arXiv subjects

Shengjie Li

Publications and source records attributed to Shengjie Li.

14 recordsLinked to original sources

arXiv2026

NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models

Vision-Language-Action (VLA) models are formulated to ground instructions in visual context and generate action sequences for robotic manipulation. Despite recent progress, VLA models still face structure-blind backbones, backbone-bound generalization, and flat single-objective optimization. To address these challenges, we propose a novel Neuro-Symbolic Vision-Language-Action (NS-VLA) framework. It introduces a Neuro-Symbolic Encoder for plan-constrained primitive inference, a Neuro-Symbolic Solver that conditions a backbone-agnostic policy on the active primitive, and Hierarchical Joint Policy Optimization with reward-granularity matching. Experiments on robotic manipulation benchmarks demonstrate that NS-VLA outperforms previous methods in both one-shot training and data-perturbed settings, while simultaneously exhibiting superior zero-shot generalizability and expanded exploration space. Our code is publicly available.

cs.RO↗

arXiv2026

ICLE++: Modeling Fine-Grained Traits for Holistic Essay Scoring

The majority of the recently-developed models for automated essay scoring (AES) are evaluated solely on the ASAP corpus. However, ASAP is not without its limitations. For instance, it is not clear whether models trained on ASAP can generalize well when evaluated on other corpora. In light of these limitations, we introduce ICLE++, a corpus of persuasive student essays annotated with both holistic scores and trait-specific scores. Not only can ICLE++ be used to test the generalizability of AES models trained on ASAP, but it can also facilitate the evaluation of models developed for newer AES problems such as multi-trait scoring and cross-prompt scoring. We believe that ICLE++, which represents a culmination of our long-term effort in annotating the essays in the ICLE corpus, contributes to the set of much-needed annotated corpora for AES research.

cs.CL↗

arXiv2026

Adaptive Loss Balancing for Noise-Robust GRPO in Generative Recommendation

Reinforcement learning (RL) presents a promising avenue for enhancing generative recommendation beyond supervised imitation, leveraging reward signals to guide policy improvement. However, its efficacy is critically contingent on the trustworthiness of the reward model for the samples it evaluates. In practice, production rankers, the widely adopted reward models, are trained on exposure-biased logs, leading to sample-dependent inaccuracies that violate this assumption. Our stratified analysis uncovers a consistent pattern: reward guidance is most beneficial when the policy exhibits uncertainty and the ranker can effectively discriminate the ground-truth item from rollout negatives. On other samples, the reward signal is either negligible or detrimental, highlighting the risk of uniform RL application. To address such an issue, we introduce AdaGRPO, a novel framework that treats reward-guided optimization as selective admission rather than uniform pressure. Training is anchored in supervised negative log-likelihood, while the GRPO objective is gated by a binary, per-sample clip determined by two rollout diagnostics: policy-side difficulty and reward discriminability. Instances failing either diagnostic default to pure supervision, ensuring stability and mitigating the amplification of noisy gradients. We validate AdaGRPO on a large-scale e-commerce dataset. At the best intermediate checkpoint, it elevates HR@10 from 11.01% to 12.18% while constraining hallucination below 0.22%, and maintains robustness at the final checkpoint (HR@10 11.63%, hallucination 0.27%), outperforming fixed NLL--GRPO mixtures across the retrieval--validity frontier. In production A/B tests, AdaGRPO achieves statistically significant gains in click-through rate and dwell time, confirming its practical utility.

cs.LG↗

arXiv2026

GenRec: A Preference-Oriented Generative Framework for Large-Scale Recommendation

Generative Retrieval (GR) offers a promising paradigm for recommendation through next-token prediction (NTP). However, scaling it to large-scale industrial systems introduces three challenges: (i) within a single request, the identical model inputs may produce inconsistent outputs due to the pagination request mechanism; (ii) the prohibitive cost of encoding long user behavior sequences with multi-token item representations based on semantic IDs, and (iii) aligning the generative policy with nuanced user preference signals. We present GenRec, a preference-oriented generative framework deployed on the JD App that addresses above challenges within a single decoder-only architecture. For training objective, we propose Page-wise NTP task, which supervises over an entire interaction page rather than each interacted item individually, providing denser gradient signal and resolving the one-to-many ambiguity of point-wise training. On the prefilling side, an asymmetric linear Token Merger compresses multi-token Semantic IDs in the prompt while preserving full-resolution decoding, reducing input length by ~2X with negligible accuracy loss. To further align outputs with user satisfaction, we introduce GRPO-SR, a reinforcement learning method that pairs Group Relative Policy Optimization with NLL regularization for training stability, and employs Hybrid Rewards combining a dense reward model with a relevance gate to mitigate reward hacking. In month-long online A/B tests serving production traffic, GenRec achieves 9.5% improvement in click count and 8.7% in transaction count over the existing pipeline.

cs.IR↗

arXiv2026

Inertial accelerated primal-dual algorithms for non-smooth convex optimization problems with linear equality constraints

This paper is devoted to the study of an inertial accelerated primal-dual algorithm, which is based on a second-order differential system with time scaling, for solving a non-smooth convex optimization problem with linear equality constraints. We first introduce a second-order differential system with time scaling associated with the non-smooth convex optimization problem, and then obtain fast convergence rates for the primal-dual gap, the feasibility violation, and the objective residual along the trajectory generated by this system. Subsequently, based on the setting of the parameters involved, we propose an inertial accelerated primal-dual algorithm from the time discretization of this system. We also establish fast convergence rates for the primal-dual gap, the feasibility violation, and the objective residual. Furthermore, we demonstrate the efficacy of the proposed algorithm through numerical experiments.

math.OC↗

arXiv2026

Primal-dual dynamical systems with closed-loop control for convex optimization in continuous and discrete time

This paper develops a primal-dual dynamical system where the coefficients are designed in closed-loop way for solving a convex optimization problem with linear equality constraints. We first introduce a ``second-order primal" + ``first-order dual'' continuous-time dynamical system, in which both the time scaling and Hessian-driven damping are governed by a feedback control of the gradient for the Lagrangian function. This system achieves the fast convergence rates for the primal-dual gap, the feasibility violation, and the objective residual along its trajectory. Subsequently, by time discretization of this system, we develop an accelerated primal-dual algorithm with a gradient-defined adaptive step size. We also obtain convergence rates for the primal-dual gap, the feasibility violation, and the objective residual. Furthermore, we provide numerical results to demonstrate the practical efficacy and superior performance of the proposed algorithm.

math.OC↗

arXiv2025

UAGLNet: Uncertainty-Aggregated Global-Local Fusion Network with Cooperative CNN-Transformer for Building Extraction

Building extraction from remote sensing images is a challenging task due to the complex structure variations of the buildings. Existing methods employ convolutional or self-attention blocks to capture the multi-scale features in the segmentation models, while the inherent gap of the feature pyramids and insufficient global-local feature integration leads to inaccurate, ambiguous extraction results. To address this issue, in this paper, we present an Uncertainty-Aggregated Global-Local Fusion Network (UAGLNet), which is capable to exploit high-quality global-local visual semantics under the guidance of uncertainty modeling. Specifically, we propose a novel cooperative encoder, which adopts hybrid CNN and transformer layers at different stages to capture the local and global visual semantics, respectively. An intermediate cooperative interaction block (CIB) is designed to narrow the gap between the local and global features when the network becomes deeper. Afterwards, we propose a Global-Local Fusion (GLF) module to complementarily fuse the global and local representations. Moreover, to mitigate the segmentation ambiguity in uncertain regions, we propose an Uncertainty-Aggregated Decoder (UAD) to explicitly estimate the pixel-wise uncertainty to enhance the segmentation accuracy. Extensive experiments demonstrate that our method achieves superior performance to other state-of-the-art methods. Our code is available at https://github.com/Dstate/UAGLNet

cs.CV↗

arXiv2024

A Depression Detection Method Based on Multi-Modal Feature Fusion Using Cross-Attention

Depression, a prevalent and serious mental health issue, affects approximately 3.8\% of the global population. Despite the existence of effective treatments, over 75\% of individuals in low- and middle-income countries remain untreated, partly due to the challenge in accurately diagnosing depression in its early stages. This paper introduces a novel method for detecting depression based on multi-modal feature fusion utilizing cross-attention. By employing MacBERT as a pre-training model to extract lexical features from text and incorporating an additional Transformer module to refine task-specific contextual understanding, the model's adaptability to the targeted task is enhanced. Diverging from previous practices of simply concatenating multimodal features, this approach leverages cross-attention for feature integration, significantly improving the accuracy in depression detection and enabling a more comprehensive and precise analysis of user emotions and behaviors. Furthermore, a Multi-Modal Feature Fusion Network based on Cross-Attention (MFFNC) is constructed, demonstrating exceptional performance in the task of depression identification. The experimental results indicate that our method achieves an accuracy of 0.9495 on the test dataset, marking a substantial improvement over existing approaches. Moreover, it outlines a promising methodology for other social media platforms and tasks involving multi-modal processing. Timely identification and intervention for individuals with depression are crucial for saving lives, highlighting the immense potential of technology in facilitating early intervention for mental health issues.

cs.CL↗

arXiv2023

The Behavior of Error Bounds via Moreau Envelopes

In this paper, we first establish the equivalence of three types of error bounds: uniformized Kurdyka-Łojasiewicz (u-KL) property, uniformized level-set subdifferential error bound (u-LSEB) and uniformized Hölder error bound (u-HEB) for prox-regular functions. Then we study the behavior of the level-set subdifferential error bound (LSEB) and the local Hölder error bound (LHEB) which is expressed respectively by Moreau envelopes, under suitable assumptions. Finally, in order to illustrate our main results and to compare them with those of recent references, some examples are also given.

math.OC↗

arXiv2023

Multimodal Propaganda Processing

Propaganda campaigns have long been used to influence public opinion via disseminating biased and/or misleading information. Despite the increasing prevalence of propaganda content on the Internet, few attempts have been made by AI researchers to analyze such content. We introduce the task of multimodal propaganda processing, where the goal is to automatically analyze propaganda content. We believe that this task presents a long-term challenge to AI researchers and that successful processing of propaganda could bring machine understanding one important step closer to human understanding. We discuss the technical challenges associated with this task and outline the steps that need to be taken to address it.

cs.CL↗

arXiv2022

End-to-End Neural Discourse Deixis Resolution in Dialogue

We adapt Lee et al.'s (2018) span-based entity coreference model to the task of end-to-end discourse deixis resolution in dialogue, specifically by proposing extensions to their model that exploit task-specific characteristics. The resulting model, dd-utt, achieves state-of-the-art results on the four datasets in the CODI-CRAC 2021 shared task.

cs.CL↗

arXiv2020

SDOD:Real-time Segmenting and Detecting 3D Object by Depth

Most existing instance segmentation methods only focus on improving performance and are not suitable for real-time scenes such as autonomous driving. This paper proposes a real-time framework that segmenting and detecting 3D objects by depth. The framework is composed of two parallel branches: one for instance segmentation and another for object detection. We discretize the objects' depth into depth categories and transform the instance segmentation task into a pixel-level classification task. The Mask branch predicts pixel-level depth categories, and the 3D branch indicates instance-level depth categories. We produce an instance mask by assigning pixels which have the same depth categories to each instance. In addition, to solve the imbalance between mask labels and 3D labels in the KITTI dataset, we introduce a coarse mask generated by the auto-annotation model to increase samples. Experiments on the challenging KITTI dataset show that our approach outperforms LklNet about 1.8 times on the speed of segmentation and 3D detection.

cs.CV↗

arXiv2020

Segmenting Epipolar Line

Identifying feature correspondence between two images is a fundamental procedure in three-dimensional computer vision. Usually the feature search space is confined by the epipolar line. Using the cheirality constraint, this paper finds that the feature search space can be restrained to one of two or three segments of the epipolar line that are defined by the epipole and a so-called virtual infinity point.

cs.CV↗

arXiv2020

Clue: Cross-modal Coherence Modeling for Caption Generation

We use coherence relations inspired by computational models of discourse to study the information needs and goals of image captioning. Using an annotation protocol specifically devised for capturing image--caption coherence relations, we annotate 10,000 instances from publicly-available image--caption pairs. We introduce a new task for learning inferences in imagery and text, coherence relation prediction, and show that these coherence annotations can be exploited to learn relation classifiers as an intermediary step, and also train coherence-aware, controllable image captioning models. The results show a dramatic improvement in the consistency and quality of the generated captions with respect to information needs specified via coherence relations.

cs.CL↗