arXiv subjects

Yixin Wang

Publications and source records attributed to Yixin Wang.

At least 55 records · Page 3Linked to original sources

arXiv2026

Goal-Oriented Influence-Maximizing Data Acquisition for Learning and Optimization

Active data acquisition is central to many learning and optimization tasks in deep neural networks, yet remains challenging because most approaches rely on predictive uncertainty estimates that are difficult to obtain reliably. To this end, we propose Goal-Oriented Influence- Maximizing Data Acquisition (GOIMDA), an active acquisition algorithm that avoids explicit posterior inference while remaining uncertainty-aware through inverse curvature. GOIMDA selects inputs by maximizing their expected influence on a user-specified goal functional, such as test loss, predictive entropy, or the value of an optimizer-recommended design. Leveraging first-order influence functions, we derive a tractable acquisition rule that combines the goal gradient, training-loss curvature, and candidate sensitivity to model parameters. We show theoretically that, for generalized linear models, GOIMDA approximates predictive-entropy minimization up to a correction term accounting for goal alignment and prediction bias, thereby, yielding uncertainty-aware behavior without maintaining a Bayesian posterior. Empirically, across learning tasks (including image and text classification) and optimization tasks (including noisy global optimization benchmarks and neural-network hyperparameter tuning), GOIMDA consistently reaches target performance with substantially fewer labeled samples or function evaluations than uncertainty-based active learning and Gaussian-process Bayesian optimization baselines.

stat.ML↗

arXiv2026

Band-Like Transport and Cation Off-Centring in Ag/Bi-Based Solar Absorbers

Ag(I)-Bi(III)-based semiconductors have gained substantial attention as nontoxic, stable alternatives to lead-halide perovskites for optoelectronics, but are widely limited by carrier localization, which severely restricts diffusion lengths. The most efficient Ag/Bi solar absorber is AgBiS2, but diffusion lengths in nanocrystal films are <50 nm. Carrier localization in this rock-salt (Fm-3m) system is believed to arise from cation disorder, and so we herein investigate the layered cation-ordered analogue. Through beyond-DFT simulations combined with neutron and X-ray powder diffraction, we reveal that off-centring of Ag+ and Bi3+ cations is energetically-favoured in this cation-ordered phase. Despite local distortions in the AgS6 and BiS6 octahedra, band-like transport takes place, which, surprisingly, also occurs in the cation-disordered rock-salt phase when these materials are made as bulk powders. The cubic-phase powders have the same degree of cation disorder as the nanocrystals that have carrier localization, which suggests that extrinsic factors play a determining role. We ascribe the intrinsic band-like transport of both phases of AgBiS2 to its close packing, ensuring high electronic dimensionality. These insights offer pathways for designing solar absorbers avoiding carrier localization limitations, and call for future efforts to enhance the efficiency of AgBiS2 photovoltaics to focus on large-grained thin films, or improved nanocrystal surface passivation.

cond-mat.mtrl-sci↗

arXiv2026

Identification and estimation of the conditional average treatment effect with nonignorable missing covariates, treatment, and outcome

Treatment effect heterogeneity is central to policy evaluation, social science, and precision medicine, where interventions can affect individuals differently. In observational studies, covariates, treatment, and outcomes are often only partially observed. When missingness depends on unobserved values (missing not at random; MNAR), standard methods can yield biased estimates of the conditional average treatment effect (CATE). This paper establishes nonparametric identification of the CATE under multivariate MNAR mechanisms that allow covariates, treatment, and outcomes to be MNAR. It also develops nonparametric and parametric estimators and proposes a sensitivity analysis framework for assessing robustness to violations of the missingness assumptions.

stat.ME↗

arXiv2026

QuEPT: Quantized Elastic Precision Transformers with One-Shot Calibration for Multi-Bit Switching

Elastic precision quantization enables multi-bit deployment via a single optimization pass, fitting diverse quantization scenarios.Yet, the high storage and optimization costs associated with the Transformer architecture, research on elastic quantization remains limited, particularly for large language models.This paper proposes QuEPT, an efficient post-training scheme that reconstructs block-wise multi-bit errors with one-shot calibration on a small data slice. It can dynamically adapt to various predefined bit-widths by cascading different low-rank adapters, and supports real-time switching between uniform quantization and mixed precision quantization without repeated optimization. To enhance accuracy and robustness, we introduce Multi-Bit Token Merging (MB-ToMe) to dynamically fuse token features across different bit-widths, improving robustness during bit-width switching. Additionally, we propose Multi-Bit Cascaded Low-Rank adapters (MB-CLoRA) to strengthen correlations between bit-width groups, further improve the overall performance of QuEPT. Extensive experiments demonstrate that QuEPT achieves comparable or better performance to existing state-of-the-art post-training quantization methods.Our code is available at https://github.com/xuke225/QuEPT

cs.CV↗

arXiv2026

Causal Representation Meets Stochastic Modeling under Generic Geometry

Learning meaningful causal representations from observations has emerged as a crucial task for facilitating machine learning applications and driving scientific discoveries in fields such as climate science, biology, and physics. This process involves disentangling high-level latent variables and their causal relationships from low-level observations. Previous work in this area that achieves identifiability typically focuses on cases where the observations are either i.i.d. or follow a latent discrete-time process. Nevertheless, many real-world settings require identifying latent variables that are continuous-time stochastic processes (e.g., multivariate point processes). To this end, we develop identifiable causal representation learning for continuous-time latent stochastic point processes. We study its identifiability by analyzing the geometry of the parameter space. Furthermore, we develop MUTATE, an identifiable variational autoencoder framework with a time-adaptive transition module to infer stochastic dynamics. Across simulated and empirical studies, we find that MUTATE can effectively answer scientific questions, such as the accumulation of mutations in genomics and the mechanisms driving neuron spike triggers in response to time-varying dynamics.

cs.LG↗

arXiv2026

Ligand Engineering for Precise Control of Ultrathin CsPbI3 Nanoplatelet Superlattices for Efficient Light-Emitting Diodes

Strongly-confined perovskite nanoplatelets (PeNPLs) offer opportunities not found in conventional isotropic nanocubes, especially in producing linearly polarized light, as well as enhancing outcoupling through control over the transition dipole moment. But this requires ultrathin nanoplatelets with three or fewer monolayers of PbI6 octahedra across the thickness, which are challenging to synthesise uniformly, and their luminescence is strongly affected by surface defects. Together, these limit the performance of ultrathin PeNPLs in light-emitting diodes (LEDs). Here, we address these challenges with an ancillary ligand engineering strategy. We demonstrate that ligands with phosphoryl functional groups strongly bind to the perovskite surface, while having an organic backbone that is not sterically bulky ensures high ligand density. By modulating nucleation and growth, these ancillary ligands lead to monodisperse PeNPLs that stack more uniformly when self-assembled into superlattices, with suppressed agglomeration. As a result, from edge-up PeNPL superlattices, we achieve enhanced degree of polarization, while from face-down PeNPL superlattices, we achieve enhanced outcoupling that results in LEDs with 13.1% external quantum efficiency, the highest reported for ultrathin PeNPL LEDs. This work establishes ancillary ligand-induced synthesis as a decisive route to achieve uniform nanoplatelets with robust orientation control, enabling full utilization of the multifunctionality of anisotropic PeNPLs.

physics.optics↗

arXiv2026

Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces

AI agents may soon become capable of autonomously completing valuable, long-horizon tasks in diverse domains. Current benchmarks either do not measure real-world tasks, or are not sufficiently difficult to meaningfully measure frontier models. To this end, we present Terminal-Bench 2.0: a carefully curated hard benchmark composed of 89 tasks in computer terminal environments inspired by problems from real workflows. Each task features a unique environment, human-written solution, and comprehensive tests for verification. We show that frontier models and agents score less than 65\% on the benchmark and conduct an error analysis to identify areas for model and agent improvement. We publish the dataset and evaluation harness to assist developers and researchers in future work at https://www.tbench.ai/ .

cs.SE↗

arXiv2026

On the Identifiability of Regime-Switching Models with Multi-Lag Dependencies

Identifiability is central to the interpretability of deep latent variable models, ensuring parameterisations are uniquely determined by the data-generating distribution. However, it remains underexplored for deep regime-switching time series. We develop a general theoretical framework for multi-lag Regime-Switching Models (RSMs), encompassing Markov Switching Models (MSMs) and Switching Dynamical Systems (SDSs). For MSMs, we formulate the model as a temporally structured finite mixture and prove identifiability of both the number of regimes and the multi-lag transitions in a nonlinear-Gaussian setting. For SDSs, we establish identifiability of the latent variables up to permutation and scaling via temporal structure, which in turn yields conditions for identifiability of regime-dependent latent causal graphs (up to regime/node permutations). Our results hold in a fully unsupervised setting through architectural and noise assumptions that are directly enforceable via neural network design. We complement the theory with a flexible variational estimator that satisfies the assumptions and validate the results on synthetic benchmarks. Across real-world datasets from neuroscience, finance, and climate, identifiability leads to more trustworthy interpretability analysis, which is crucial for scientific discovery.

stat.ML↗

arXiv2025

COPE: Chain-Of-Thought Prediction Engine for Open-Source Large Language Model Based Stroke Outcome Prediction from Clinical Notes

Predicting outcomes in acute ischemic stroke (AIS) guides clinical decision-making, patient counseling, and resource allocation. Clinical notes contain rich contextual information, but their unstructured nature limits their use in traditional predictive models. We developed and evaluated the Chain-of-Thought (CoT) Outcome Prediction Engine (COPE), a reasoning-enhanced large language model framework, for predicting 90-day functional outcomes after AIS from unstructured clinical notes. This study included 464 AIS patients with discharge summaries and 90-day modified Rankin Scale (mRS) scores. COPE uses a two-step CoT framework based on sequential open-source LLaMA-3-8B models: the first generates clinical reasoning, and the second outputs an mRS prediction. We compared COPE with GPT-4.1, ClinicalBERT, a structured variable-based machine learning model (Clinical ML), and a single-step LLM without CoT. Performance was evaluated using mean absolute error (MAE), accuracy within +/-1 mRS point, and exact accuracy. COPE achieved an MAE of 1.01 (95% CI 0.92-1.11), +/-1 accuracy of 74.4% (69.9, 78.8%), and exact accuracy of 32.8% (28.0, 37.6%), comparable to GPT-4.1 and superior to ClinicalBERT [MAE 1.24 (1.13-1.36)], Clinical ML [1.28 (1.18-1.39)], and the single-step LLM [1.20 (1.09-1.33)]. Subgroup analyses showed consistent performance across sex and age, with slightly higher error among older patients, those undergoing thrombectomy, and those with longer summaries. These findings demonstrate that COPE, a lightweight, interpretable, and privacy-preserving open-source framework, provides an accurate and practical solution for outcome prediction from unstructured clinical text.

cs.AI↗

arXiv2025

An Efficient Smoothing Damped Newton Method for Large-Scale Mathematical Programs with Equilibrium Constraints

Bilevel hyperparameter optimization has received growing attention thanks to the fast development of machine learning. Due to the tremendous size of data sets, the scale of bilevel hyperparameter optimization problem could be extremely large, posing great challenges in designing efficient numerical algorithms. In this paper, we focus on solving the large-scale mathematical programs with equilibrium constraints (MPEC) derived from hyperparameter selection of L1 support vector classification (L1-SVC). We propose a highly efficient smoothing damped Newton method (SDNM) for solving such MPEC. Compared with most existing algorithms where subproblems are solved by packages, our approach fully takes advantage of the structure of MPEC and therefore is package-free. Moreover, the proposed SDNM converges to C-stationary point under MPEC-LICQ with subproblem enjoys a quadratic convergence rate under proper assumptions. Extensive numerical results over LIBSVM dataset show the superior performance of SDNM over other state-of-art algorithms.

math.OC↗

arXiv2025

Permutative Preference Alignment from Listwise Ranking of Human Judgments

Aligning Large Language Models (LLMs) with human preferences is crucial in ensuring desirable and controllable model behaviors. Current methods, such as Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO), rely on the Bradley-Terry (B-T) model to maximize the likelihood of pairwise choices. However, when multiple responses are available, the B-T model fails to guarantee an accurate list ranking of the responses. To address this issue, we propose Permutative Preference Alignment (PPA), a novel offline listwise approach that incorporates the Normalized Discounted Cumulative Gain (NDCG), a widely-used ranking metric, as an alternative training objective for LLM alignment. We develop an end-to-end alignment algorithm by approximating NDCG with a differentiable surrogate loss. Experiments demonstrate that PPA outperforms existing pairwise and listwise methods on evaluation sets and general benchmarks such as AlpacaEval. Furthermore, we show that NDCG-based approaches improve ranking accuracy more effectively than B-T-based methods and provide a theoretical explanation for this improvement.

cs.CL↗

arXiv2025

On the Challenges and Opportunities in Generative AI

The field of deep generative modeling has grown rapidly in the last few years. With the availability of massive amounts of training data coupled with advances in scalable unsupervised learning paradigms, recent large-scale generative models show tremendous promise in synthesizing high-resolution images and text, as well as structured data such as videos and molecules. However, we argue that current large-scale generative AI models exhibit several fundamental shortcomings that hinder their widespread adoption across domains. In this work, our objective is to identify these issues and highlight key unresolved challenges in modern generative AI paradigms that should be addressed to further enhance their capabilities, versatility, and reliability. By identifying these challenges, we aim to provide researchers with insights for exploring fruitful research directions, thus fostering the development of more robust and accessible generative AI solutions.

cs.LG↗

arXiv2025

Evaluation of 3D Counterfactual Brain MRI Generation

Counterfactual generation offers a principled framework for simulating hypothetical changes in medical imaging, with potential applications in understanding disease mechanisms and generating physiologically plausible data. However, generating realistic structural 3D brain MRIs that respect anatomical and causal constraints remains challenging due to data scarcity, structural complexity, and the lack of standardized evaluation protocols. In this work, we convert six generative models into 3D counterfactual approaches by incorporating an anatomy-guided framework based on a causal graph, in which regional brain volumes serve as direct conditioning inputs. Each model is evaluated with respect to composition, reversibility, realism, effectiveness and minimality on T1-weighted brain MRIs (T1w MRIs) from the Alzheimer's Disease Neuroimaging Initiative (ADNI). In addition, we test the generalizability of each model with respect to T1w MRIs of the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA). Our results indicate that anatomically grounded conditioning successfully modifies the targeted anatomical regions; however, it exhibits limitations in preserving non-targeted structures. Beyond laying the groundwork for more interpretable and clinically relevant generative modeling of brain MRIs, this benchmark highlights the need for novel architectures that more accurately capture anatomical interdependencies.

eess.IV↗

arXiv2025

Counterfactual Voting Adjustment for Quality Assessment and Fairer Voting in Online Platforms with Helpfulness Evaluation

Efficient access to high-quality information is vital for online platforms. To promote more useful information, users not only create new content but also evaluate existing content, often through helpfulness voting. Although aggregated votes help service providers rank their user content, these votes are often biased by disparate accessibility per position and the cascaded influence of prior votes. For a fairer assessment of information quality, we propose the Counterfactual Voting Adjustment (CVA), a causal framework that accounts for the context in which individual votes are cast. Through preliminary and semi-synthetic experiments, we show that CVA effectively models the position and herding biases, accurately recovering the predefined content quality. In a real experiment, we demonstrate that reranking content based on the learned quality by CVA exhibits stronger alignment with both user sentiment and quality evaluation assessed by GPT-4o, outperforming system rankings based on aggregated votes and model-based rerankings without causal inference. Beyond the individual quality inference, our embeddings offer comparative insights into the behavioral dynamics of expert user groups across 120 major StackExchange communities.

cs.CE↗

arXiv2025

Explanation Design in Strategic Learning: Sufficient Explanations that Induce Non-harmful Responses

We study explanation design in algorithmic decision making with strategic agents, individuals who may modify their inputs in response to explanations of a decision maker's (DM's) predictive model. As the demand for transparent algorithmic systems continues to grow, most prior work assumes full model disclosure as the default solution. In practice, however, DMs such as financial institutions typically disclose only partial model information via explanations. Such partial disclosure can lead agents to misinterpret the model and take actions that unknowingly harm their utility. A key open question is how DMs can communicate explanations in a way that avoids harming strategic agents, while still supporting their own decision-making goals, e.g., minimising predictive error. In this work, we analyse well-known explanation methods, and establish a necessary condition to prevent explanations from misleading agents into self-harming actions. Moreover, with a conditional homogeneity assumption, we prove that action recommendation-based explanations (ARexes) are sufficient for non-harmful responses, mirroring the revelation principle in information design. To demonstrate how ARexes can be operationalised in practice, we propose a simple learning procedure that jointly optimises the predictive model and explanation policy. Experiments on synthetic and real-world tasks show that ARexes allow the DM to optimise their model's predictive performance while preserving agents' utility, offering a more refined strategy for safe and effective partial model disclosure.

cs.AI↗

arXiv2025

Last Layer Empirical Bayes

The task of quantifying the inherent uncertainty associated with neural network predictions is a key challenge in artificial intelligence. Bayesian neural networks (BNNs) and deep ensembles are among the most prominent approaches to tackle this task. Both approaches produce predictions by computing an expectation of neural network outputs over some distribution on the corresponding weights; this distribution is given by the posterior in the case of BNNs, and by a mixture of point masses for ensembles. Inspired by recent work showing that the distribution used by ensembles can be understood as a posterior corresponding to a learned data-dependent prior, we propose last layer empirical Bayes (LLEB). LLEB instantiates a learnable prior as a normalizing flow, which is then trained to maximize the evidence lower bound; to retain tractability we use the flow only on the last layer. We show why LLEB is well motivated, and how it interpolates between standard BNNs and ensembles in terms of the strength of the prior that they use. LLEB performs on par with existing approaches, highlighting that empirical Bayes is a promising direction for future research in uncertainty quantification.

cs.LG↗

arXiv2025

Systematic Bernoulli Generator Matrix Codes

This paper is concerned with the systematic Bernoulli generator matrix~(BGM) codes, which have been proved to be capacity-achieving over binary-input output-symmetric~(BIOS) channels in terms of bit-error rate~(BER). We prove that the systematic BGM codes are also capacity-achieving over BIOS channels in terms of frame-error rate (FER). To this end, we present a new framework to prove the coding theorems for binary linear codes. Different from the widely-accepted approach via ensemble enlargement, the proof directly applies to the systematic binary linear codes. The new proof indicates that the pair-wise independence condition is not necessary for proving the binary linear code ensemble to achieve the capacity of the BIOS channel. The Bernoulli parity-check~(BPC) codes, which fall within the framework of the systematic BGM codes with parity-check bits known at the decoder can also be proved to achieve the capacity. The presented framework also reveals a new mechanism pertained to the systematic linear codes that the systematic bits and the corresponding parity-check bits play different roles. Precisely, the noisy systematic bits are used to limit the list size of candidate codewords, while the noisy parity-check bits are used to select from the list the maximum likelihood codeword. For systematic BGM codes with finite length, we derive the lower bounds on the BER and FER, which can be used to predict the error floors. Numerical results show that the systematic BGM codes match well with the derived error floors. The performance in water-fall region can be improved with approaches in statistical physics and the error floors can be significantly improved by implementing the concatenated codes with the systematic BGM codes as the inner codes.

cs.IT↗

arXiv2025

Let Me Grok for You: Accelerating Grokking via Embedding Transfer from a Weaker Model

''Grokking'' is a phenomenon where a neural network first memorizes training data and generalizes poorly, but then suddenly transitions to near-perfect generalization after prolonged training. While intriguing, this delayed generalization phenomenon compromises predictability and efficiency. Ideally, models should generalize directly without delay. To this end, this paper proposes GrokTransfer, a simple and principled method for accelerating grokking in training neural networks, based on the key observation that data embedding plays a crucial role in determining whether generalization is delayed. GrokTransfer first trains a smaller, weaker model to reach a nontrivial (but far from optimal) test performance. Then, the learned input embedding from this weaker model is extracted and used to initialize the embedding in the target, stronger model. We rigorously prove that, on a synthetic XOR task where delayed generalization always occurs in normal training, GrokTransfer enables the target model to generalize directly without delay. Moreover, we demonstrate that, across empirical studies of different tasks, GrokTransfer effectively reshapes the training dynamics and eliminates delayed generalization, for both fully-connected neural networks and Transformers.

cs.LG↗