Searcharxiv⌕ Search

arXiv subjects

Qi Wang

Publications and source records attributed to Qi Wang.

At least 127 records · Page 7Linked to original sources
arXiv2026

LoRA in LoRA: Towards Parameter-Efficient Architecture Expansion for Continual Visual Instruction Tuning

Continual Visual Instruction Tuning (CVIT) enables Multimodal Large Language Models (MLLMs) to incrementally learn new tasks over time. However, this process is challenged by catastrophic forgetting, where performance on previously learned tasks deteriorates as the model adapts to new ones. A common approach to mitigate forgetting is architecture expansion, which introduces task-specific modules to prevent interference. Yet, existing methods often expand entire layers for each task, leading to significant parameter overhead and poor scalability. To overcome these issues, we introduce LoRA in LoRA (LiLoRA), a highly efficient architecture expansion method tailored for CVIT in MLLMs. LiLoRA shares the LoRA matrix A across tasks to reduce redundancy, applies an additional low-rank decomposition to matrix B to minimize task-specific parameters, and incorporates a cosine-regularized stability loss to preserve consistency in shared representations over time. Extensive experiments on a diverse CVIT benchmark show that LiLoRA consistently achieves superior performance in sequential task learning while significantly improving parameter efficiency compared to existing approaches. The code is available at https://github.com/chanceche/LiLoRA.

cs.CV↗
arXiv2026

$ϕ$-Balancing for Mixture-of-Experts Training

Mixture-of-Experts (MoE) models rely on balanced expert utilization to fully realize their scalability. However, existing load-balancing methods are largely heuristic and operate on noisy mini-batch assignment statistics, introducing bias relative to population-level objectives. We propose $ϕ$-balancing, a principled framework that directly targets population-level expert balance by minimizing a strictly convex, symmetric, and differentiable potential of the expected routing distribution. Using convex duality, we derive an equivalent min-max formulation and obtain a simple online algorithm via mirror descent, yielding an efficient EMA-based routing adjustment with negligible overhead. Across large-scale pretraining and downstream fine-tuning, $ϕ$-balancing consistently outperforms prior Switch-style and loss-free baselines, demonstrating more stable and effective expert utilization.

cs.LG↗
arXiv2026

Collaborative Parameter Learning: Mitigating Forgetting via Parameter-Level Gradient Analysis

Catastrophic forgetting during knowledge injection impairs the ability of large language models to acquire new knowledge without overwriting previously mastered knowledge. Recent studies analyze forgetting from a gradient similarity perspective and mitigate forgetting through vector projection. However, these methods primarily characterize gradient similarity at the aggregate direction level, leaving the parameter wise contributions to forgetting underexplored. In this paper, we decompose gradient similarity into parameter wise contributions and identify two types of parameters during forgetting: Conflicting Parameters, whose updates contribute to forgetting and typically account for 50 percent to 75 percent of parameters, and Collaborative Parameters, whose updates mitigate forgetting and account for 25 percent to 50 percent. Based on this analysis, we propose Collaborative Parameter Learning, CPL, a parameter wise training rule that freezes Conflicting Parameters and updates only Collaborative Parameters. Experiments comparing CPL with seven baseline methods show that CPL learns 20.2% to 48.2% more questions with negligible forgetting, while reducing peak VRAM by approximately 3 GB per billion model parameters and computation time by 16.5 percent. Extensive evaluations on parameter consumption, out of set generalization, cross prompt generalization, multimodal tasks, open ended question answering, and multilingual settings demonstrate that CPL effectively mitigates forgetting across diverse scenarios.

cs.LG↗
arXiv2026

OrbiSim: World Models as Differentiable Physics Engines for Embodied Intelligence

We present OrbiSim, a novel robotic simulation paradigm that redefines world models as a fully differentiable physics engine for embodied intelligence. Unlike prior world models that focus on unconstrained imagination in latent or visual domains, OrbiSim establishes a unified, physically-grounded pathway that bridges structured scene assets, neural dynamics, and downstream reinforcement learning. By enabling end-to-end differentiability throughout the entire simulation loop -- spanning from explicit state transitions to visual observation generation -- OrbiSim supports tasks traditionally intractable for classical simulators, such as differentiable contact modeling, gradient-based policy optimization under sparse rewards, and intuitive physical inference. Empirical results demonstrate that OrbiSim significantly outperforms state-of-the-art world models in both predictive fidelity and control performance. Furthermore, its consistent responsiveness to asset configurations and physical parameters suggests its potential as a differentiable tool for enhancing robot simulation and policy training.

cs.RO↗
arXiv2026

A Breast Vision Pathology Foundation Model for Real-world Clinical Utility

Pathology foundation models have shown strong retrospective performance, but whether such systems can support clinically relevant use remains unclear. This challenge is particularly important in breast cancer, where pathological assessment serves as the gold standard for diagnosis and guides treatment planning, surgical decision-making and risk stratification across pre-, intra- and post-operative stages. Here we present \textbf{BRAVE}, a breast-adaptive pathology foundation model developed and evaluated using a total resource of 101,638 breast whole-slide images from 32 sources across Asia, Europe and North America. We assessed BRAVE across 34 tasks in 82 cohorts spanning pre-operative biopsy, intra-operative frozen section and post-operative resection, using an evidence chain comprising retrospective benchmarking, clinically challenging scenarios, workflow-oriented clinical impact simulations, prospective observational validation with the thresholds locked in the retrospective cohorts and crossover pathologist-AI interaction studies. Across these settings, BRAVE supported practical roles in the clinical workflow, including safe exclusion of low-risk cases from routine review, AI-assisted second-review rescue of initially missed positives and prioritization of cases for further assessment. In prospective validation across three centres, BRAVE excluded 76.9% of negative biopsy cases (NPV 0.953) and 70.1% of negative frozen-section cases (NPV 0.973), and triaged 78.8% of post-operative subtyping cases as high-confidence clear-cut cases (NPV 1.000). In reader studies, AI assistance improved balanced accuracy from 88.5% to 95.1% (OR 3.14, P<0.001), with better efficiency, confidence and inter-rater agreement. BRAVE-derived scores also independently predicted disease-free survival (adjusted HR 4.79, P<0.001) and overall survival (adjusted HR 8.14, P<0.001).

cs.CV↗
arXiv2026

MoCapAnything: Unified 3D Motion Capture for Arbitrary Skeletons from Monocular Videos

Motion capture now underpins content creation far beyond digital humans, yet most existing pipelines remain species- or template-specific. We formalize this gap as Category-Agnostic Motion Capture (CAMoCap): given a monocular video and an arbitrary rigged 3D asset as a prompt, the goal is to reconstruct a rotation-based animation such as BVH that directly drives the specific asset. We present MoCapAnything, a reference-guided, factorized framework that first predicts 3D joint trajectories and then recovers asset-specific rotations via constraint-aware inverse kinematics. The system contains three learnable modules and a lightweight IK stage: (1) a Reference Prompt Encoder that extracts per-joint queries from the asset's skeleton, mesh, and rendered images; (2) a Video Feature Extractor that computes dense visual descriptors and reconstructs a coarse 4D deforming mesh to bridge the gap between video and joint space; and (3) a Unified Motion Decoder that fuses these cues to produce temporally coherent trajectories. We also curate Truebones Zoo with 1038 motion clips, each providing a standardized skeleton-mesh-render triad. Experiments on both in-domain benchmarks and in-the-wild videos show that MoCapAnything delivers high-quality skeletal animations and exhibits meaningful cross-species retargeting across heterogeneous rigs, enabling scalable, prompt-driven 3D motion capture for arbitrary assets. Project page: https://animotionlab.github.io/MoCapAnything/

cs.CV↗
arXiv2026

Some Large Language Models Exhibit Consistent Risk Attitudes

As artificial intelligence systems are deployed in open-ended, high-stakes settings, a critical dimension remains unmeasured: how perceived risk is translated into action. We test whether large language models (LLMs) exhibit systematic and consistent risk attitudes under uncertainty. We introduce a cross-domain framework that decouples contextual risk belief from categorical decision, and apply it to six representative LLMs and 100 human participants across spatial navigation, clinical triage, and financial allocation tasks. Using regression models, we extract each agents belief-to-decision mapping and quantify risk sensitivity and risk attitude bias. We find that most tested LLMs exhibit (i) robust intra-task consistency, indicating stable mappings from contextual belief to risk decision within a fixed task domain; (ii) cross-domain rank-order stability, preserving relative risk posture across tasks; and (iii) a convergence toward a restricted risk-attitude distribution relative to the broader human baseline. These results reveal risk attitude as a stable and previously uncharacterized dimension of LLM behavior, establishing a foundation for evaluating and aligning AI systems in open-ended decision-making and motivating further investigation into the origins of these intrinsic behavioral dispositions.

cs.AI↗
arXiv2026

Agentic Artificial Intelligence in Finance: A Comprehensive Survey

The emergence of agentic artificial intelligence (AI) represents a fundamental transformation in financial markets, characterized by autonomous systems capable of reasoning, planning, and adaptive decision-making with minimal human intervention. This comprehensive survey synthesizes recent advances in agentic AI across multiple dimensions of financial operations, including system architecture, market applications, regulatory frameworks, and systemic implications. We examine how agentic AI differs from traditional algorithmic trading and generative AI through its capacity for goal-oriented autonomy, continuous learning, and multi-agent coordination. Our analysis shows that while agentic AI offers substantial potential for enhanced market efficiency, liquidity provision, and risk management, it also introduces novel challenges related to market stability, regulatory compliance, interpretability, and systemic risk. Through a systematic review of foundational research, technical architectures, market applications, and governance frameworks, this survey provides scholars and practitioners with a structured understanding of how agentic AI is reshaping financial markets and identifies critical research directions for ensuring that these systems enhance both operational efficiency and market resilience.

econ.EM↗
arXiv2026

True random number generation through stochastic magnonic bistability

True random number generators (TRNGs) underpin modern cryptography, yet existing implementations face fundamental trade-offs between speed, scalability, and entropy quality. Here, we demonstrate that stochastic switching in the bistable regime of spin-wave dynamics provides a physical entropy source for high-quality random number generation. Our magnonic random number generator (mRNG), based on a lithography-patterned microstrip on yttrium iron garnet (YIG), exploits thermal fluctuations near the nonlinear bistable regime to generate random bitstreams that pass all 15 NIST SP 800-22 statistical tests at rates with 20 Mb/s. We implement a random-bit multiplier using synchronized mRNG units and demonstrate scalability to 200-nm-wide nanoscale waveguides, establishing spin-wave bistability as a viable physical entropy source for integrated random number generation.

cond-mat.mtrl-sci↗
arXiv2026

Planar Gaussian Splatting with Bilinear Spatial Transformer for Wireless Radiance Field Reconstruction

Wireless radiance field (WRF) reconstruction aims to learn a continuous, queryable representation of radio frequency characteristics over 3D space and direction, from which specific quantities, such as the spatial power spectrum (SPS) at a receiver given a transmitter position, can be predicted. While Gaussian splatting (GS)-based method has surpassed Neural Radiance Fields (NeRF)-based method for this task, existing adaptations largely transplant vision pipelines, limiting physical interpretability and accuracy. We introduce BiSplat-WRF, a planar GS framework that retains the expressiveness of 3D GS while removing unnecessary projections and incorporating global EM coupling and mutual scattering among primitives. Each primitive is a 2D planar Gaussian with 3D coordinates, rendered directly on the angular domain of the SPS. A bilinear spatial transformer (BST) aggregates inter-primitive relations on an angular grid and, via attention, captures long-range electromagnetic dependencies, thereby enforcing globally aware EM interactions that reflect the complex physics of the wireless environment. On spatial spectrum synthesis task, BiSplat-WRF surpasses NeRF-based and prior GS-based baselines with respect to the Structural Similarity Index (SSIM); comprehensive ablation studies validate the contribution of BST. We also provide a larger BiSplat-WRF+ variant that further increases SSIM at a higher computation cost, serving as a strong reference for future studies.

eess.SP↗
arXiv2026

Judge Like Human Examiners: A Weighted Importance Multi-Point Evaluation Framework for Generative Tasks with Long-form Answers

Evaluating the quality of model responses remains challenging in generative tasks with long-form answers, as the expected answers usually contain multiple semantically distinct yet complementary factors that should be factorized for fine-grained assessment. Recent evaluation methods resort to relying on either task-level rubrics or question-aware checklists. However, they still 1) struggle to assess whether a response is genuinely grounded in provided contexts; 2) fail to capture the heterogeneous importance of different aspects of reference answers. Inspired by human examiners, we propose a Weighted Importance Multi-Point Evaluation (WIMPE) framework, which factorizes each reference answer into weighted context-bound scoring points. Two complementary metrics, namely Weighted Point-wise Alignment (WPA) and Point-wise Conflict Penalty (PCP), are designed to measure the alignment and contradiction between model responses and reference answers. Extensive experiments on 10 generative tasks demonstrate that WIMPE achieves higher correlations with human annotations.

cs.CL↗
arXiv2026

JoyAI-LLM Flash: Advancing Mid-Scale LLMs with Token Efficiency

We introduce JoyAI-LLM Flash, an efficient Mixture-of-Experts (MoE) language model designed to redefine the trade-off between strong performance and token efficiency in the sub-50B parameter regime. JoyAI-LLM Flash is pretrained on a massive corpus of 20 trillion tokens and further optimized through a rigorous post-training pipeline, including supervised fine-tuning (SFT), Direct Preference Optimization (DPO), and large-scale reinforcement learning (RL) across diverse environments. To improve token efficiency, JoyAI-LLM Flash strategically balances \emph{thinking} and \emph{non-thinking} cognitive modes and introduces FiberPO, a novel RL algorithm inspired by fibration theory that decomposes trust-region maintenance into global and local components, providing unified multi-scale stability control for LLM policy optimization. To enhance architectural sparsity, the model comprises 48B total parameters while activating only 2.7B parameters per forward pass, achieving a substantially higher sparsity ratio than contemporary industry leading models of comparable scale. To further improve inference throughput, we adopt a joint training-inference co-design that incorporates dense Multi-Token Prediction (MTP) and Quantization-Aware Training (QAT). We release the checkpoints for both JoyAI-LLM-48B-A3B Base and its post-trained variants on Hugging Face to support the open-source community.

cs.CL↗
arXiv2026

LLM2Manim: Pedagogy-Aware AI Generation of STEM Animations

High-quality STEM animations can be useful for learning, but they are still not common in daily teaching, mostly because they take time and special skills to make. In this paper, we present a semi-automated, human-in-the-loop (HITL) pipeline that uses a large language model (LLM) to help convert math and physics concepts into narrated animations with the Python library Manim. The pipeline also tries to follow multimedia learning ideas like segmentation, signaling, and dual coding, so the narration and the visuals are more aligned. To keep the outputs stable, we use constrained prompt templates, a symbol ledger to keep symbols consistent, and we regenerate only the parts that have errors. We also include expert review before the final rendering, because sometimes the generated code or explanation is not fully correct. We tested the approach with 100 undergraduate students in a within-subject A-B study. Each student learned two similar STEM topics, one with the LLM-generated animations and one with PowerPoint slides. In general, the animation-based instruction gives slightly better post-test scores (83% vs.78%, p < .001), and students show higher learning gains (d=0.67). They also report higher engagement (d=0.94) and lower cognitive load (d=0.41). Students finished the tasks faster, and many of them said they prefer the animated format. Overall, these results suggest LLM-assisted animation can make STEM content creation easier, and it may be a practical option for more classrooms.

cs.MM↗
arXiv2026

Batch Loss Score for Dynamic Data Pruning

Dynamic data pruning accelerates deep learning by selectively omitting less informative samples during training. While per-sample loss is a common importance metric, obtaining it can be challenging or infeasible for complex models or loss functions, often requiring significant implementation effort. This work proposes the Batch Loss Score (BLS), a computationally efficient alternative using an Exponential Moving Average (EMA) of readily available batch losses to assign scores to individual samples. We frame the batch loss, from the perspective of a single sample, as a noisy measurement of its scaled individual loss, with noise originating from stochastic batch composition. It is formally shown that the EMA mechanism functions as a first-order low-pass filter, attenuating high-frequency batch composition noise. This yields a score approximating the smoothed and persistent contribution of the individual sample to the loss, providing a theoretical grounding for BLS as a proxy for sample importance. BLS demonstrates remarkable code integration simplicity (\textbf{three-line injection}) and readily adapts existing per-sample loss-based methods (\textbf{one-line proxy}). Its effectiveness is demonstrated by enhancing two such methods to losslessly prune \textbf{20\%-50\%} of samples across \textit{14 datasets}, \textit{11 tasks} and \textit{18 models}, highlighting its utility and broad applicability, especially for complex scenarios where per-sample loss is difficult to access. Code is available at https://github.com/mrazhou/BLS.

cs.LG↗
arXiv2026

Disentangled World Models: Learning to Transfer Semantic Knowledge from Distracting Videos for Reinforcement Learning

Training visual reinforcement learning (RL) in practical scenarios presents a significant challenge, $\textit{i.e.,}$ RL agents suffer from low sample efficiency in environments with variations. While various approaches have attempted to alleviate this issue by disentangled representation learning, these methods usually start learning from scratch without prior knowledge of the world. This paper, in contrast, tries to learn and understand underlying semantic variations from distracting videos via offline-to-online latent distillation and flexible disentanglement constraints. To enable effective cross-domain semantic knowledge transfer, we introduce an interpretable model-based RL framework, dubbed Disentangled World Models (DisWM). Specifically, we pretrain the action-free video prediction model offline with disentanglement regularization to extract semantic knowledge from distracting videos. The disentanglement capability of the pretrained model is then transferred to the world model through latent distillation. For finetuning in the online environment, we exploit the knowledge from the pretrained model and introduce a disentanglement constraint to the world model. During the adaptation phase, the incorporation of actions and rewards from online environment interactions enriches the diversity of the data, which in turn strengthens the disentangled representation learning. Experimental results validate the superiority of our approach on various benchmarks.

cs.CV↗
arXiv2026

Goal-Driven Reward by Video Diffusion Models for Reinforcement Learning

Reinforcement Learning (RL) has achieved remarkable success in various domains, yet it often relies on carefully designed programmatic reward functions to guide agent behavior. Designing such reward functions can be challenging and may not generalize well across different tasks. To address this limitation, we leverage the rich world knowledge contained in pretrained video diffusion models to provide goal-driven reward signals for RL agents without ad-hoc design of reward. Our key idea is to exploit off-the-shelf video diffusion models pretrained on large-scale video datasets as informative reward functions in terms of video-level and frame-level goals. For video-level rewards, we first finetune a pretrained video diffusion model on domain-specific datasets and then employ its video encoder to evaluate the alignment between the latent representations of agent's trajectories and the generated goal videos. To enable more fine-grained goal-achievement, we derive a frame-level goal by identifying the most relevant frame from the generated video using CLIP, which serves as the goal state. We then employ a learned forward-backward representation that represents the probability of visiting the goal state from a given state-action pair as frame-level reward, promoting more coherent and goal-driven trajectories. Experiments on Meta-World and Distracting Control Suite demonstrate the effectiveness of our approach.

cs.LG↗
arXiv2026

Differential Mental Disorder Detection with Psychology-Inspired Multimodal Stimuli

Differential diagnosis of mental disorders remains a fundamental challenge in real-world clinical practice, where multiple conditions often exhibit overlapping symptoms. However, most existing public datasets are developed under single-disorder settings and rely on limited data elicitation paradigms, restricting their ability to capture disorder-specific patterns. In this work, we investigate differential mental disorder detection through psychology-inspired multimodal stimuli, designed to elicit diverse emotional, cognitive, and behavioral responses grounded in findings from experimental psychology. Based on this paradigm, we collect a large-scale multimodal mental health dataset (MMH) covering depression, anxiety, and schizophrenia, with all diagnostic labels clinically verified by licensed psychiatrists. To effectively model the heterogeneous signals induced by diverse elicitation tasks, we further propose a paradigm-aware multimodal framework that leverages inter-disorder differences prior knowledge as prompt-guided semantic descriptions to capture task-specific affective and interaction contexts for multimodal representation learning in the new differential mental disorder detection task. Extensive experiments show that our framework consistently outperforms existing baselines, underscoring the value of psychology-inspired stimulus design for differential mental disorder detection.

cs.MM↗
arXiv2026

Swarm-Based Inertial Methods for Optimization

We introduce a new class of swarm-based inertial methods (SBIMs) for global minimization, formulated as coupled dissipative inertial dynamical systems derived from the generalized Onsager principle. The proposed framework identifies the friction operator and the scaling of the potential energy, namely the objective function to be minimized, as the key ingredients governing relaxation dynamics over the energy landscape. Within this framework, we propose a new underdamped inertial dynamics whose damping mechanisms incorporate both gradient and Hessian information, allowing the system to adjust damping or acceleration according to the agent trajectories and the curvature of the landscape. Under suitable conditions, we prove that the underdamped system satisfies an energy dissipation law, from which we establish an upper bound on the asymptotic decay rate of the gap between the objective function and its global minimum, given by $O(1/δ(t))$ (defined in §3). We further construct structure-preserving discretizations that retain both discrete energy dissipation and the convergence rate estimate, $O(1/δ_k)$ (defined in \S3). In addition, we present several other efficient numerical algorithms for the dynamical system. Numerical experiments for all proposed algorithms validate the theory on convex test problems and demonstrate convergence rates in function values that are substantially faster than the theoretical guarantees ($O(1/δ_k)$). On nonconvex benchmark problems, the proposed methods achieve high success rates in reaching the global minimum, and exhibit more stable energy decay than swarm-based gradient descent and Nesterov methods. Overall, this work provides a systematic framework for the construction and analysis of SBIMs from an energy-dissipative perspective.

math.OC↗