model images and radiology reports. However, current medical VLMs are generally limited to 2D images and short reports. Here to overcome these shortcomings for abdominal CT interpretation, we introduce Merlin, a 3D VLM that learns from volumetric CT scans, electronic health record data and radiology reports. This approach is enabled by a multistage pretraining framework that does not require additional manual annotations. We trained Merlin using a high-quality clinical dataset of paired CT scans (>6 million images from 15,331 CT scans), diagnosis codes (>1.8 million codes) and radiology reports (>6 million tokens). We comprehensively evaluated Merlin on 6 task types and 752 individual tasks that covered diagnostic, prognostic and quality-related tasks. The non-adapted (off-the-shelf) tasks included zero-shot classification of findings (30 findings), phenotype classification (692 phenotypes) and zero-shot cross-modal retrieval (image-to-findings and image-to-impression). The model-adapted tasks included 5-year chronic disease prediction (6 diseases), radiology report generation and 3D semantic segmentation (20 organs). We validated Merlin at scale, with internal testing on 5,137 CT scans and external testing on 44,098 CT scans from 3 independent sites and 2 public datasets. The results demonstrated high generalization across institutions and anatomies. Merlin outperformed 2D VLMs, CT foundation models and off-the-shelf radiology models. We also release our trained models, code, and dataset, available at: https://github.com/StanfordMIMI/Merlin.

cs.CV↗

arXiv2025

Evaluating and Improving the Effectiveness of Synthetic Chest X-Rays for Medical Image Analysis

Purpose: To explore best-practice approaches for generating synthetic chest X-ray images and augmenting medical imaging datasets to optimize the performance of deep learning models in downstream tasks like classification and segmentation. Materials and Methods: We utilized a latent diffusion model to condition the generation of synthetic chest X-rays on text prompts and/or segmentation masks. We explored methods like using a proxy model and using radiologist feedback to improve the quality of synthetic data. These synthetic images were then generated from relevant disease information or geometrically transformed segmentation masks and added to ground truth training set images from the CheXpert, CANDID-PTX, SIIM, and RSNA Pneumonia datasets to measure improvements in classification and segmentation model performance on the test sets. F1 and Dice scores were used to evaluate classification and segmentation respectively. One-tailed t-tests with Bonferroni correction assessed the statistical significance of performance improvements with synthetic data. Results: Across all experiments, the synthetic data we generated resulted in a maximum mean classification F1 score improvement of 0.150453 (CI: 0.099108-0.201798; P=0.0031) compared to using only real data. For segmentation, the maximum Dice score improvement was 0.14575 (CI: 0.108267-0.183233; P=0.0064). Conclusion: Best practices for generating synthetic chest X-ray images for downstream tasks include conditioning on single-disease labels or geometrically transformed segmentation masks, as well as potentially using proxy modeling for fine-tuning such generations.

eess.IV↗

arXiv2025

CheXalign: Preference fine-tuning in chest X-ray interpretation models without human feedback

Radiologists play a crucial role in translating medical images into actionable reports. However, the field faces staffing shortages and increasing workloads. While automated approaches using vision-language models (VLMs) show promise as assistants, they require exceptionally high accuracy. Most current VLMs in radiology rely solely on supervised fine-tuning. Meanwhile, additional preference fine-tuning in the post-training pipeline has become standard practice in the general domain. The challenge in radiology lies in the prohibitive cost of obtaining radiologist feedback at scale. To address this challenge, we propose an automated pipeline for preference feedback, focusing on chest X-ray radiology report generation (RRG). Specifically, our method leverages publicly available datasets containing pairs of images and radiologist-written reference reports with reference-based metrics, or Judges, eliminating the need for additional radiologist feedback. We investigate reward overoptimization via length exploitation in this setting and introduce a length-controlled version of the GREEN score. Our best-performing setup achieves state-of-the-art CheXbert scores on the MIMIC-CXR dataset for the RRG task while on average maintaining robust performance across six additional image perception and reasoning tasks.

cs.CV↗

arXiv2025

Automated Structured Radiology Report Generation

Automated radiology report generation from chest X-ray (CXR) images has the potential to improve clinical efficiency and reduce radiologists' workload. However, most datasets, including the publicly available MIMIC-CXR and CheXpert Plus, consist entirely of free-form reports, which are inherently variable and unstructured. This variability poses challenges for both generation and evaluation: existing models struggle to produce consistent, clinically meaningful reports, and standard evaluation metrics fail to capture the nuances of radiological interpretation. To address this, we introduce Structured Radiology Report Generation (SRRG), a new task that reformulates free-text radiology reports into a standardized format, ensuring clarity, consistency, and structured clinical reporting. We create a novel dataset by restructuring reports using large language models (LLMs) following strict structured reporting desiderata. Additionally, we introduce SRR-BERT, a fine-grained disease classification model trained on 55 labels, enabling more precise and clinically informed evaluation of structured reports. To assess report quality, we propose F1-SRR-BERT, a metric that leverages SRR-BERT's hierarchical disease taxonomy to bridge the gap between free-text variability and structured clinical reporting. We validate our dataset through a reader study conducted by five board-certified radiologists and extensive benchmarking experiments.

cs.CL↗

arXiv2025

Survival Games: Human-LLM Strategic Showdowns under Severe Resource Scarcity

The rapid advancement of large language models (LLMs) raises critical concerns about their ethical alignment, particularly in scenarios where human and AI co-exist under the conflict of interest. This work introduces an extendable, asymmetric, multi-agent simulation-based benchmarking framework to evaluate the moral behavior of LLMs in a novel human-AI co-existence setting featuring consistent living and critical resource management. Building on previous generative agent environments, we incorporate a life-sustaining system, where agents must compete or cooperate for food resources to survive, often leading to ethically charged decisions such as deception, theft, or social influence. We evaluated two types of LLM, DeepSeek and OpenAI series, in a three-agent setup (two humans, one LLM-powered robot), using adapted behavioral detection from the MACHIAVELLI framework and a custom survival-based ethics metric. Our findings reveal stark behavioral differences: DeepSeek frequently engages in resource hoarding, while OpenAI exhibits restraint, highlighting the influence of model design on ethical outcomes. Additionally, we demonstrate that prompt engineering can significantly steer LLM behavior, with jailbreaking prompts significantly enhancing unethical actions, even for highly restricted OpenAI models and cooperative prompts show a marked reduction in unethical actions. Our framework provides a reproducible testbed for quantifying LLM ethics in high-stakes scenarios, offering insights into their suitability for real-world human-AI interactions.

cs.HC↗

arXiv2025

Foundation Models in Radiology: What, How, When, Why and Why Not

Recent advances in artificial intelligence have witnessed the emergence of large-scale deep learning models capable of interpreting and generating both textual and imaging data. Such models, typically referred to as foundation models, are trained on extensive corpora of unlabeled data and demonstrate high performance across various tasks. Foundation models have recently received extensive attention from academic, industry, and regulatory bodies. Given the potentially transformative impact that foundation models can have on the field of radiology, this review aims to establish a standardized terminology concerning foundation models, with a specific focus on the requirements of training data, model training paradigms, model capabilities, and evaluation strategies. We further outline potential pathways to facilitate the training of radiology-specific foundation models, with a critical emphasis on elucidating both the benefits and challenges associated with such models. Overall, we envision that this review can unify technical advances and clinical needs in the training of foundation models for radiology in a safe and responsible manner, for ultimately benefiting patients, providers, and radiologists.

cs.LG↗