arXiv subjects

Kexin Chen

Publications and source records attributed to Kexin Chen.

At least 19 recordsLinked to original sources

arXiv2026

Renting, Owning, and Moving: Housing Tenure Choice with Transaction Costs

The wealth thresholds at which a household switches between renting and owning depend on its preferences, but the dates at which some owner sells or every renter buys do not. In our finite-horizon model, a household chooses consumption, a portfolio, and costly tenure changes while rents, ownership expenses, and transaction costs vary over time on announced dates. When financial markets span house-price risk, these dates are those of a deterministic minimum cost schedule for housing, discounted at the risk-adjusted housing carry and computed from a scalar reflected equation. Preferences, housing services, and labor income only locate the thresholds inside these timing windows, and house-price coefficients enter the dates only through the carry, because financial hedging offsets the exposure of the home. When owning costs at least as much as renting, the windows concern selling, and an anticipated selling surcharge closes the selling window before the surcharge peaks and keeps it closed for years afterwards. We also establish strong duality and the regularity of the thresholds and compute them by integral equations. Numerically, the windows disappear when short sales or borrowing are restricted but do not move with the preferences we vary under an uninsurable increase in the ownership cost.

math.OC↗

arXiv2026

Does AI Help or Harm? Endogenous Information Acquisition with AI Advice

AI advice increasingly enters human decision workflows before costly information acquisition and final action. This paper studies how such advice affects decision quality, measured by the gross utility of the final action. Advice shifts beliefs before the decision maker chooses an acquisition experiment with uniformly posterior-separable (UPS) costs. The analysis characterizes gross effects through the continuation gross value generated by net-optimal acquisition. Convexity of this value function is necessary and sufficient for every AI signal to weakly improve gross utility at every prior, while nonconvexity allows a binary signal to cause gross harm. The criterion gives tight state-space results. Every AI signal is weakly gross-improving in binary-state problems under arbitrary actions, payoffs, and UPS costs. A minimal three-state problem can generate gross harm through concentrated shutdown, where valuable acquisition stops. For classification problems with Shannon acquisition costs, the criterion becomes an active-set slope test. It guarantees weak gross improvement for homogeneous classification and all three-state weighted classification problems, while a minimal four-state failure arises from attention dilution, where acquisition shifts away from high-stake distinctions. Along a Blackwell-increasing mixture path, gross utility can first decrease and then increase. With endogenous interpretation, cheaper interpretation can reduce gross utility.

cs.GT↗

arXiv2026

Band Structure Modulation of ZrO2 Nanoparticles for Control of CO Adsorption Properties: A Combined Density Functional Theory - Density Functional Tight Binding Study

We present a combined density functional theory (DFT) and density functional tight binding (DFTB) study of zirconia (ZrO2) nanoparticles of experimentally relevant sizes of several nanometers and their interactions with the CO molecule. A hybrid DFTB - Force Field (DFTB-FF) framework is developed, whereby band structure calculations rely on an existing Slater-Koster framework, while the accuracy of structural optimization and adsorption properties is controlled by the introduction of classical long-range interatomic potentials into DFTB instead of the traditional repulsive potentials. Additionally, coordination-dependent Zr-C potentials are introduced to account for the distinct local chemical environments of bulk-like facet sites and under-coordinated tip and edge sites, thereby improving the description of CO adsorption. This hybrid DFTB-FF approach substantially improves the robustness of geometry optimization and provides a practical strategy for extending the applicability of DFTB to complex oxide nanostructures. The calculations reveal that termination stoichiometry can be used to engineer intrinsic, p-type, or n-type electronic structures and thereby tune the adsorption activity of zirconia nanoparticles. While stoichiometric nanoparticles do not activate the C-O bond, low-coordinated sites in Zr-rich (n-type) nanoparticles exhibit chemisorption accompanied by charge donation into a CO antibonding LUMO-derived orbital, resulting in C-O bond activation. These results demonstrate that stoichiometry-controlled electronic structure and under-coordinated surface sites introduced by nanostructuring play a key role in governing the adsorption strength and reactivity of zirconia nanoparticles.

cond-mat.mtrl-sci↗

arXiv2026

Kimi K3: Open Frontier Intelligence

We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window. Kimi K3 is built on Kimi Delta Attention and Attention Residuals, which improve information flow across sequence length and model depth. Together with Stable LatentMoE, which effectively activates 16 of 896 routed experts per token, and refined training and data recipes, these advances yield an approximately 2.5x improvement in overall scaling efficiency over Kimi K2. Post-training highlights reinforcement learning across general, agentic, and coding domains and multiple reasoning-effort levels, enabling compositional generalization and robust long-horizon execution. At 2.8T scale, Kimi K3 is supported by infrastructure advances in multiple areas: algorithm-system co-design for KDA, perfectly balanced expert-parallel training with efficient memory management, million-token agentic RL with persistent rollout and sandbox states, and deployment innovations. Extensive evaluations show that Kimi K3 achieves frontier-level performance across long-horizon coding, agentic, knowledge, reasoning, and vision tasks. While its overall performance still trails the most powerful proprietary models, namely Claude Fable 5 and GPT-5.6 Sol, Kimi K3 consistently outperforms other open and proprietary models evaluated in our suite. We release the full Kimi K3 model weights to facilitate future research and accelerate the broader deployment and adoption of frontier intelligence.

cs.CL↗

arXiv2026

When Extreme Darkness Meets Motion Blur: MeanFlow for Unified RAW Restoration

Extremely low-light RAW enhancement aims to recover severely attenuated sensor signals, yet existing methods often focus on illumination and noise while overlooking the motion-induced degradations inherent in practical low-light imaging. We present a framework for robust extremely low-light RAW enhancement under realistic acquisition degradations. First, we introduce See in the Degraded Extremely Dark (SIDED), a new dataset that applies controlled motion degradation to extremely low-light RAW pairs while retaining their original sensor noise. Second, we propose a unified RAW tokenizer equipped with explicit domain-conditioned representation calibration to align extremely low-light and well-exposed RAW data, followed by a MeanFlow that performs enhancement in a single function evaluation. To our knowledge, this is the first work to formulate extremely low-light RAW enhancement under realistic motion-degraded acquisition and address it with MeanFlow. We further introduce a physics-guided refinement model to strengthen illumination--reflectance consistency, pixel fidelity, and color preservation without incurring additional inference cost. Extensive experiments demonstrate that our framework achieves state-of-the-art performance in extremely low-light RAW enhancement, and robustly handles coupled motion and noise degradations.

cs.CV↗

arXiv2026

Evaluation of Winning Solutions of 2025 Low Power Computer Vision Challenge

The IEEE Low-Power Computer Vision Challenge (LPCVC) aims to promote the development of efficient vision models for edge devices, balancing accuracy with constraints such as latency, memory capacity, and energy use. The 2025 challenge featured three tracks: (1) Image classification under various lighting conditions and styles, (2) Open-Vocabulary Segmentation with Text Prompt, and (3) Monocular Depth Estimation. This paper presents the design of LPCVC 2025, including its competition structure and evaluation framework, which integrates the Qualcomm AI Hub for consistent and reproducible benchmarking. The paper also introduces the top-performing solutions from each track and outlines key trends and observations. The paper concludes with suggestions for future computer vision competitions.

cs.CV↗

arXiv2026

Decoding Hidden Deception in Reasoning LLMs: Activation Explainers for Deception Auditing

As LLMs acquire stronger reasoning capabilities, deceptive behavior becomes an increasingly serious safety concern. Existing deception monitors either score visible transcripts or derive scalar probe scores from representation vectors, leaving little inspectable evidence about why a response is suspicious. We introduce STATEWITNESS, an activation explainer for deception auditing. A separate decoder reads a target model's hidden states, then answers natural-language queries or emits structured reports about them. We evaluate STATEWITNESS on two target reasoning LLMs across seven deception datasets. STATEWITNESS reaches 0.916 mean AUROC, a relative gain of 11.6% over the best black-box text monitor and 25.0% over the best activation-probe baseline under the same evaluation protocol. When combined with existing monitors, STATEWITNESS reduces missed deceptive examples in simple threshold ensembles. Beyond scalar detection, the decoder returns query-level answers, schema reports, and token- or sentence-level evidence traces for human inspection. We view this interface as a potential building block for broader interpretability and alignment tools.

cs.CL↗

arXiv2026

Provable Knowledge Acquisition and Extraction in One-Layer Transformers

Large language models may encounter factual knowledge during pre-training yet fail to reliably use that knowledge after fine-tuning. Despite growing empirical evidence that MLP layers store factual associations and fine-tuning affects factual recall, the training-dynamics mechanisms linking next-token pre-training, knowledge storage, and post-fine-tuning extraction remain poorly understood. We study this problem in a stylized one-layer transformer with self-attention and MLP modules, trained by next-token prediction and subsequently fine-tuned on question-answering data. Under suitable regularity conditions, we first prove that the model reaches near-optimal pre-training loss while learning structured attention patterns and relation-specific feature directions, giving a mechanism for factual knowledge acquisition. We then show that fine-tuning can turn the Q&A prompt format into a trigger for pre-trained relation features, enabling the model to extract facts that are not revisited during fine-tuning. Our analysis yields a relation-covering characterization of knowledge extraction: fine-tuning need not revisit every stored subject-answer pair, but it must cover enough latent relation-template directions through which facts were encoded during pre-training. Consequently, extraction improves with pre-training multiplicity and fine-tuning coverage, but becomes harder as the relation-template universe grows. Conversely, insufficient coverage leads to a failure regime in which facts may be stored but remain inaccessible, providing a stylized mechanism for hallucination. The theory applies to both full and low-rank fine-tuning, offering insight into why low-rank adaptation can recover pre-trained factual knowledge when relation coverage is sufficient. Experiments on synthetic data and PopQA-based GPT-2/Llama models support the predicted trends.

cs.LG↗

arXiv2026

VeriRAG: A Retrieval-Augmented Framework for Automated RTL Testability Repair

Large language models (LLMs) have demonstrated immense potential in computer-aided design (CAD), particularly for automated debugging and verification within electronic design automation (EDA) tools. However, Design for Testability (DFT) remains a relatively underexplored area. This paper presents VeriRAG, the first LLM-assisted DFT-EDA framework. VeriRAG leverages a Retrieval-Augmented Generation (RAG) approach to enable LLM to revise code to ensure DFT compliance. VeriRAG integrates (1) an autoencoder-based similarity measurement model for precise retrieval of reference RTL designs for the LLM, and (2) an iterative code revision pipeline that allows the LLM to ensure DFT compliance while maintaining synthesizability. To support VeriRAG, we introduce VeriDFT, a Verilog-based DFT dataset curated for DFT-aware RTL repairs. VeriRAG retrieves structurally similar RTL designs from VeriDFT, each paired with a rigorously validated correction, as references for code repair. With VeriRAG and VeriDFT, we achieve fully automated DFT correction -- resulting in a 7.72-fold improvement in successful repair rate compared to the zero-shot baseline (Fig. 5 in Section V). Ablation studies further confirm the contribution of each component of the VeriRAG framework. We open-source our data, models, and scripts at https://github.com/HarminChee/VeriRAG.

cs.AR↗

arXiv2026

LLM-VA: Resolving the Jailbreak-Overrefusal Trade-off via Vector Alignment

Safety-aligned LLMs suffer from two failure modes: jailbreak (answering harmful inputs) and over-refusal (declining benign queries). Existing vector steering methods adjust the magnitude of answer vectors, but this creates a fundamental trade-off -- reducing jailbreak increases over-refusal and vice versa. We identify the root cause: LLMs encode the decision to answer (answer vector $v_a$) and the judgment of input safety (benign vector $v_b$) as nearly orthogonal directions, treating them as independent processes. We propose LLM-VA, which aligns $v_a$ with $v_b$ through closed-form weight updates, making the model's willingness to answer causally dependent on its safety assessment -- without fine-tuning or architectural changes. Our method identifies vectors at each layer using SVMs, selects safety-relevant layers, and iteratively aligns vectors via minimum-norm weight modifications. Experiments on 12 LLMs demonstrate that LLM-VA achieves 11.45% higher F1 than the best baseline while preserving 95.92% utility, and automatically adapts to each model's safety bias without manual tuning. Code and models are available at https://hotbento.github.io/LLM-VA-Web/.

cs.LG↗

arXiv2026

Robust dividend policy: Equivalence of Epstein-Zin and Maenhout preferences

In a continuous-time economy, this paper formulates the Epstein-Zin preference for discounted dividends received by an investor as an Epstein-Zin singular control utility. We introduce a backward stochastic differential equation with an aggregator integrated with respect to a singular control, prove its well-posedness, and show that it coincides with the Epstein-Zin singular control utility. We then establish that this formulation is equivalent to a robust dividend policy chosen by the firm's executive under the Maenhout's ambiguity-averse preference. In particular, the robust dividend policy takes the form of a threshold strategy on the firm's surplus process, where the threshold level is characterized as the free boundary of a Hamilton-Jacobi-Bellman variational inequality. Therefore, dividend-caring investors can choose firms that match their preferences by examining stock's dividend policies and financial statements, whereas executives can make use of dividend to signal their confidence, in the form of ambiguity aversion, on realizing the earnings implied by their financial statements.

q-fin.MF↗

arXiv2026

Differential Privacy in Two-Layer Networks: How DP-SGD Harms Fairness and Robustness

Differentially private learning is essential for training models on sensitive data, but empirical studies consistently show that it can degrade performance, introduce fairness issues like disparate impact, and reduce adversarial robustness. The theoretical underpinnings of these phenomena in modern, non-convex neural networks remain largely unexplored. This paper introduces a unified feature-centric framework to analyze the feature learning dynamics of differentially private stochastic gradient descent (DP-SGD) in two-layer ReLU convolutional neural networks. Our analysis establishes test loss bounds governed by a crucial metric: the feature-to-noise ratio (FNR). We demonstrate that the noise required for privacy leads to suboptimal feature learning, and specifically show that: 1) imbalanced FNRs across classes and subpopulations cause disparate impact; 2) even in the same class, noise has a greater negative impact on semantically long-tailed data; and 3) noise injection exacerbates vulnerability to adversarial attacks. Furthermore, our analysis reveals that the popular paradigm of public pre-training and private fine-tuning does not guarantee improvement, particularly under significant feature distribution shifts between datasets. Experiments on synthetic and real-world data corroborate our theoretical findings.

cs.LG↗

arXiv2025

RP-CATE: Recurrent Perceptron-based Channel Attention Transformer Encoder for Industrial Hybrid Modeling

Nowadays, industrial hybrid modeling which integrates both mechanistic modeling and machine learning-based modeling techniques has attracted increasing interest from scholars due to its high accuracy, low computational cost, and satisfactory interpretability. Nevertheless, the existing industrial hybrid modeling methods still face two main limitations. First, current research has mainly focused on applying a single machine learning method to one specific task, failing to develop a comprehensive machine learning architecture suitable for modeling tasks, which limits their ability to effectively represent complex industrial scenarios. Second, industrial datasets often contain underlying associations (e.g., monotonicity or periodicity) that are not adequately exploited by current research, which can degrade model's predictive performance. To address these limitations, this paper proposes the Recurrent Perceptron-based Channel Attention Transformer Encoder (RP-CATE), with three distinctive characteristics: 1: We developed a novel architecture by replacing the self-attention mechanism with channel attention and incorporating our proposed Recurrent Perceptron (RP) Module into Transformer, achieving enhanced effectiveness for industrial modeling tasks compared to the original Transformer. 2: We proposed a new data type called Pseudo-Image Data (PID) tailored for channel attention requirements and developed a cyclic sliding window method for generating PID. 3: We introduced the concept of Pseudo-Sequential Data (PSD) and a method for converting industrial datasets into PSD, which enables the RP Module to capture the underlying associations within industrial dataset more effectively. An experiment aimed at hybrid modeling in chemical engineering was conducted by using RP-CATE and the experimental results demonstrate that RP-CATE achieves the best performance compared to other baseline models.

cs.LG↗

arXiv2025

Generative AI as Digital Representatives in Collective Decision-Making: A Game-Theoretical Approach

Generative Artificial Intelligence (GenAI) enables digital representatives to make decisions on behalf of team members in collaborative tasks, but faces challenges in accurately representing preferences. While supplying GenAI with detailed personal information improves representation fidelity, feasibility constraints make complete information access impractical. We bridge this gap by developing a game-theoretic framework that models strategic information revelation to GenAI in collective decision-making. The technical challenges lie in characterizing members' equilibrium behaviors under interdependent strategies and quantifying the imperfect preference learning outcomes by digital representatives. Our contribution includes closed-form equilibrium characterizations that reveal how members strategically balance team decision preference against communication costs. Our analysis yields an interesting finding: Conflicting preferences between team members drive competitive information revelation, with members revealing more information than those with aligned preferences. While digital representatives produce aggregate preference losses no smaller than direct participation, individual members may paradoxically achieve decisions more closely aligned with their preferences when using digital representatives, particularly when manual participation costs are high or when GenAI systems are sufficiently advanced.

cs.GT↗

arXiv2025

ORFuzz: Fuzzing the "Other Side" of LLM Safety -- Testing Over-Refusal

Large Language Models (LLMs) increasingly exhibit over-refusal - erroneously rejecting benign queries due to overly conservative safety measures - a critical functional flaw that undermines their reliability and usability. Current methods for testing this behavior are demonstrably inadequate, suffering from flawed benchmarks and limited test generation capabilities, as highlighted by our empirical user study. To the best of our knowledge, this paper introduces the first evolutionary testing framework, ORFuzz, for the systematic detection and analysis of LLM over-refusals. ORFuzz uniquely integrates three core components: (1) safety category-aware seed selection for comprehensive test coverage, (2) adaptive mutator optimization using reasoning LLMs to generate effective test cases, and (3) OR-Judge, a human-aligned judge model validated to accurately reflect user perception of toxicity and refusal. Our extensive evaluations demonstrate that ORFuzz generates diverse, validated over-refusal instances at a rate (6.98% average) more than double that of leading baselines, effectively uncovering vulnerabilities. Furthermore, ORFuzz's outputs form the basis of ORFuzzSet, a new benchmark of 1,855 highly transferable test cases that achieves a superior 63.56% average over-refusal rate across 10 diverse LLMs, significantly outperforming existing datasets. ORFuzz and ORFuzzSet provide a robust automated testing framework and a valuable community resource, paving the way for developing more reliable and trustworthy LLM-based software systems.

cs.SE↗

arXiv2025

OD-MoE: On-Demand Expert Loading for Cacheless Edge-Distributed MoE Inference

Mixture-of-Experts (MoE), while offering significant advantages as a Large Language Model (LLM) architecture, faces substantial challenges when deployed on low-cost edge devices with tight memory constraints. Expert offloading mitigates this issue by storing expert parameters in CPU memory and caching a subset of popular experts in GPU memory. Although this approach improves GPU memory utilization by caching only the likely-used experts, the GPU memory reserved for expert caching is underutilized compared with dense LLMs. This paper presents OD-MoE, a distributed MoE inference framework that obviates the need for expert caches via fully on-demand expert loading. OD-MoE is built upon two key mechanisms: 1) parallelizing expert loading and expert computation across distributed edge nodes, and 2) an ultra-accurate emulative predictor that forecasts expert activations multiple layers ahead while expert computation is ongoing. With these innovations, OD-MoE dynamically loads each target expert to one of the distributed nodes just-in-time before its activation and promptly evicts it afterward, freeing GPU memory for subsequent experts. We comprehensively benchmark OD-MoE against state-of-the-art MoE offloading systems on a ten-node testbed. Experimental results show that: 1) OD-MoE achieves 99.94% expert activation prediction accuracy, substantially surpassing all existing methods; and 2) OD-MoE delivers approximately 75% of the decoding speed of a fully GPU-cached MoE deployment while using only 1/3 of the GPU memory. More importantly, by eliminating the need for expert caches, OD-MoE enables MoE inference on edge nodes with less-than-1GB GPU memory, paving the way for practical MoE deployment of low-cost IoT devices at the edge in the LLM era.

cs.DC↗

arXiv2025

DynamicRTL: RTL Representation Learning for Dynamic Circuit Behavior

There is a growing body of work on using Graph Neural Networks (GNNs) to learn representations of circuits, focusing primarily on their static characteristics. However, these models fail to capture circuit runtime behavior, which is crucial for tasks like circuit verification and optimization. To address this limitation, we introduce DR-GNN (DynamicRTL-GNN), a novel approach that learns RTL circuit representations by incorporating both static structures and multi-cycle execution behaviors. DR-GNN leverages an operator-level Control Data Flow Graph (CDFG) to represent Register Transfer Level (RTL) circuits, enabling the model to capture dynamic dependencies and runtime execution. To train and evaluate DR-GNN, we build the first comprehensive dynamic circuit dataset, comprising over 6,300 Verilog designs and 63,000 simulation traces. Our results demonstrate that DR-GNN outperforms existing models in branch hit prediction and toggle rate prediction. Furthermore, its learned representations transfer effectively to related dynamic circuit tasks, achieving strong performance in power estimation and assertion prediction.

cs.LG↗

arXiv2025

Transmitter-Side Beyond-Diagonal RIS-Enabled Integrated Sensing and Communications

Beyond diagonal reconfigurable intelligent surfaces (BD-RIS) have emerged as a promising technology for 6G wireless networks, offering more advanced control over electromagnetic wave propagation than conventional diagonal RIS. This paper proposes a novel integrated sensing and communication (ISAC) framework that incorporates BD-RIS at the transmitter. This not only opens the door to enhanced sensing and communication performance, but also alleviates the need for large-scale fully digital radio frequency (RF) chains at the transmitter. Based on the proposed system model, we formulate a normalized weighted optimization problem to jointly design the active beamforming and the BD-RIS scattering matrix with the aim of jointly minimizing the trace of the Cramér-Rao bound (CRB) for sensing targets and maximizing the sum rate (SR) for communication users. To address this highly coupled optimization problem, we propose a novel and low-complexity iterative algorithm that efficiently solves the active beamforming and scattering matrix subproblems by transforming each into a series of tractable projection problems with closed-form solutions. Numerical results show the appealing capability of the transmitter-side BD-RIS-aided ISAC over conventional diagonal RIS-aided ISAC in enhancing both sensing and communication performance. Moreover, compared to the classic iterative algorithm, the proposed algorithm offers enhanced dual-functional performance while significantly reducing the computational complexity.

eess.SP↗