arXiv subjects

Yikun Wang

Publications and source records attributed to Yikun Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

GeometryZero: Advancing Geometry Solving via Group Contrastive Policy Optimization

Recent progress in large language models (LLMs) has boosted mathematical reasoning, yet geometry remains challenging where auxiliary construction is often essential. Prior methods either underperform or depend on very large models (e.g., GPT-4o), making them costly. We argue that reinforcement learning with verifiable rewards (e.g., GRPO) can train smaller models to couple auxiliary construction with solid geometric reasoning. However, naively applying GRPO yields unconditional rewards, encouraging indiscriminate and sometimes harmful constructions. We propose Group Contrastive Policy Optimization (GCPO), an RL framework with two components: (1) Group Contrastive Masking, which assigns positive/negative construction rewards based on contextual utility, and (2) a Length Reward that encourages longer reasoning chains. On top of GCPO, we build GeometryZero, an affordable family of geometry reasoning models that selectively use auxiliary construction. Experiments on Geometry3K and MathVista show GeometryZero consistently outperforms RL baselines (e.g., GRPO, ToRL). The code has been available at https://github.com/ekonwang/GeometryZero.

cs.CL↗

arXiv2026

VideoThinker: Building Agentic VideoLLMs with LLM-Guided Tool Reasoning

Long-form video understanding remains a fundamental challenge for current Video Large Language Models. Most existing models rely on static reasoning over uniformly sampled frames, which weakens temporal localization and leads to substantial information loss in long videos. Agentic tools such as temporal retrieval, spatial zoom, and temporal zoom offer a natural way to overcome these limitations by enabling adaptive exploration of key moments. However, constructing agentic video understanding data requires models that already possess strong long-form video comprehension, creating a circular dependency. We address this challenge with VideoThinker, an agentic Video Large Language Model trained entirely on synthetic tool interaction trajectories. Our key idea is to convert videos into rich captions and employ a powerful agentic language model to generate multi-step tool use sequences in caption space. These trajectories are subsequently grounded back to video by replacing captions with the corresponding frames, yielding a large-scale interleaved video and tool reasoning dataset without requiring any long-form understanding from the underlying model. Training on this synthetic agentic dataset equips VideoThinker with dynamic reasoning capabilities, adaptive temporal exploration, and multi-step tool use. Remarkably, VideoThinker significantly outperforms both caption-only language model agents and strong video model baselines across long-video benchmarks, demonstrating the effectiveness of tool augmented synthetic data and adaptive retrieval and zoom reasoning for long-form video understanding.

cs.CV↗

arXiv2026

Learning to Jointly Optimize Antenna Positioning and Beamforming for Movable Antenna-Aided Systems

The recently emerged movable antenna (MA) and fluid antenna technologies offer promising solutions to enhance the spatial degrees of freedom in wireless systems by dynamically adjusting the positions of transmit or receive antennas within given regions. In this paper, we aim to address the joint optimization problem of antenna positioning and beamforming in MA-aided multi-user downlink transmission systems. This problem involves mixed discrete antenna position and continuous beamforming weight variables, along with coupled distance constraints on antenna positions, which pose significant challenges for optimization algorithm design. To overcome these challenges, we propose an end-to-end deep learning framework, consisting of a positioning model that handles the discrete variables and the coupled constraints, and a beamforming model that handles the continuous variables. Simulation results demonstrate that the proposed framework achieves superior sum rate performance, yet with much reduced computation time compared to existing methods.

eess.SP↗

arXiv2026

NTK-Guided Implicit Neural Teaching

Implicit Neural Representations (INRs) parameterize continuous signals via multilayer perceptrons (MLPs), enabling compact, resolution-independent modeling for tasks like image, audio, and 3D reconstruction. However, fitting high-resolution signals demands optimizing over millions of coordinates, incurring prohibitive computational costs. To address it, we propose NTK-Guided Implicit Neural Teaching (NINT), which accelerates training by dynamically selecting coordinates that maximize global functional updates. Leveraging the Neural Tangent Kernel (NTK), NINT scores examples by the norm of their NTK-augmented loss gradients, capturing both fitting errors and heterogeneous leverage (self-influence and cross-coordinate coupling). This dual consideration enables faster convergence compared to existing methods. Through extensive experiments, we demonstrate that NINT significantly reduces training time by nearly half while maintaining or improving representation quality, establishing state-of-the-art acceleration among recent sampling-based strategies.

cs.LG↗

arXiv2026

HPTune: Hierarchical Proactive Tuning for Collision-Free Model Predictive Control

Parameter tuning is a powerful approach to enhance adaptability in model predictive control (MPC) motion planners. However, existing methods typically operate in a myopic fashion that only evaluates executed actions, leading to inefficient parameter updates due to the sparsity of failure events (e.g., obstacle nearness or collision). To cope with this issue, we propose to extend evaluation from executed to non-executed actions, yielding a hierarchical proactive tuning (HPTune) framework that combines both a fast-level tuning and a slow-level tuning. The fast one adopts risk indicators of predictive closing speed and predictive proximity distance, and the slow one leverages an extended evaluation loss for closed-loop backpropagation. Additionally, we integrate HPTune with the Doppler LiDAR that provides obstacle velocities apart from position-only measurements for enhanced motion predictions, thus facilitating the implementation of HPTune. Extensive experiments on high-fidelity simulator demonstrate that HPTune achieves efficient MPC tuning and outperforms various baseline schemes in complex environments. It is found that HPTune enables situation-tailored motion planning by formulating a safe, agile collision avoidance strategy.

cs.RO↗

arXiv2026

VideoPro: Adaptive Program Reasoning for Long Video Understanding

Large language models (LLMs) have shown promise in generating program workflows for visual tasks. However, previous approaches often rely on closed-source models, lack systematic reasoning, and struggle with long-form video question answering (videoQA). To address these challenges, we introduce the FS-VisPR framework, an adaptive visual program reasoning approach that balances fast reasoning for simple queries with slow reasoning for difficult ones. First, we design efficient visual modules (e.g., key clip retrieval and subtitle retrieval) to support long-form video tasks. Then, we construct a diverse and high-quality fast-slow reasoning dataset with a strong LLM to align open-source language models' ability to generate visual program workflows as FS-LLM. Next, we design a fast-slow reasoning framework with FS-LLM: Simple queries are directly solved by VideoLLMs, while difficult ones invoke visual program reasoning, motivated by human-like reasoning processes. During this process, low-confidence fast-thinking answers will trigger a second-stage slow-reasoning process, and a fallback mechanism to fast reasoning is activated if the program execution fails. Moreover, we improve visual programs through parameter search during both training and inference. By adjusting the parameters of the visual modules within the program, multiple variants are generated: during training, programs that yield correct answers are selected, while during inference, the program with the highest confidence result is applied. Experiments show that FS-VisPR improves both efficiency and reliability in visual program workflows. It achieves 50.4% accuracy on LVBench, surpassing GPT-4o, matching the performance of Qwen2.5VL-72B on VideoMME.

cs.CV↗

arXiv2026

MoIIE: Mixture of Intra- and Inter-Modality Experts for Large Vision Language Models

Large Vision-Language Models (LVLMs) have demonstrated remarkable performance across multi-modal tasks by scaling model size and training data. However, these dense LVLMs incur significant computational costs and motivate the exploration of sparse Mixture of Experts (MoE) architectures. While MoE improve parameter efficiency, effectively applying MoE to simultaneously model modality-specific features and cross-modal associations in LVLMs remains challenging. In this work, we propose to incorporate Mixture of Intra- and Inter-Modality Experts (MoIIE) to LVLMs. For each token, expert routing is guided by its modality, directing tokens to their respective intra-modality experts as well as a shared pool of inter-modality experts, enabling the model to jointly learn rich intra-modal features and cross-modal interactions. We further introduce an effective and straightforward two-stage training strategy, which facilitates the direct activation of both MoE and multi-modal capabilities. Extensive experiments across different data scales and LLM backbone demonstrate the effectiveness, efficiency and generality of our approach. Notably, our MoIIE models with 5.5B and 11.3B activated parameters match or even surpass the performance of existing advanced open-source MoE-LLMs based multi-modal models that involve more activated parameters. The code is available at https://github.com/AlenjandroWang/MoIIE.

cs.CV↗

arXiv2026

Autoregressive Semantic Visual Reconstruction Helps VLMs Understand Better

Typical large vision-language models (LVLMs) apply autoregressive supervision solely to textual sequences, without fully incorporating the visual modality into the learning process. This results in three key limitations: (1) an inability to utilize images without accompanying captions, (2) the risk that captions omit critical visual details, and (3) the challenge that certain vision-centric content cannot be adequately conveyed through text. As a result, current LVLMs often prioritize vision-to-language alignment while potentially overlooking fine-grained visual information. While some prior works have explored autoregressive image generation, effectively leveraging autoregressive visual supervision to enhance image understanding remains an open challenge. In this paper, we introduce Autoregressive Semantic Visual Reconstruction (ASVR), which enables joint learning of visual and textual modalities within a unified autoregressive framework. We show that autoregressively reconstructing the raw visual appearance of images does not enhance and may even impair multimodal understanding. In contrast, autoregressively reconstructing the semantic representation of images consistently improves comprehension. Notably, we find that even when models are given continuous image features as input, they can effectively reconstruct discrete semantic tokens, resulting in stable and consistent improvements across a wide range of multimodal understanding benchmarks. Our approach delivers significant performance gains across varying data scales (556k-2M) and types of LLM bacbones. Specifically, ASVR improves LLaVA-1.5 by 5% in average scores across 14 multimodal benchmarks. The code is available at https://github.com/AlenjandroWang/ASVR.

cs.CV↗

arXiv2025

GeoVista: Web-Augmented Agentic Visual Reasoning for Geolocalization

Current research on agentic visual reasoning enables deep multimodal understanding but primarily focuses on image manipulation tools, leaving a gap toward more general-purpose agentic models. In this work, we revisit the geolocalization task, which requires not only nuanced visual grounding but also web search to confirm or refine hypotheses during reasoning. Since existing geolocalization benchmarks fail to meet the need for high-resolution imagery and the localization challenge for deep agentic reasoning, we curate GeoBench, a benchmark that includes photos and panoramas from around the world, along with a subset of satellite images of different cities to rigorously evaluate the geolocalization ability of agentic models. We also propose GeoVista, an agentic model that seamlessly integrates tool invocation within the reasoning loop, including an image-zoom-in tool to magnify regions of interest and a web-search tool to retrieve related web information. We develop a complete training pipeline for it, including a cold-start supervised fine-tuning (SFT) stage to learn reasoning patterns and tool-use priors, followed by a reinforcement learning (RL) stage to further enhance reasoning ability. We adopt a hierarchical reward to leverage multi-level geographical information and improve overall geolocalization performance. Experimental results show that GeoVista surpasses other open-source agentic models on the geolocalization task greatly and achieves performance comparable to closed-source models such as Gemini-2.5-flash and GPT-5 on most metrics.

cs.CV↗

arXiv2025

Detecting gravitational signatures of dark matter with atom gradiometers

We study the purely gravitational signatures of dark matter from the ultralight to the ultraheavy mass range in proposed long-baseline atom gradiometers, focusing on terrestrial designs, such as AION-km and MAGIS-km, as well as space-based concepts, such as MAGIS-space, AEDGE and AEDGE+. Due to its exceptional acceleration sensitivity and depending on astrophysical backgrounds, a detector similar to AEDGE+ could detect a dark matter subcomponent which constitutes $\mathcal{O}(10\%)$ of the local dark matter energy density and is populated by compact clumps of mass between $10^6$~kg and $10^{10}$~kg ($10^{-25}~M_\odot\lesssim M \lesssim 10^{-21}~M_\odot$) in an otherwise unexplored region of dark matter model space. Furthermore, because the gravitational observable depends on the relative gravitational time delay measured by spatially separated atomic clouds, we find that atom gradiometers are parametrically more sensitive than laser interferometers, such as LIGO and LISA, to fast-oscillating spacetime perturbations sourced by energy density and pressure fluctuations of ultralight dark matter. Depending on astrophysical backgrounds, a detector akin to AEDGE+ could probe a DM overdensity of $\mathcal{O}(10)$ times the local dark matter energy density for masses $m\lesssim 10^{-17}$~eV.

hep-ph↗

arXiv2025

EnvX: Agentize Everything with Agentic AI

The widespread availability of open-source repositories has led to a vast collection of reusable software components, yet their utilization remains manual, error-prone, and disconnected. Developers must navigate documentation, understand APIs, and write integration code, creating significant barriers to efficient software reuse. To address this, we present EnvX, a framework that leverages Agentic AI to agentize GitHub repositories, transforming them into intelligent, autonomous agents capable of natural language interaction and inter-agent collaboration. Unlike existing approaches that treat repositories as static code resources, EnvX reimagines them as active agents through a three-phase process: (1) TODO-guided environment initialization, which sets up the necessary dependencies, data, and validation datasets; (2) human-aligned agentic automation, allowing repository-specific agents to autonomously perform real-world tasks; and (3) Agent-to-Agent (A2A) protocol, enabling multiple agents to collaborate. By combining large language model capabilities with structured tool integration, EnvX automates not just code generation, but the entire process of understanding, initializing, and operationalizing repository functionality. We evaluate EnvX on the GitTaskBench benchmark, using 18 repositories across domains such as image processing, speech recognition, document analysis, and video manipulation. Our results show that EnvX achieves a 74.07% execution completion rate and 51.85% task pass rate, outperforming existing frameworks. Case studies further demonstrate EnvX's ability to enable multi-repository collaboration via the A2A protocol. This work marks a shift from treating repositories as passive code resources to intelligent, interactive agents, fostering greater accessibility and collaboration within the open-source ecosystem.

cs.AI↗

arXiv2025

A New Measure of Assembly Bias using the Environment Dependence of the Luminosity Function

Assembly bias is the variation in the clustering of dark matter halos and galaxies that arises from correlations between the halo assembly history and the large-scale environment at fixed halo mass. In this work, we use the cosmological magneto-hydrodynamical simulation TNG300 to investigate how assembly bias affects the environment-dependent galaxy luminosity function. We measure the luminosity functions in bins of large-scale environment for the original simulated galaxy sample and for a shuffled sample, where the galaxies are randomly reassigned among halos of similar mass to remove assembly bias. By comparing them, we find distinct signatures, showing variations in the number of galaxies at the $\sim10\%$ level across all luminosities. Assembly bias increases the tendency of galaxies to reside in denser environments and further dilutes underdense regions, beyond the trends governed by halo mass. When separating by color, we see that assembly bias has a much bigger effect on red galaxies fainter than $\mathrm{M_r} - 5 \log{h} = -18.5$, which accounts for a $\sim20\%$ increase in the number of galaxies in the densest environment and a remarkable $50\%$ decrease in the least dense regions. The ratio of these measurements for the densest and least dense regions provides a significant assembly bias signal for the faint red galaxies, larger than a factor of two. Overall, our results provide a novel sensitive measure of assembly bias, offering valuable insight for modeling the effect and a potential new route to detect it in observations.

astro-ph.GA↗

arXiv2025

Efficient SRAM-PIM Co-design by Joint Exploration of Value-Level and Bit-Level Sparsity

Processing-in-memory (PIM) is a transformative architectural paradigm designed to overcome the Von Neumann bottleneck. Among PIM architectures, digital SRAM-PIM emerges as a promising solution, offering significant advantages by directly integrating digital logic within the SRAM array. However, rigid crossbar architecture and full array activation pose challenges in efficiently utilizing traditional value-level sparsity. Moreover, neural network models exhibit a high proportion of zero bits within non-zero values, which remain underutilized due to architectural constraints. To overcome these limitations, we present Dyadic Block PIM (DB-PIM), a groundbreaking algorithm-architecture co-design framework to harness both value-level and bit-level sparsity. At the algorithm level, our hybrid-grained pruning technique, combined with a novel sparsity pattern, enables effective sparsity management. Architecturally, DB-PIM incorporates a sparse network and customized digital SRAM-PIM macros, including input pre-processing unit (IPU), dyadic block multiply units (DBMUs), and Canonical Signed Digit (CSD)-based adder trees. It circumvents structured zero values in weights and bypasses unstructured zero bits within non-zero weights and block-wise all-zero bit columns in input features. As a result, the DB-PIM framework skips a majority of unnecessary computations, thereby driving significant gains in computational efficiency. Results demonstrate that our DB-PIM framework achieves up to 8.01x speedup and 85.28% energy savings, significantly boosting computational efficiency in digital SRAM-PIM systems.

cs.AR↗

arXiv2025

Signatures of linearized gravity in atom interferometers: A simplified computational framework

We develop a general framework for calculating the leading-order, general relativistic contributions to the gravitational phase shift in single-photon atom interferometers within the context of linearized gravity. We show that the atom gradiometer observable, which only depends on the atom interferometer propagation phase, can be written in terms of three distinct contributions: the Doppler phase shift, which accounts for the tidal displacement of atoms along the baseline, the Shapiro phase shift, which accounts for the delay in the arrival time of photons at atom-light interaction points, and the Einstein phase shift, which accounts for the gravitational redshift measured by the atoms. For specific atom gradiometer configurations, we derive the signal and response functions for two physically motivated scenarios: (i) transient gravitational waves in the transverse-traceless gauge and, for the first time, in the proper detector frame, and (ii) transient massive objects sourcing weak and slow-varying Newtonian potentials. We find that the Doppler contribution of realistic Newtonian noise sources (e.g., a freight truck or a piece of space debris) at proposed atom gradiometer experiments, such as AION, MAGIS and AEDGE, can exceed the shot noise level and thus affect physics searches if not properly subtracted.

gr-qc↗

arXiv2025

CIMFlow: An Integrated Framework for Systematic Design and Evaluation of Digital CIM Architectures

Digital Compute-in-Memory (CIM) architectures have shown great promise in Deep Neural Network (DNN) acceleration by effectively addressing the "memory wall" bottleneck. However, the development and optimization of digital CIM accelerators are hindered by the lack of comprehensive tools that encompass both software and hardware design spaces. Moreover, existing design and evaluation frameworks often lack support for the capacity constraints inherent in digital CIM architectures. In this paper, we present CIMFlow, an integrated framework that provides an out-of-the-box workflow for implementing and evaluating DNN workloads on digital CIM architectures. CIMFlow bridges the compilation and simulation infrastructures with a flexible instruction set architecture (ISA) design, and addresses the constraints of digital CIM through advanced partitioning and parallelism strategies in the compilation flow. Our evaluation demonstrates that CIMFlow enables systematic prototyping and optimization of digital CIM architectures across diverse configurations, providing researchers and designers with an accessible platform for extensive design space exploration.

cs.AR↗

arXiv2025

VisuoThink: Empowering LVLM Reasoning with Multimodal Tree Search

Recent advancements in Large Vision-Language Models have showcased remarkable capabilities. However, they often falter when confronted with complex reasoning tasks that humans typically address through visual aids and deliberate, step-by-step thinking. While existing methods have explored text-based slow thinking or rudimentary visual assistance, they fall short of capturing the intricate, interleaved nature of human visual-verbal reasoning processes. To overcome these limitations and inspired by the mechanisms of slow thinking in human cognition, we introduce VisuoThink, a novel framework that seamlessly integrates visuospatial and linguistic domains. VisuoThink facilitates multimodal slow thinking by enabling progressive visual-textual reasoning and incorporates test-time scaling through look-ahead tree search. Extensive experiments demonstrate that VisuoThink significantly enhances reasoning capabilities via inference-time scaling, even without fine-tuning, achieving state-of-the-art performance in tasks involving geometry and spatial reasoning.

cs.CL↗

arXiv2024

Contrast Loss from Astrophysical Backgrounds in Space-Based Matter-Wave Interferometers

Atom and matter interferometers are precise quantum sensing experiments that can probe differential forces along separated spacetime paths. Various atom and matter interferometer experiments have been proposed to study dark matter, gravitational waves, and exotic new physics. Increasingly, these experimental concepts have proposed space-based designs to maximize interrogation times and baselines. However, decoherence and phase shifts caused by astrophysical backgrounds could largely undermine or destroy the target sensitivity of the experiments. We calculate the decoherence effects induced by solar photons, the solar wind, cosmic rays, solar neutrinos and zodiacal dust on space-based atom and matter interferometers. We find that, in future space-based atom and matter interferometers, the solar wind generically produces decoherence beyond the quantum noise limit, without proper shielding. In addition, solar photons are also an important background for matter interferometers.

quant-ph↗

arXiv2024

Atom Interferometer Tests of Dark Matter

Direct detection experiments for dark matter are increasingly ruling out large parameter spaces. However, light dark matter models with particle masses $<$ GeV are still largely unconstrained. Here we examine a proposal to use atom interferometers to detect a light dark matter subcomponent at sub-GeV masses. We describe the decoherence and phase shifts caused by dark matter scattering off of one "arm" of an atom interferometer using a generalized dark matter direct detection framework. This allows us to consider multiple channels: nuclear recoils, hidden photon processes, and axion interactions. We apply this framework to several proposed atom interferometer experiments. Because atom interferometers are sensitive to extremely low momentum deposition and their coherent atoms may give them a boost in sensitivity, these experiments will be highly competitive and complementary to other direct detection methods. In particular, atom interferometers are uniquely able to probe a dark matter sub-component with $m_χ\lesssim 10~\rm{keV}$. We find that, for a mediator mass $m_ϕ=10^{-5}m_χ$, future atom interferometers could close a gap in the existing constraints on nuclear recoils down to $\barσ_n \sim 10^{-42}~\rm{cm}^2$ for $m_χ\sim 10^{-5} - 10^{-1}~\rm{MeV}$ dark matter masses.

hep-ph↗