arXiv subjects

Yingda Li

Publications and source records attributed to Yingda Li.

5 recordsLinked to original sources

arXiv2026

Style over Substance: A Shortcut Audit of Emotion-Description Preference Evaluation

Preference over model-generated emotion descriptions is emerging as a standard evaluation metric for multimodal emotion understanding, exemplified by the MER2026 MER-Prefer track on EmoPrefer. Such benchmarks assume that predicting the preferred description requires grounded cross-modal understanding of the video. We conduct a systematic shortcut audit of EmoPrefer using content-blind probes. A simple logistic regression using only description length and generator identity, without processing the text, video, or audio, performs comparably to LoRA-finetuned 7B text and audio-visual judges (65.8 versus 66.8 WAF on EmoPrefer-V2). Generator identity is recoverable from description text with 99.5 percent accuracy, every candidate pair contrasts two distinct generators, and the human preference labels agree with a fold-exclusive per-generator win-rate prior on 66 percent of the evaluated pairs. When the human label conflicts with this prior, trained judges still follow the style prior on 63 to 80 percent of the pairs. On a length-matched subset that neutralizes verbosity bias, the tested media configurations yield no statistically significant improvement, while an ODIN-inspired diagnostic that decouples the style shortcut leaves its content head near chance. These results do not imply that human preferences are inherently stylistic or that the descriptions contain no emotional information. Instead, they show that the current scores can be reached without verifying either description against the video. We recommend source-balanced pairing, strict length control, counter-stereotypical sliced reporting, and multi-annotator consensus for future cross-generator evaluations. Code is available at https://github.com/jiabingyang01/EmoPrefer-Audit.

cs.CV↗

arXiv2026

UAOR: Uncertainty-aware Observation Reinjection for Vision-Language-Action Models

Vision-Language-Action (VLA) models leverage pretrained Vision-Language Models (VLMs) as backbones to map images and instructions to actions, demonstrating remarkable potential for generalizable robotic manipulation. To enhance performance, existing methods often incorporate extra observation cues (e.g., depth maps, point clouds) or auxiliary modules (e.g., object detectors, encoders) to enable more precise and reliable task execution, yet these typically require costly data collection and additional training. Inspired by the finding that Feed-Forward Network (FFN) in language models can act as "key-value memory", we propose Uncertainty-aware Observation Reinjection (UAOR), an effective, training-free and plug-and-play module for VLA models. Specifically, when the current language model layer exhibits high uncertainty, measured by Action Entropy, it reinjects key observation information into the next layer's Feed-Forward Network (FFN) through attention retrieval. This mechanism directly augments the hidden states with observation evidence at high-uncertainty layers, enabling more accurate and reliable action generation. Comprehensive experiments show that our method consistently improves diverse VLA models across simulation and real-world tasks with minimal overhead. Notably, UAOR eliminates the need for additional observation cues or modules, making it a versatile and practical plug-in for existing VLA pipelines. The project page is at https://uaor.jiabingyang.cn.

cs.CV↗

arXiv2024

An Adaptive Online Smoother with Closed-Form Solutions and Information-Theoretic Lag Selection for Conditional Gaussian Nonlinear Systems

Data assimilation (DA) combines partial observations with dynamical models to improve state estimation. Filter-based DA uses only past and present data and is the prerequisite for real-time forecasts. Smoother-based DA exploits both past and future observations. It aims to fill in missing data, provide more accurate estimations, and develop high-quality datasets. However, the standard smoothing procedure requires using all historical state estimations, which is storage-demanding, especially for high-dimensional systems. This paper develops an adaptive-lag online smoother for a large class of complex dynamical systems with strong nonlinear and non-Gaussian features, which has important applications to many real-world problems. The adaptive lag allows the utilization of observations only within a nearby window, thus reducing computational complexity and storage needs. Online lag adjustment is essential for tackling turbulent systems, where temporal autocorrelation varies significantly over time due to intermittency, extreme events, and nonlinearity. Based on the uncertainty reduction in the estimated state, an information criterion is developed to systematically determine the adaptive lag. Notably, the mathematical structure of these systems facilitates the use of closed analytic formulae to calculate the online smoother and adaptive lag, avoiding empirical tunings as in ensemble-based DA methods. The adaptive online smoother is applied to studying three important scientific problems. First, it helps detect online causal relationships between state variables. Second, the advantage of reduced computational storage expenditure is illustrated via Lagrangian DA, a high-dimensional nonlinear problem. Finally, the adaptive smoother advances online parameter estimation with partial observations, emphasizing the role of the observed extreme events in accelerating convergence.

eess.SY↗

arXiv2021

Conditional Gaussian Nonlinear System: a Fast Preconditioner and a Cheap Surrogate Model For Complex Nonlinear Systems

Developing suitable approximate models for analyzing and simulating complex nonlinear systems is practically important. This paper aims at exploring the skill of a rich class of nonlinear stochastic models, known as the conditional Gaussian nonlinear system (CGNS), as both a cheap surrogate model and a fast preconditioner for facilitating many computationally challenging tasks. The CGNS preserves the underlying physics to a large extent and can reproduce intermittency, extreme events and other non-Gaussian features in many complex systems arising from practical applications. Three interrelated topics are studied. First, the closed analytic formulae of solving the conditional statistics provide an efficient and accurate data assimilation scheme. It is shown that the data assimilation skill of a suitable CGNS approximate forecast model outweighs that by applying an ensemble method even to the perfect model with strong nonlinearity, where the latter suffers from filter divergence. Second, the CGNS allows the development of a fast algorithm for simultaneously estimating the parameters and the unobserved variables with uncertainty quantification in the presence of only partial observations. Utilizing an appropriate CGNS as a preconditioner significantly reduces the computational cost in accurately estimating the parameters in the original complex system. Finally, the CGNS advances rapid and statistically accurate algorithms for computing the probability density function and sampling the trajectories of the unobserved state variables. These fast algorithms facilitate the development of an efficient and accurate data-driven method for predicting the linear response of the original system with respect to parameter perturbations based on a suitable CGNS preconditioner.

math.NA↗

arXiv2021

BAMCAFE: A Bayesian Machine Learning Advanced Forecast Ensemble Method for Complex Turbulent Systems with Partial Observations

Ensemble forecast based on physics-informed models is one of the most widely used forecast algorithms for complex turbulent systems. A major difficulty in such a method is the model error that is ubiquitous in practice. Data-driven machine learning (ML) forecasts can reduce the model error but they often suffer from the partial and noisy observations. In this paper, a simple but effective Bayesian machine learning advanced forecast ensemble (BAMCAFE) method is developed, which combines an available imperfect physics-informed model with data assimilation (DA) to facilitate the ML ensemble forecast. In the BAMCAFE framework, a Bayesian ensemble DA is applied to create the training data of the ML model, which reduces the intrinsic error in the imperfect physics-informed model simulations and provides the training data of the unobserved variables. Then a generalized DA is employed for the initialization of the ML ensemble forecast. In addition to forecasting the optimal point-wise value, the BAMCAFE also provides an effective approach of quantifying the forecast uncertainty utilizing a non-Gaussian probability density function that characterizes the intermittency and extreme events. It is shown using a two-layer Lorenz 96 model that the BAMCAFE method can significant improve the forecasting skill compared to the typical reduced-order imperfect models with bare truncation or stochastic parameterization for both the observed and unobserved large-scale variables. It is also shown via a nonlinear conceptual model that the BAMCAFE leads to a comparable non-Gaussian forecast uncertainty as the perfect model while the associated imperfect physics-informed model suffers from large forecast biases.

physics.ao-ph↗