arXiv subjects

Di Zhang

Publications and source records attributed to Di Zhang.

At least 55 records · Page 3Linked to original sources

arXiv2026

AGE-MIL: Anchor-Guided Evidence Learning for Patient-Level Prediction

Existing computational pathology methods predominantly operate within whole-slide image (WSI)-level multiple instance learning (MIL) paradigms, while patient-level modeling remains underexplored. In routine pathological practice, however, pathologists derive diagnostic and prognostic conclusions by integrating evidence across multiple WSIs rather than relying on any single slide. This discrepancy creates a fundamental misalignment when patient-level supervision is directly imposed on conventional MIL frameworks, often leading to unstable optimization and degraded predictive reliability. To address this issue, we propose Anchor-Guided Evidence MIL (AGE-MIL), a weakly supervised framework for patient-level prediction. AGE-MIL constructs a patient-level anchor from slide representations to capture global pathological context and guide the retrieval and integration of diagnostically relevant local patches, enabling robust patient-level modeling. Patient-level risk is further modeled as an evidence accumulation process, promoting stable optimization under weak supervision. AGE-MIL is evaluated on six clinically relevant patient-level prediction tasks from two independent cohorts. Experimental results show that the proposed framework consistently outperforms eight state-of-the-art MIL methods. Code is available at https://github.com/wodeniua/AGE-MIL.

cs.CV↗

arXiv2026

Potentially Optimal Joint Actions Recognition for Cooperative Multi-Agent Reinforcement Learning

Value function factorization is widely used in cooperative multi-agent reinforcement learning (MARL). Existing approaches often impose monotonicity constraints between the joint action value and individual action values to enable decentralized execution. However, such constraints limit the expressiveness of value factorization, restricting the range of joint action values that can be represented and hindering the learning of optimal policies. To address this, we propose Potentially Optimal Joint Actions Weighting (POW), a method that ensures optimal policy recovery where existing approximate weighting strategies may fail. POW iteratively identifies potentially optimal joint actions and assigns them higher training weights through a theoretically grounded iterative weighted training process. We prove that this mechanism guarantees recovery of the true optimal policy, overcoming the limitations of prior heuristic weighting strategies. POW is architecture-agnostic and can be seamlessly integrated into existing value factorization algorithms. Extensive experiments on matrix games, difficulty-enhanced predator-prey tasks, SMAC, SMACv2, and a highway-env intersection scenario show that POW substantially improves stability and consistently surpasses state-of-the-art value-based MARL methods.

cs.LG↗

arXiv2026

On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters

Parameter-efficient fine-tuning (PEFT) is usually treated as a cheaper alternative to full fine-tuning. We study a broader role: small trainable adapters as persistent local state on top of strong shared foundation models. In this framing, the base model provides shared competence while adapters carry instance-specific behavior such as preferences, skills, tool habits, and memory-like updates. We organize the problem around three scaling axes: Scale Up, where stronger shared priors make small local updates more useful; Scale Down, where we study how small adapters can be while remaining reliable; and Scale Out, where many persistent adapted instances coexist. MinT provides one infrastructure example for managing adapter identity, revision, provenance, evaluation, and serving residency. Together, the results suggest that PEFT can be a compact substrate for persistent personal models rather than only a budget substitute for full fine-tuning.

cs.LG↗

arXiv2026

Tailoring pure valley-Zeeman spin-orbit coupling in WSe$_2$-encapsulated monolayer graphene

Engineering proximity effects in twisted van der Waals heterostructures offers a powerful platform for designing electronic properties. While theoretical predictions of quantum interference in transition metal dichalcogenide-encapsulated graphene can selectively control the spin-orbit coupling component, experimental realizations have remained elusive. Here, we report pure valley-Zeeman spin-orbit coupling in monolayer graphene, achieved by encapsulation between two parallel twisted WSe$_2$ monolayers. We observed a symmetry-enforced reordering of Landau levels, which is driven by the competition between the fixed valley-Zeeman energy and the magnetic-field-dependent cyclotron energy. This reordering is characterized by a transition from symmetry-broken states in the quantum Hall effect to a restored fourfold degeneracy with integer or half-integer quantum Hall sequences. We also demonstrate the ability to completely quench the proximity spin-orbit coupling by tuning the encapsulated geometry.

cond-mat.mes-hall↗

arXiv2026

MinT: Managed Infrastructure for Training and Serving Millions of LLMs

We present MindLab Toolkit (MinT), a managed infrastructure system for Low-Rank Adaptation (LoRA) post-training and online serving. MinT targets a setting where many trained policies are produced over a small number of expensive base-model deployments. Instead of materializing each policy as a merged full checkpoint, MinT keeps the base model resident and moves exported LoRA adapter revisions through rollout, update, export, evaluation, serving, and rollback, hiding distributed training, serving, scheduling, and data movement behind a service interface. MinT scales this path along three axes. Scale Up extends LoRA RL to frontier-scale dense and MoE architectures, including MLA and DSA attention paths, with training and serving validated beyond 1T total parameters. Scale Down moves only the exported LoRA adapter, which can be under 1% of base-model size in rank-1 settings; adapter-only handoff reduces the measured step by 18.3x on a 4B dense model and 2.85x on a 30B MoE, while concurrent multi-policy GRPO shortens wall time by 1.77x and 1.45x without raising peak memory. Scale Out separates durable policy addressability from CPU/GPU working sets: a tensor-parallel deployment supports 10^6-scale addressable catalogs (measured single-engine sweeps through 100K) and thousand-adapter active waves at cluster scale, with cold loading treated as scheduled service work and packed MoE LoRA tensors improving live engine loading by 8.5-8.7x. MinT thus manages million-scale LoRA policy catalogs while training and serving selected adapter revisions over shared 1T-class base models.

cs.LG↗

arXiv2026

PathNavigate: A Training-Free Pathology Agent with Surprise-Guided Scan and Shared Slide Memory for Whole-Slide Image VQA

Whole-slide image visual question answering (WSI-VQA) frames pathology as an extreme-context search problem: to answer a free-form clinical query, a system must first navigate a gigapixel slide under a strict inspection budget to locate sparse, high-resolution evidence. Existing approaches largely fall into two paradigms: i) supervised pathology multimodal large language models (MLLMs) and agents can absorb localization and reasoning into learned modules, but they often couple navigation to task-specific supervision and retraining, limiting their practicality; ii) training-free pathology agents avoid this cost by keeping core models frozen, but often follow a question-first design, constructing the initial candidate set mainly from query-conditioned relevance. This can miss decisive morphology that is not named in the question, and force heavier inference-time scaffolding. To address this challenge, we introduce PathNavigate, a training-free pathology agent built around a scan-search-readout routine. Before question matching, PathNavigate scans the current slide at low magnification with a shared online memory module over frozen pathology features, producing a slide-specific surprise field that marks an abnormal-region pool. It then applies question-conditioned PLIP relevance only within this pool to select high-magnification search targets. Finally, it extracts local high-magnification evidence and answers with a frozen perceptor-adjudicator stack, using the same online memory as slide-level context. Experiments on WSI-VQA and SlideBench-BCNB show that the proposed scan-search-readout design improves answer accuracy and yields more interpretable evidence-selection trajectories with higher efficiency.The code is available online.

cs.CV↗

arXiv2026

Building a physics-aware AI ecosystem for solid-state hydrogen storage materials

Hydrogen storage remains a central bottleneck for scalable hydrogen energy systems due to the multiscale and coupled nature of the thermodynamics, kinetics, and microstructural evolution of hydrogen storage materials (HSMs). Although artificial intelligence (AI) has accelerated materials discovery, current approaches remain constrained by fragmented data, limited physical consistency, and weak integration with experimental validation. Here, we propose a unified framework that integrates coherent data infrastructure, physics-grounded modeling, and AI-driven inverse design within a closed-loop discovery paradigm. By embedding physical constraints and experimental feedback, this approach enables adaptive, physically consistent optimization, thereby establishing a pathway toward autonomous, digital-twin-enabled discovery of HSMs.

cond-mat.mtrl-sci↗

arXiv2026

Learning Native Continuation for Action Chunking Flow Policies

Action chunking enables Vision Language Action (VLA) models to run in real time, but naive chunked execution often exhibits discontinuities at chunk boundaries. Real-Time Chunking (RTC) alleviates this issue but is external to the policy, leading to spurious multimodal switching and trajectories that are not intrinsically smooth. We propose Legato, a training-time continuation method for action-chunked flow-based VLA policies. Specifically, Legato initializes denoising from a schedule-shaped mixture of known actions and noise, exposing the model to partial action information. Moreover, Legato reshapes the learned flow dynamics to ensure that the denoising process remains consistent between training and inference under per-step guidance. Legato further uses randomized schedule condition during training to support varying inference delays and achieve controllable smoothness. Empirically, Legato produces smoother trajectories and reduces spurious multimodal switching during execution, leading to less hesitation and shorter task completion time. Extensive real-world experiments show that Legato consistently outperforms RTC across five manipulation tasks, achieving approximately 10% improvements in both trajectory smoothness and task completion time.

cs.RO↗

arXiv2026

Is Class Signal Clustered or Routed in Task-Induced Implicit Neural Representation Weight Spaces?

Implicit neural representations (INRs) encode images as neural-network weights, making image classification a problem of weight-space classifiability. A natural geometric hypothesis is that classifier feedback should make image-specific weights cluster by class in the shared-anchor coordinate. We test this hypothesis in the SIREN-based Meta Weight Transformer (MWT) regime, where end-to-end training meta-learns a shared initialization and inner-loop update schedule for fitting image-specific SIRENs. We find that this prediction fails. Exposed weight-space geometry and supervised clustering pressure do not reliably track trained-reader accuracy; clustering can even make local neighborhoods more class-consistent while making the trained reader worse. Crucially, the reader constructs rather than inherits class-aligned geometry: token-flow diagnostics show that class-aligned neighborhoods become strongly predictive of trained-reader accuracy only after late reader interactions, not in the input coordinate. We further identify the native SIREN bias column in the augmented weight token as a low-dimensional, sample-dependent causal readout route for the trained reader; targeted controls rule out generic scalar-column and marginal-distribution artifacts. The diagnosis motivates interventions that strengthen reader routing, add an explicit bias route, or use denser inner-loop fitting; under the lane-specific training conventions used here, route-directed variants often outperform the shared-anchor baseline but interact non-additively. Task-induced INR weights are classifiable not because they form raw geometric clusters, but because their class signal is routed through the reader.

cs.CV↗

arXiv2026

Risk-aware stochastic scheduling of multi-market energy storage systems

Energy storage promotes the integration of renewables by operating with charge and discharge policies that balance an intermittent power supply. A key challenge in this emerging sector is how to optimize the operation of storage assets given future price uncertainties and the need to recover the costs of project finance while ensuring an attractive return on equity and hedging against downside risk. This study investigates the scheduling of energy storage assets under price uncertainty, with a focus on electricity markets. A two-stage stochastic risk-constrained approach is employed, whereby electricity price trajectories or specific power markets are observed, allowing for recourse in the schedule. Conditional value-at-risk is used to quantify risk in the optimization problems; this allows for explicit specification of a probabilistic risk limit. The proposed approach is tested in an integrated hydrogen system (IHS) and a battery energy storage system (BESS). In the joint design and operation context for the IHS, the risk constraint results in large installed unit capacities, increasing capital cost but enabling more inventory to buffer price uncertainty. In both case studies, there is an operational trade-off between risk and expected reward; this is reflected in higher expected costs (or lower expected profits) with increasing risk aversion. Despite the decrease in expected reward (up to 500\$k), both systems exhibit substantial benefits of increasing risk aversion (up to 1.5\$mn) with respect to risk-neutral settings. This work provides a general method to address uncertainties in energy storage scheduling, allowing operators to input their level of risk tolerance on asset decisions.

math.OC↗

arXiv2026

Steering Visual Generation in Unified Multimodal Models with Understanding Supervision

Unified multimodal models are envisioned to bridge the gap between understanding and generation. Yet, to achieve competitive performance, state-of-the-art models adopt largely decoupled understanding and generation components. This design, while effective for individual tasks, weakens the connection required for mutual enhancement, leaving the potential synergy empirically uncertain. We propose to explicitly restore this synergy by introducing Understanding-Oriented Post-Training (UNO), a lightweight framework that treats understanding not only as a distinct task, but also a direct supervisory signal to steer generative representations. By incorporating objectives that encode semantic abstraction (captioning) and structural details (visual regression), we enable effective gradient flow from understanding to generation. Extensive experiments on image generation and editing demonstrate that understanding can serve as an effective catalyst for generation.

cs.CV↗

arXiv2026

MolReFlect: Towards In-Context Fine-grained Alignments between Molecules and Texts

Molecule discovery is a pivotal research field, impacting everything from medicine to materials. Recently, Large Language Models (LLMs) have been widely adopted in molecular understanding and generation, serving as a bridge between the molecular space and the natural language space, yet the alignment between molecules and their corresponding captions remains a significant challenge. Previous endeavors typically treat molecules as monolithic inputs, lacking an intermediate reasoning process and sacrificing explainability. In this work, we define fine-grained alignments as the precise correspondence between a molecule's sub-structures and the textual phrases that explain their properties. These alignments are crucial for LLMs to understand molecules in a more accurate and explainable manner. Normally, such fine-grained alignments require expert annotation, which is both costly and time-consuming. To allow LLMs to automatically label and learn the fine-grained alignments, we propose MolReFlect, a novel teacher-student framework, where a teacher LLM first generates and refines mappings between caption phrases and SMILES substructures and then explicitly teaches these detailed alignments to a student LLM. Experimental results demonstrate that MolReFlect enables LLMs to significantly outperform previous baselines, achieving the state-of-the-art performance in the molecule-caption translation task. Our codes are available via: https://github.com/phenixace/MolReFlect.

cs.CL↗

arXiv2026

The Structural Attention Tax: How Retrieval Format Hijacks In-Context Learning Independent of Content

Retrieval-augmented generation (RAG) systems inject external knowledge to improve LLM outputs, yet the format of injected content -- distinct from its semantic relevance -- can independently distort the model's attention distribution. We identify and formalise a phenomenon we term the structural attention tax: knowledge graph (KG) triples, due to their relational delimiters and repeated slot patterns, capture 2-3x more attention per token than semantically equivalent natural-language text ($\hat{o}$(KG) $\approx$ 0.70 vs. $\hat{o}$(neutral) $\approx$ 0.25), compressing demonstration attention by up to 42% -- regardless of whether the triples are relevant or noise. We develop a formal framework decomposing attention scores into semantic and structural components (Eq. 2), derive a compression bound (Proposition 1) connecting token-level format bias to demonstration attention loss, and show that the structural term governs how much attention is diverted while the semantic term governs whether this helps or hurts. This decoupling reveals two orthogonal axes for improving retrieval-augmented ICL: optimising retrieval quality (semantic axis) and reducing format-driven attention capture (structural axis). Empirically, across two model families (Mistral-7B, LLaMA-3-8B) and three QA benchmarks, we observe that source-task alignment dominates: task-matched BM25 retrieval achieves 58-62% on HotpotQA vs. ConceptNet's 25-27%, a >30 pp gap that dwarfs all gating strategies ($\leq$2 pp). We derive five structure-aware mitigation strategies from the framework, ranging from zero-cost prompt modifications to training-time regularisation; format flattening (S3) is validated by both accuracy and attention-level evidence from a verbalized-triple control, while structural dispersal (S1) yields mixed results that illuminate the challenges of format-level intervention.

cs.CL↗

arXiv2026

Spin State versus Potential of Zero Charge as Predictors of Density-Dependent Oxygen Reduction in M-N-C Electrocatalysts

Metal-site density strongly influences oxygen reduction activity and selectivity in M-N-C electrocatalysts, but the descriptors that predict these trends remain under debate. Here, we compare spin state and the potential of zero charge as predictors of density-dependent oxygen reduction behavior in Fe-N-C and Co-N-C catalysts. Using constrained-magnetization calculations combined with Landau analysis, we find that the ground-state magnetic moments vary only weakly across a broad range of metal-site densities, suggesting that magnetic descriptors alone cannot account for the pronounced performance changes. In contrast, explicit-solvent simulations reveal systematic density-dependent shifts in PZC, which alter the interfacial electric field and thereby modulate field-sensitive adsorption energetics of ORR intermediates. Incorporating these PZC shifts into a pH-field-coupled microkinetic model captures the density-dependent activity trends and reproduces the experimentally observed increase in two-electron selectivity at lower site densities under acidic conditions. Experimental PZC measurements further support the predicted trend. Together, these results show that PZC is a more effective predictor than spin state for density-dependent oxygen reduction activity and selectivity in M-N-C electrocatalysts.

cond-mat.mtrl-sci↗

arXiv2026

SelfBudgeter: Adaptive Token Allocation for Efficient LLM Reasoning

Recently, large reasoning models demonstrate exceptional performance on various tasks. However, reasoning models always consume excessive tokens even for simple queries, leading to resource waste and prolonged user latency. To address this challenge, we propose SelfBudgeter - a self-adaptive reasoning strategy for efficient and controllable reasoning. Specifically, we first train the model to self-estimate the required reasoning budget based on the query. We then introduce budget-guided GRPO for reinforcement learning, which effectively maintains accuracy while reducing output length. Experimental results demonstrate that SelfBudgeter dynamically allocates budgets according to problem complexity, achieving an average response length compression of 61% on math reasoning tasks while maintaining accuracy. Furthermore, SelfBudgeter allows users to see how long generation will take and decide whether to continue or stop. Additionally, users can directly control the reasoning length by setting token budgets upfront.

cs.AI↗

arXiv2026

Multi-Task Learning for High-Dimensional Regression with Many Weak Instruments

Many weak instrumental variables (IVs) are routinely used in the health and social sciences to improve identification and inference of the treatment effect of interest, along with a broad collection of data on potential confounding factors in the hope that the IV assumptions hold within each data stratum. We propose a new debiased continuous-updating generalized method of moments estimator with multi-task learning of the IV propensity scores to simultaneously address the biases from a diverging number of weak IVs as well as first-step regularized estimation of nuisance regression functions in high-dimensional potential confounding factors. We develop a new multi-task learning theory for generalized linear models under a general sub-Gaussian design to establish valid inference in the many weak IVs asymptotic regime under appropriate sparsity conditions. We evaluate the proposed method via extensive Monte Carlo studies and an empirical application to investigate the returns to education.

stat.ME↗

arXiv2026

A unified descriptor framework for hydrogen storage capacity and equilibrium pressure in interstitial hydrides

Hydrogen is a promising energy carrier, yet its practical deployment is limited by the lack of storage materials that simultaneously achieve high storage capacity ($w$) and practical equilibrium pressure at room temperature ($P_{\rm eq,RT}$). Interstitial metal hydrides offer fast kinetics and favorable thermodynamics (high $P_{\rm eq,RT}$) but suffer from intrinsically low w. Here, we establish a physically interpretable, data-driven framework to uncover descriptor-property relationships in interstitial hydrides using a curated database of pressure-composition-temperature measurements (Digital Hydrogen Platform, DigHyd) and white-box symbolic regression. Strikingly, the analysis reveals a clear separation of governing mechanisms, in which $w$ is governed by geometric and lattice conditions, captured by the average atomic radius ($\left\langle r_M \right\rangle$) and average thermal conductivity ($\left\langleκ\right\rangle$), with an optimal regime of $r_M \sim 1.47 Å$ and relatively low $\left\langleκ\right\rangle$. In contrast, $P_{\rm eq,RT}$ is governed by elastic properties, captured by the average shear modulus ($\left\langle G \right\rangle$) and average Poisson's ratio ($\left\langle ν\right\rangle$), reflecting the role of lattice rigidity and mechanical compliance. These relationships are translated into compositional optimization pathways that follow the descriptor trends above, enabling the design of candidate materials with enhanced w under practical equilibrium conditions ($P_{\rm eq,RT} \sim 0.1$ MPa). This work establishes a general, interpretable strategy for physics-informed design of energy materials systems.

cond-mat.mtrl-sci↗

arXiv2026

PhysInOne: Visual Physics Learning and Reasoning in One Suite

We present PhysInOne, a large-scale synthetic dataset addressing the critical scarcity of physically-grounded training data for AI systems. Unlike existing datasets limited to merely hundreds or thousands of examples, PhysInOne provides 2 million videos across 153,810 dynamic 3D scenes, covering 71 basic physical phenomena in mechanics, optics, fluid dynamics, and magnetism. Distinct from previous works, our scenes feature multiobject interactions against complex backgrounds, with comprehensive ground-truth annotations including 3D geometry, semantics, dynamic motion, physical properties, and text descriptions. We demonstrate PhysInOne's efficacy across four emerging applications: physics-aware video generation, long-/short-term future frame prediction, physical property estimation, and motion transfer. Experiments show that fine-tuning foundation models on PhysInOne significantly enhances physical plausibility, while also exposing critical gaps in modeling complex physical dynamics and estimating intrinsic properties. As the largest dataset of its kind, orders of magnitude beyond prior works, PhysInOne establishes a new benchmark for advancing physics-grounded world models in generation, simulation, and embodied AI.

cs.CV↗