Searcharxiv⌕ Search

arXiv subjects

Yi Zhang

Publications and source records attributed to Yi Zhang.

At least 37 records · Page 2Linked to original sources
arXiv2026

VLA-Trace: Diagnosing Vision-Language-Action Models through Representation and Behavior Tracing

Understanding how Vision-Language-Action (VLA) models transform multimodal knowledge into embodied control remains an open challenge. We present VLA-Trace, a progressive diagnostic framework that analyzes VLA models through a unified evidence chain from representation dynamics to causal control attribution and behavioral manifestation. It specifically combines cross-modal and checkpoint-drift centered kernel alignment (CKA) to trace representation evolution, attention knockout interventions to identify modality-specific control pathways, and rollout-level behavioral probes to examine grounding, shortcut dependence, and semantic following. Experiments on $π_{0.5}$ and OpenVLA reveal three key findings. First, the two models exhibit distinct modality-specific adaptation dynamics during VLA finetuning. Second, they rely on different multimodal routing strategies and layer-wise dependencies during action decoding. Third, although VLA policies excel at visually grounded trajectory generation, they remain limited in fine-grained semantic following. These findings highlight future directions for representation-preserving adaptation, causal VLA circuits, and compositional semantic control.

cs.AI↗
arXiv2026

MemPoison: Uncovering Persistent Memory Threats and Structural Blind Spots in LLM Agents

Persistent external memory enhances agent continuity but introduces persistent security vulnerabilities: adversarial content can be injected via standard interaction channels, retained across turns, and later distort downstream behavior. To address this challenge, we propose MemPoison, a comprehensive benchmark and analysis framework featuring 1227 hand-validated cases across four attack types, three injection channels, and three representative memory substrates, evaluated on seven open-weight and three closed-weight model families. We introduce a three-tier taxonomy: (L1) direct single-record corruption, (L2) compositional multi-record corruption and (L3) context-triggered dormant corruption. Our evaluations reveal a distinct defense frontier: while baseline write-time defenses, such as consistency checks, substantially suppress direct L1 attacks, they fail to reliably suppress L2 and L3 attacks. Through mechanistic influence decomposition (MID), we demonstrate structural blind spots in write-time defenses, which admit seemingly benign records that later become harmful through joint retrieval composition or trigger-conditioned activation. Our findings advocate for shifting from static filtering to adaptive, context-sensitive memory defense strategies.

cs.CR↗
arXiv2026

Rota-Baxter operators and factorization of Lie affgebras

We introduce Rota-Baxter operators on Lie affgebras by an identity independent of the choice of origin and prove that they induce descendant Lie affgebra structures. We establish a bijection between ordered affine Lie factorizations and idempotent operators split at a chosen idempotent fixed point. A cokernel obstruction determines the existence of this point, while invariance under the structural endomorphisms characterizes splitting. We also determine when weighted descendants satisfy the affine Jacobi identity and when operators of nonzero weight give affine doubles. For canonical Lie affgebras, we describe all affine Rota-Baxter operators by Lie subalgebra decompositions and translation vectors, and classify them up to automorphism in dimensions at most two.

math.RA↗
arXiv2026

Peer Pressure: Execution-Verified Co-Evolution of Attacker and Defender Populations for Code Security

LLM code agents increasingly ship code with little oversight, yet defenders trained on static vulnerability datasets fall behind adapting adversaries. We co-train attacker and defender instead, and let execution decide every reward: an attacker (Red) edits a real repository and earns nothing unless its patch applies, builds, passes the tests, completes its development task and reproduces a real exploit in an isolated container, whose path also labels the entry point, the critical operation and the trace between them. Peer-Relative Policy Optimization (PRPO) makes both sides populations of LoRA adapters over one frozen base and credits each against same-role peers that faced an identical condition, placing the population inside the policy-gradient baseline rather than the opponent pool, with a tournament preserving diversity. Matching cancels the condition's difficulty: at equal coverage it is worth 2.3 points of balanced accuracy over a baseline computed on other conditions, and it matches a within-policy control at half the verifier spend. PRPO's 7B diff reviewer reaches 64.8% balanced accuracy on four unseen applications in three untrained languages, 5.5 points above supervised security fine-tuning and 4.4 above a frozen frontier reviewer (Claude Opus 4.8), which it leads by 15.8 on the application it trains on. Its 9B repository agent recovers 23 of 172 held-out VulnGym advisories, against 10 after supervised fine-tuning and 7 for the same backbone untrained under the same scaffold, and locates 17 of 43 reachable entry points against 13 for mini-swe-agent with GLM-5.2. Run unchanged on jailbreak defense, the same algorithm matches or exceeds WildGuard's static F1 at equal corpus and model size. Under a judge distinct from its training oracle it cuts attack success from 36.0% to 4.6%, against WildGuard's 8.9%.

cs.CR↗
arXiv2026

Interplay between high-entropy effects and the superionic transition in fluorite oxides: A liquid-like perspective

Minor actinide (MA)-bearing mixed oxide (MOX), a prominent high-entropy ceramic (HEC), is a promising candidate for advanced fast reactor fuels. Using a hybrid Monte Carlo/Molecular Dynamics (MC/MD) approach, we reveal that Np and Am segregation establishes a distinct short-range order (SRO) that induces severe lattice distortion. We uncover a competitive mechanism between localized high-entropy blocking effects and collective superionic dynamics, where the anionic kinetic and structural evolution within the 2000-2800 K range is partitioned and governed by the Tammann crossover temperature and superionic transition temperature. In the low-temperature regime, severe lattice distortions induced by the SRO structures create localized electrostatic and strain-coupled potential wells, which spatially confine the O-ions and lead to their sluggish diffusion. The system exhibits increased dynamical heterogeneity as thermal activation alleviates these structural constraints. Crucially, these localized traps spontaneously transform into a highly interconnected, string-like polyhedral topological network. Driven by the maximization of configurational entropy, this structural change creates low-viscosity transport pathways that trigger confinement-enhanced diffusion. Consequently, a dramatic kinetic inversion occurs, where the O-ion mobility of SRO model surpasses that of random solid solution (RSS) model via an entropy-driven compensation mechanism. Finally, through Gibbs free energy minimization and defect analysis, structural and thermodynamic origins of the superionic transition in HEC are elucidated. This study unveils how local chemical ordering regulates ion transport under extreme conditions, providing theoretical support for the design of high-performance nuclear fuels with both superior radiation tolerance and optimized mass transfer capabilities.

cond-mat.mtrl-sci↗
arXiv2026

Rethinking GRPO for Large Reasoning Model Safety Alignment

Large reasoning models (LRMs) exhibit strong reasoning capabilities but remain vulnerable to unsafe outputs. While Group Relative Policy Optimization (GRPO) is widely used for safety alignment, its group-relative optimization may provide insufficient safety-directed learning signals. In this paper, we identify two failure modes: advantage collapse, where homogeneous rollout groups yield vanishing advantages, and unsafe diversification, where unsafe behaviors persist and become increasingly diverse. Our theoretical analysis reveals that behavioral consistency weakens within-group safety contrasts, while entropy regularization can diversify unsafe responses when relative signals vanish. To address these challenges, we propose Beacon, comprising Safety Signal Recovery (SSR) and Safety-Utility Decoupled Alignment (SUDA). SSR combines advantage-guided trajectory filtering and safe reference supervision to recover effective safety learning signals, while SUDA employs category-aware rewards and benign-only KL anchoring to mitigate over-refusal without compromising safety. Experiments across four Qwen3 and DeepSeek-R1-Distill variants and eight benchmarks show that our framework achieves the best average safety and reasoning performance, reducing average safety error by up to 32.83 percentage points while maintaining low over-refusal.

cs.AI↗
arXiv2026

Computing Multi-Slot First-Price Pacing Equilibria at Scale via Entropic Smoothing

Pacing equilibria describe how budget-constrained advertisers clear an ad market, but computing one at production scale is hard when every request fills several slots: the allocation jumps whenever two campaigns' scores cross. We consider the first-price pacing equilibrium (FPPE) of auctions with $K$ homogeneous slots and unit-demand campaigns, whose Eisenberg-Gale dual reduces, via a TopK identity, to a convex program over bids. Regularizing each request's allocation with the binary (Fermi-Dirac) entropy gives a smooth, strictly convex dual whose gradient is exactly spend minus budget and whose evaluation splits into independent one-dimensional threshold solves, one per request. The smoothed equilibrium converges to the hard one at an explicit $O(\sqrt{T})$ rate, and a duality gap certifies how far any bids are from the hard equilibrium. We implement an efficient multi-GPU solver and test it on a production ad workload. We use it to compute an equilibrium price anchor: each campaign's hindsight equilibrium cost per key result.

math.OC↗
arXiv2026

EPIC-Bench: A Perception-Centric Benchmark for Fine-Grained Embodied Visual Grounding in Vision-Language Models

While large vision-language models (VLMs) are increasingly adopted as the perceptual backbone for embodied agents, their fundamental vision-language grounding (VLG) capability in embodied scenarios remains unexplored. We present \textbf{EPIC-Bench}, \textbf{E}mbodied \textbf{P}ercept\textbf{I}on Ben\textbf{C}hmark, a fine-grained grounding benchmark designed to systematically evaluate the VLG capabilities of VLMs in real-world embodied environments. EPIC-Bench comprises 6.6k meticulously human-annotated tuples (Image, Text, Mask) spanning three core stages of the embodied interaction pipeline: Target Localization, Navigation, and Manipulation. With 23 fine-grained tasks, multi-dimensional settings and evaluation over 80 leading VLMs, EPIC-Bench provides a diagnostic analysis.

cs.CV↗
arXiv2026

Modules and operator rings of Reynolds algebras

We develop a module theory for weighted Reynolds algebras by constructing an associa?tive ring that records the algebra action and the module operator. The categories of left and right Reynolds modules are equivalent to the corresponding module categories over this ring, and their tensor products agree under these equivalences. We construct free Reynolds modules and give an explicit graded description of the operator ring for idempotent classical Reynolds operators fixing the identity. When the Reynolds operator on the algebra is bijective, modules with bijective module operators correspond to differential modules satisfying an invertibility condition. The operator ring transfers the Hom-tensor adjunction, Morita theory and derived functors to Reynolds modules. Explicit computations over a field show that the identity operator gives an operator ring of global dimension one, whereas the zero operator gives one of infinite global dimension.

math.RT↗
arXiv2026

Odd-to-Even-Layer Switch of Spontaneous Excitonic State in $\rm{TiSe_{2}}$ Thin Films

$\rm{TiSe_{2}}$ is a candidate of high temperature excitonic condensates with spontaneous exciton formation in addition to its charge density wave (CDW). However, direct observation and manipulation of this electron-hole paring state formation is still lacking. Here, using varying-temperature angle-resolved photoemission spectroscopy (VT-ARPES), we observe a distinct odd-to-even-layer switch of spontaneous exciton state formation with the evident correlated band structure in $\rm{TiSe_{2}}$ thin films. While the odd-layer $\rm{TiSe_{2}}$ films show a sharp band structure in single-particle picture superimposed by CDW renormalization, in the even-layer $\rm{TiSe_{2}}$ films, the full-filling valence band (VB) opens a distinct notch at the top, and the conductive band (CB) shows a significant extra signal extending above Fermi level with spectra broadening. These features indicate that the extra holes (notched VB) and electrons (broadening CB above Fermi level) form spontaneous excitons in even-layer $\rm{TiSe_{2}}$. Based on a correlated electron-hole model extracted from the tight-binding description, we reproduce the characterized band structures with spontaneous exciton formation, where the interlayer inter-band coupling in $\rm{TiSe_{2}}$ plays a key role, which leads to an odd-to-even-layer switch of the excitonic state formation. Our results provide direct evidences of the spontaneous exciton formation, promising an excitonic condensate as versatile platform for realizing Bose-Einstein condensate (BEC) and investigating novel correlated phenomena.

cond-mat.str-el↗
arXiv2026

Testing Conditional Distribution Equality Using Generative Models

Conditional distribution equality testing is a core problem with applications in causal inference, model diagnostics, and transfer learning. Existing methods often lose power when the conditioning variable is high-dimensional. We propose a new test that leverages generative models to compare conditional distributions using both observed data and synthetic samples. A sample-splitting and studentization scheme is introduced to control estimation error. Under the null, the test statistic is asymptotically normal, and we establish local power against contiguous alternatives. Our theory under the null and fixed alternative requires only consistency of the learned conditional distribution, allowing for a broad class of generative models. Empirical results demonstrate accurate size control and strong power in high-dimensional settings compared to representative kernel-based conditional and joint two-sample tests.

stat.ME↗
arXiv2026

Efficient Quantum Fully Homomorphic Encryption

Quantum fully homomorphic encryption (QFHE) enables arbitrary quantum computations on encrypted data, but prior constructions require prohibitive quantum resources--specifically, O(lambda^2) EPR pairs per T-gate evaluation using the Barrington-based approach (DSS16). This paper introduces a unified framework achieving exponential improvement over the generic Barrington-based approach in program length. The central innovation is a novel modular arithmetic program (MA-Program) tailored to learning with errors (LWE) decryption. We show that LWE decryption computes the inner product mod q, a modular inner product that is NOT a symmetric function. Thus, prior symmetric-function optimizations (Sinha's O(n)-state branching programs) do not apply. Our MA-Program tracks partial sums modulus q with state space Z_q requiring O(log q) bits, yielding programs of state count O(lambda) with binary encoding O(log lambda) and length O(lambda log lambda). This reduces the quantum gadget size from O(lambda^2) to O(lambda log^2 lambda) EPR pairs. To achieve a fully classical client, we transfer all quantum resources (EPR preparation, Bell measurements, adaptive error correction) to the server via the MA-Program gadget framework. Clients only perform classical LWE key generation, Pauli key encryption under classical FHE, and no quantum operations; a layered key structure further eliminates circular security assumptions. For parallel computation, we adopt the MBQC framework with flow functions, supporting up to O(log lambda) parallel measurements per layer. This separates offline resource preparation from online adaptive measurement, enabling parallel processing while maintaining deterministic evaluation.

quant-ph↗
arXiv2026

Monoidal structure on Rota-Baxter algebras via augmentation, and free cocycle bialgebras

A tensor product is known for commutative Rota--Baxter algebras through a universal construction, but its underlying algebra need not be the ordinary tensor product $A\otimes B$ and can be substantially larger. We show that, after passing to augmented Rota--Baxter algebras, the ordinary tensor product itself carries a natural Rota--Baxter operator which makes the category monoidal with the usual associator and unit constraints. The resulting structure is intrinsically nonsymmetric: over a field of characteristic zero, this operator and its flip-conjugated counterpart are the only possibilities among linear combinations of five natural terms, while over any field the full category admits no braiding. Its comonoid objects are precisely Rota--Baxter cocycle bialgebras. Moreover, the free Rota--Baxter adjunction yields a bimonad and free cocycle bialgebras on ordinary bialgebras, with compatible Hopf structures for connected filtered inputs, in both the associative and commutative settings.

math.CT↗
arXiv2026

Code2Worlds-2: Physically Plausible Agent Harness for Embodied World Simulation

Recent coding large language models (LLMs) can synthesize dynamic 4D worlds from natural language, but the resulting content remains a visual artifact. It lacks physical semantics such as the collision geometry, mass, friction, inertia, and joint structure, required to enable it to simulate the physical world, and its dynamics are scripted to look plausible rather than designed to execute. What is needed is an executable world: one that a physics engine can load and step, resolving contact physically. We present Code2Worlds-2, an extension of Code2Worlds that turns 4D world generation into an embodied world simulation. We keep the original dual-stream generation pipeline intact and append two modules. First, a SimReady Scene Compiler that turns a generated visual scene into a simulator-compatible representation: it assigns collision meshes via convex decomposition, estimates physical attributes with a vision-language model, infers a lightweight support graph, instantiates the engine's contact, friction, and dynamics constraints, and applies a gravity settling pass, before exporting the scene as MuJoCo MJCF. Second, a Simulator-in-the-Loop 4D Refiner that implements an automated agent harness-a compile, simulate, critique and refine loop-that converts visual dynamics into executable rigid and articulated dynamics and evolves the physical settings rather than only the code, replacing the purely visual motion critic with a simulator critic that inspects penetration depth, collision pairs, stability, contact forces, joint limits, and robot reachability. We further propose SimCode4D, a state-based benchmark with five task templates for evaluating embodied simulation. Experiments on SimCode4D and across natural, indoor, industrial, urban, and service scenes show that Code2Worlds-2 produces physically plausible, simulator-compatible, and robot-interactable scenes.

cs.RO↗
arXiv2026

Preliminary Exploration of Large Models Based on Self-Supervised Learning for Health Monitoring of Power Electronic Converters

Current artificial intelligence (AI) methods for power electronics converters make limited use of the abundant unlabeled data generated during real-world operation. This underexplored resource holds significant potential for advancing model performance. To address this gap, this paper presents PE-TSFM, a time-series foundation model tailored for power module health monitoring. The model adopts the PatchTST architecture combined with self-supervised learning and is trained on more than 143 million timestamps, where over 98\% of this training data is unlabeled and collected directly from an operating traction converter. In degradation classification tasks, PE-TSFM achieves an accuracy of 87\%, considerably surpassing five conventional time-series AI models whose performance ranges from 27\% to 45\%. These results demonstrate the model's strong few-shot learning capability and generalization across different operating conditions. This work promotes a shift from narrowly focused models in power electronics toward unified, scalable foundation models that can unlock value from large-scale unlabeled operational data.

eess.SY↗
arXiv2026

GDEPO: Group Dual-dynamic and Equal-right Advantage Policy Optimization for Sample-Efficient Reinforcement Learning

Automated Theorem Proving (ATP) is a fundamental AI challenge, requiring machine-verifiable proofs in formal languages like Lean to evaluate reasoning. Reinforcement Learning (RL), especially Group Relative Policy Optimization (GRPO), is a mainstream approach. However, GRPO faces two critical issues in ATP: its relative advantage estimation may conflict with binary verifier feedback under composite rewards, and its static sampling strategy discards batches without valid proofs, causing data waste. Repeated training on mastered problems also wastes computation. To address this, we propose Group Dual-dynamic and Equal-right Advantage Policy Optimization (GDEPO), incorporating three core mechanisms: 1) dynamic additional sampling, resampling ineffective batches until a valid proof is found; 2) equal-right advantage, decoupling the advantage function's sign (correctness) from its magnitude (auxiliary rewards) to ensure stable policy updates while increasing effective computation; and 3) dynamic additional iterations, applying extra gradient steps to initially failed but eventually successful samples to accelerate learning. Experiments on MinF2F-test, MathOlympiadBench, and PutnamBench confirm GDEPO's effectiveness, with ablation studies validating its synergistic components. The method enhances data utilization and optimization efficiency, offering a novel training paradigm for ATP.

cs.AI↗
arXiv2026

PWM: Personalized World Models with Online Reinforcement Learning

Pretrained world models can generate diverse environments, yet users often want to explore a particular scene specified by their own video. This requires learning the scene's visual identity while retaining the quality of action-conditioned generation. We introduce Personalized World Models (PWM), a framework for customizing interactive world models from short scene videos through online reinforcement learning. In PWM, the support trajectory and its associated controls provide reward feedback on continuations sampled from the current policy. In the GRPO instantiation, group-relative optimization updates a compact LoRA adapter using a unified reward for scene appearance, visual continuity, and motion, while base-policy anchoring regularizes changes to the pretrained generation prior of a frozen Yume-5B backbone. The same adaptation procedure is applied across real and rendered environments. We also instantiate PWM with DiffusionNFT as an alternative reward-guided optimization method for learning the scene-specific adapter. We also introduce PWM-Bench, comprising 150 customization tasks across Indoor, Outdoor, and Gaming, with paired evaluation on held-out continuations. The GRPO and DiffusionNFT instantiations of PWM improve customization over native Yume in 71.3% and 65.3% of the evaluated scenes, respectively, with positive mean gains across all three domains. For the GRPO instantiation, matched SFT comparisons further demonstrate higher mean customization gains and better mean image-quality scores in every domain, while retaining frame-level visual quality close to the pretrained model.

cs.CV↗
arXiv2026

A physics-coupled graph neural operator for mid-plane six-component stress tensor field prediction in concrete penetration

The penetration-resistant design of concrete structures requires the spatiotemporal evolution of the stress field within the target, whereas in design only the strength grade, the geometry and the projectile parameters are known, and the aggregate arrangement is not. Empirical formulae and scalar-output surrogates serve design screening, and mesoscale finite element models resolve the aggregate-scale response, but one solution cannot cover random aggregate samples at scale, and no framework links mesoscale heterogeneity to field-level prediction. A physics-coupled graph neural operator (PC-GNO) is proposed, which learns in function space the evolution of the mid-plane six-component stress tensor field. The training data are generated by full-scale, aggregate-resolved finite element models and comprise 359 cases at the impact velocities of 100, 135 and 165 m/s, corresponding to compressive strengths of 14, 23 and 48 MPa; the data source is verified by laboratory measurements. The coupling strength of message passing is constructed from the impedance contrast across the material interface, and the update is constrained by spectral normalization. Under the single-step primary setting, the mean accuracy of the three levels is 80.16% (91.45%, 77.80% and 71.22%), i.e. 27.81, 42.46 and 34.16 percentage points above the preceding-frame field of the same window; with 20 664 parameters it exceeds the 87.9% of the best Fourier neural operator configuration (about 812 times as many parameters) and the 71.0% of the closed-form linear regression. The output is the complete six-component tensor field rather than a fixed engineering quantity, so that strength criteria and energy indicators follow from one set of weights. For field-level response prediction in concrete penetration, this result provides a data-driven route, with conditions of applicability defined by the training and the evaluation.

cs.LG↗