arXiv subjects

Peng Gao

Publications and source records attributed to Peng Gao.

At least 91 records · Page 5Linked to original sources

arXiv2025

CAML: Collaborative Auxiliary Modality Learning for Multi-Agent Systems

Multi-modal learning has emerged as a key technique for improving performance across domains such as autonomous driving, robotics, and reasoning. However, in certain scenarios, particularly in resource-constrained environments, some modalities available during training may be absent during inference. While existing frameworks effectively utilize multiple data sources during training and enable inference with reduced modalities, they are primarily designed for single-agent settings. This poses a critical limitation in dynamic environments such as connected autonomous vehicles (CAV), where incomplete data coverage can lead to decision-making blind spots. Conversely, some works explore multi-agent collaboration but without addressing missing modality at test time. To overcome these limitations, we propose Collaborative Auxiliary Modality Learning (CAML), a novel multi-modal multi-agent framework that enables agents to collaborate and share multi-modal data during training, while allowing inference with reduced modalities during testing. Experimental results in collaborative decision-making for CAV in accident-prone scenarios demonstrate that CAML achieves up to a 58.1% improvement in accident detection. Additionally, we validate CAML on real-world aerial-ground robot data for collaborative semantic segmentation, achieving up to a 10.6% improvement in mIoU.

cs.RO↗

arXiv2025

Distinctive Feature Codec: An Adaptive Efficient Speech Representation for Depression Detection

Large Language Models (LLMs) have demonstrated remarkable success across diverse fields, establishing a powerful paradigm for complex information processing. This has inspired the integration of speech into LLM frameworks, often by tokenizing continuous audio via neural speech codecs, enabling powerful speech language models. However, this dominant tokenization strategy relies on uniform frame-based processing at fixed time intervals. This fixed-rate approach, while effective for linguistic content, destroys the temporal dynamics. These dynamics are not noise but are established as primary biomarkers in clinical applications such as depression detection. To address this gap, we introduce the Distinctive Feature Codec (DFC), an adaptive framework engineered to preserve this vital timing information. Drawing from linguistic theory, DFC abandons fixed-interval processing and instead learns to dynamically segment the signal at perceptually significant acoustic transitions. This generates variable-length tokens that efficiently encode the temporal structure. As a key contribution, this work is the first to integrate traditional distinctive features into a modern deep learning codec for a temporally sensitive task such as depression detection. We also introduce the Group-wise Scalar Quantization (GSQ) approach to stably quantize these variable-length segments. Our distinctive feature-based approach offers a promising alternative to conventional frame-based processing and advances interpretable representation learning in the modern deep learning speech depression detection framework.

eess.AS↗

arXiv2025

A Dual-Branch Local-Global Framework for Cross-Resolution Land Cover Mapping

Cross-resolution land cover mapping aims to produce high-resolution semantic predictions from coarse or low-resolution supervision, yet the severe resolution mismatch makes effective learning highly challenging. Existing weakly supervised approaches often struggle to align fine-grained spatial structures with coarse labels, leading to noisy supervision and degraded mapping accuracy. To tackle this problem, we propose DDTM, a dual-branch weakly supervised framework that explicitly decouples local semantic refinement from global contextual reasoning. Specifically, DDTM introduces a diffusion-based branch to progressively refine fine-scale local semantics under coarse supervision, while a transformer-based branch enforces long-range contextual consistency across large spatial extents. In addition, we design a pseudo-label confidence evaluation module to mitigate noise induced by cross-resolution inconsistencies and to selectively exploit reliable supervisory signals. Extensive experiments demonstrate that DDTM establishes a new state-of-the-art on the Chesapeake Bay benchmark, achieving 66.52\% mIoU and substantially outperforming prior weakly supervised methods. The code is available at https://github.com/gpgpgp123/DDTM.

cs.CV↗

arXiv2025

Upper Bounds for low moments of twisted Fourier coefficients of modular forms

For any large prime $q$, $1 \leq x \leq q$ and any real $0 \leq k \leq 1$, we prove an upper bound for the following $2k$-th moment $$\displaystyle \sum_{\substack{χ\bmod q}} \Big| \sum_{n\leq x} χ(n)λ(n)\Big|^{2k},$$ where $λ(n)$ denotes the Fourier coefficients of a fixed modular form. In particular, our result implies that $$\displaystyle \frac 1{q-1}\sum_{\substack{χ\bmod q}} \Big| \sum_{n\leq x} χ(n)λ(n)\Big|= o(\sqrt{x}),$$ when both $x$ and $q/x$ tend to infinity with $q$.

math.NT↗

arXiv2025

First moment of central values of Hecke $L$-functions with Fixed Order Characters

We evaluate asymptotically the smoothed first moment of central values of families of quadratic, cubic, quartic and sextic Hecke $L$-functions over various imaginary quadratic number fields of class number one, using the method of double Dirichlet series. In particular, we obtain asymptotic formulas for the quadratic families with error terms of size $O(X^{1/4+\varepsilon})$ under the generalized Riemann hypothesis.

math.NT↗

arXiv2025

Decoupled DMD: CFG Augmentation as the Spear, Distribution Matching as the Shield

Diffusion model distillation has emerged as a powerful technique for creating efficient few-step and single-step generators. Among these, Distribution Matching Distillation (DMD) and its variants stand out for their impressive performance, which is widely attributed to their core mechanism of matching the student's output distribution to that of a pre-trained teacher model. In this work, we challenge this conventional understanding. Through a rigorous decomposition of the DMD training objective, we reveal that in complex tasks like text-to-image generation, where CFG is typically required for desirable few-step performance, the primary driver of few-step distillation is not distribution matching, but a previously overlooked component we identify as CFG Augmentation (CA). We demonstrate that this term acts as the core ``engine'' of distillation, while the Distribution Matching (DM) term functions as a ``regularizer'' that ensures training stability and mitigates artifacts. We further validate this decoupling by demonstrating that while the DM term is a highly effective regularizer, it is not unique; simpler non-parametric constraints or GAN-based objectives can serve the same stabilizing function, albeit with different trade-offs. This decoupling of labor motivates a more principled analysis of the properties of both terms, leading to a more systematic and in-depth understanding. This new understanding further enables us to propose principled modifications to the distillation process, such as decoupling the noise schedules for the engine and the regularizer, leading to further performance gains. Notably, our method has been adopted by the Z-Image ( https://github.com/Tongyi-MAI/Z-Image ) project to develop a top-tier 8-step image generation model, empirically validating the generalization and robustness of our findings.

cs.CV↗

arXiv2025

Strongly confined Mid-infrared to Terahertz Phonon Polaritons in Ultra-thin SrTiO3

Phonon polaritons (PhPs) enable subwavelength light control for infrared sensing, imaging, and optoelectronics, but conventional polar materials have narrow Reststrahlen bands, limiting applications. Materials that support PhPs with broad spectral range, strong field confinement, slow group velocity, and high quality factor are therefore needed. Here, using monochromatic electron energy loss spectroscopy in a scanning transmission electron microscope, we demonstrate that ultra-thin SrTiO3 membranes possess the desired properties. Systematic measurements across varying thicknesses reveal two PhP branches with wide spectral dispersion, strong confinement, and anomalously slow group velocities spanning from the mid-infrared to terahertz range. Notably, in 3-nm-thick membranes, these polaritons exhibit unprecedented confinement factors exceeding 500 and group velocities as low as ~ 7 x 10-5c, rivaling the best-performing van der Waals materials. These findings establish perovskite oxide such as SrTiO3 as versatile platforms for tailoring light-matter interactions at the nanoscale, providing critical insights for the design of next-generation photonic devices requiring broadband operation and enhanced optical confinement.

physics.optics↗

arXiv2025

EnerVerse: Envisioning Embodied Future Space for Robotics Manipulation

We introduce EnerVerse, a generative robotics foundation model that constructs and interprets embodied spaces. EnerVerse employs a chunk-wise autoregressive video diffusion framework to predict future embodied spaces from instructions, enhanced by a sparse context memory for long-term reasoning. To model the 3D robotics world, we adopt a multi-view video representation, providing rich perspectives to address challenges like motion ambiguity and 3D grounding. Additionally, EnerVerse-D, a data engine pipeline combining generative modeling with 4D Gaussian Splatting, forms a self-reinforcing data loop to reduce the sim-to-real gap. Leveraging these innovations, EnerVerse translates 4D world representations into physical actions via a policy head (EnerVerse-A), achieving state-of-the-art performance in both simulation and real-world tasks. For efficiency, EnerVerse-A reuses features from the first denoising step and predicts action chunks, achieving about 280 ms per 8-step action chunk on a single RTX 4090. Further video demos, dataset samples could be found in our project page.

cs.RO↗

arXiv2025

Remember Me: Bridging the Long-Range Gap in LVLMs with Three-Step Inference-Only Decay Resilience Strategies

Large Vision-Language Models (LVLMs) have achieved impressive performance across a wide range of multimodal tasks. However, they still face critical challenges in modeling long-range dependencies under the usage of Rotary Positional Encoding (ROPE). Although it can facilitate precise modeling of token positions, it induces progressive attention decay as token distance increases, especially with progressive attention decay over distant token pairs, which severely impairs the model's ability to remember global context. To alleviate this issue, we propose inference-only Three-step Decay Resilience Strategies (T-DRS), comprising (1) Semantic-Driven DRS (SD-DRS), amplifying semantically meaningful but distant signals via content-aware residuals, (2) Distance-aware Control DRS (DC-DRS), which can purify attention by smoothly modulating weights based on positional distances, suppressing noise while preserving locality, and (3) re-Reinforce Distant DRS (reRD-DRS), consolidating the remaining informative remote dependencies to maintain global coherence. Together, the T-DRS recover suppressed long-range token pairs without harming local inductive biases. Extensive experiments on Vision Question Answering (VQA) benchmarks demonstrate that T-DRS can consistently improve performance in a training-free manner. The code can be accessed in https://github.com/labixiaoq-qq/Remember-me

cs.CV↗

arXiv2025

Lower Bounds on High Moments of Twisted Fourier coefficients of Modular Forms

For any large prime $q$, $x \leq 1$ and any real $k\geq 2$, we prove a lower bound for the following $2k$-th moment \begin{equation*} \sum_{\substack{χ\in X_q^*}} \Big| \sum_{n\leq x} χ(n)λ(n)\Big|^{2k}, \end{equation*} where $X_q^*$ denotes the set of primitive Dirichlet characters modulo $q$ and $λ(n)$ the Fourier coefficients of a fixed modular form. The bound we obtain is sharp up to a constant factor under the generalized Riemann Hypothesis.

math.NT↗

arXiv2025

From Ferromagnet to Antiferromagnet: Dimensional Crossover in (111) SrRuO3 Ultrathin Films

SrRuO3 is a canonical itinerant ferromagnet, yet its properties in the extreme two-dimensional limit on a (111) crystal plane remain largely unexplored. Here, we demonstrate a complete transformation of its ground state driven by dimensional reduction. As the thickness of (111)-oriented SrRuO3 films is reduced to a few unit cells, the system transitions from a metallic ferromagnet to a semiconducting antiferromagnet. This emergent antiferromagnetism is evidenced by a vanishing magnetic remanence and most strikingly, by the appearance of an unconventional twelve-fold anisotropic magnetoresistance. First-principles calculations confirm that an A-type antiferromagnetic order is the stable ground state in the ultrathin limit. Our findings establish (111) dimensional engineering as a powerful route to manipulate correlated electron states and uncover novel functionalities for antiferromagnetic spintronics.

cond-mat.str-el↗

arXiv2025

The mollified fourth moment of Dirichlet $L$-functions

We prove an asymptotic formula with a power saving error term for the fourth moment of the family of Dirichlet $L$-functions to modulus $q$ mollified by a Dirichlet polynomial of length $q^{\frac1{22}-\ve}$, valid for all moduli $q\not\equiv2 \pmod 4$. This result was previously known only for restricted sets of moduli with smaller power savings. As a special case, when no Dirichlet polynomial is enclosed, this leads to a significant improvement on X. Wu's asymptotic evaluation of the fourth moment of Dirichlet $L$-functions at the cental point.

math.NT↗

arXiv2025

Visionary-R1: Mitigating Shortcuts in Visual Reasoning with Reinforcement Learning

Learning general-purpose reasoning capabilities has long been a challenging problem in AI. Recent research in large language models (LLMs), such as DeepSeek-R1, has shown that reinforcement learning techniques like GRPO can enable pre-trained LLMs to develop reasoning capabilities using simple question-answer pairs. In this paper, we aim to train visual language models (VLMs) to perform reasoning on image data through reinforcement learning and visual question-answer pairs, without any explicit chain-of-thought (CoT) supervision. Our findings indicate that simply applying reinforcement learning to a VLM -- by prompting the model to produce a reasoning chain before providing an answer -- can lead the model to develop shortcuts from easy questions, thereby reducing its ability to generalize across unseen data distributions. We argue that the key to mitigating shortcut learning is to encourage the model to interpret images prior to reasoning. Therefore, we train the model to adhere to a caption-reason-answer output format: initially generating a detailed caption for an image, followed by constructing an extensive reasoning chain. When trained on 273K CoT-free visual question-answer pairs and using only reinforcement learning, our model, named Visionary-R1, outperforms strong multimodal models, such as GPT-4o, Claude3.5-Sonnet, and Gemini-1.5-Pro, on multiple visual reasoning benchmarks.

cs.CV↗

arXiv2025

Spatial Preference Rewarding for MLLMs Spatial Understanding

Multimodal large language models~(MLLMs) have demonstrated promising spatial understanding capabilities, such as referencing and grounding object descriptions. Despite their successes, MLLMs still fall short in fine-grained spatial perception abilities, such as generating detailed region descriptions or accurately localizing objects. Additionally, they often fail to respond to the user's requirements for desired fine-grained spatial understanding. This issue might arise because existing approaches primarily focus on tuning MLLMs to model pre-annotated instruction data to inject spatial knowledge, without direct supervision of MLLMs' actual responses. We address this issue by SPR, a Spatial Preference Rewarding~(SPR) approach that enhances MLLMs' spatial capabilities by rewarding MLLMs' detailed responses with precise object localization over vague or inaccurate responses. With randomly selected image regions and region descriptions from MLLMs, SPR introduces semantic and localization scores to comprehensively evaluate the text quality and localization quality in MLLM-generated descriptions. We also refine the MLLM descriptions with better localization accuracy and pair the best-scored refinement with the initial descriptions of the lowest score for direct preference optimization, thereby enhancing fine-grained alignment with visual input. Extensive experiments over standard referring and grounding benchmarks show that SPR improves MLLM spatial understanding capabilities effectively with minimal overhead in training. Data and code will be released at https://github.com/hanqiu-hq/SPR

cs.CV↗

arXiv2025

Inverse obstacle scattering with a single moving emitter

This paper is concerned with time domain forward scattering and inverse scattering problems with a single moving point source as the emitter. Approximate solutions are provided for the forward scattering problem with a moving emitter. Regarding the inverse problem, in addition to a basic indicator function based on the approximate solutions, a novel indicator function is developed to construct the direct sampling method to recover both point-like and extended scatterers. Numerical experiments demonstrate that the proposed algorithms are effective in reconstructing both two-dimensional and three-dimensional scatterers with a single moving emitter.

math.NA↗

arXiv2025

ACPO: Adaptive Curriculum Policy Optimization for Aligning Vision-Language Models in Complex Reasoning

Aligning large-scale vision-language models (VLMs) for complex reasoning via reinforcement learning is often hampered by the limitations of existing policy optimization algorithms, such as static training schedules and the rigid, uniform clipping mechanism in Proximal Policy Optimization (PPO). In this work, we introduce Adaptive Curriculum Policy Optimization (ACPO), a novel framework that addresses these challenges through a dual-component adaptive learning strategy. First, ACPO employs a dynamic curriculum that orchestrates a principled transition from a stable, near on-policy exploration phase to an efficient, off-policy exploitation phase by progressively increasing sample reuse. Second, we propose an Advantage-Aware Adaptive Clipping (AAAC) mechanism that replaces the fixed clipping hyperparameter with dynamic, sample-wise bounds modulated by the normalized advantage of each token. This allows for more granular and robust policy updates, enabling larger gradients for high-potential samples while safeguarding against destructive ones. We conduct extensive experiments on a suite of challenging multimodal reasoning benchmarks, including MathVista, LogicVista, and MMMU-Pro. Results demonstrate that ACPO consistently outperforms strong baselines such as DAPO and PAPO, achieving state-of-the-art performance, accelerated convergence, and superior training stability.

cs.AI↗

arXiv2025

Adaptive Conformal Guidance for Learning under Uncertainty

Learning with guidance has proven effective across a wide range of machine learning systems. Guidance may, for example, come from annotated datasets in supervised learning, pseudo-labels in semi-supervised learning, and expert demonstration policies in reinforcement learning. However, guidance signals can be noisy due to domain shifts and limited data availability and may not generalize well. Blindly trusting such signals when they are noisy, incomplete, or misaligned with the target domain can lead to degraded performance. To address these challenges, we propose Adaptive Conformal Guidance (AdaConG), a simple yet effective approach that dynamically modulates the influence of guidance signals based on their associated uncertainty, quantified via split conformal prediction (CP). By adaptively adjusting to guidance uncertainty, AdaConG enables models to reduce reliance on potentially misleading signals and enhance learning performance. We validate AdaConG across diverse tasks, including knowledge distillation, semi-supervised image classification, gridworld navigation, and autonomous driving. Experimental results demonstrate that AdaConG improves performance and robustness under imperfect guidance, e.g., in gridworld navigation, it accelerates convergence and achieves over $6\times$ higher rewards than the best-performing baseline. These results highlight AdaConG as a broadly applicable solution for learning under uncertainty.

cs.LG↗

arXiv2025

Reconstruction of acoustic sources from the initial arrival time of waves

In this paper, a novel time domain sampling method based on the initial arrival time of waves is proposed to reconstruct acoustic sources, including point sources, curve sources, surface sources and block sources. The uniqueness of reconstructing sources whose spatial support is a convex region is proved. Theoretical analyses are provided to demonstrate the validity of the proposed sampling method in reconstructing various types of sound sources. The proposed algorithm does not involve the time integral, exhibits high computational efficiency, and demonstrates strong noise resistance. Numerical experiments are conducted to show the effectiveness of the proposed method.

math-ph↗