arXiv subjects

Song Wang

Publications and source records attributed to Song Wang.

At least 127 records · Page 7Linked to original sources

arXiv2026

Probing to Refine: Reinforcement Distillation of LLMs via Explanatory Inversion

Distilling robust reasoning capabilities from large language models (LLMs) into smaller, computationally efficient student models remains an unresolved challenge. Despite recent advances, distilled models frequently suffer from superficial pattern memorization and subpar generalization. To overcome these limitations, we introduce a novel distillation framework that moves beyond simple mimicry to instill a deeper conceptual understanding. Our framework features two key innovations. \underline{\textit{First}}, to address pattern memorization, Explanatory Inversion (EI) generates targeted ``explanatory probes'' that compel the student to articulate the underlying logic behind an answer, rather than just memorizing it. \underline{\textit{Second}}, to improve generalization, Explanatory GRPO (\texttt{EXGRPO}) uses a reinforcement learning algorithm with a novel Dialogue Structure Utility Bonus, which explicitly rewards the student for maintaining a coherent reasoning process across these probes. Extensive evaluations on 12 datasets demonstrate significant improvements. Using Gemma-7b as the student model, our method yields an average \textbf{20.39\%} increase over zero-shot performance and a \textbf{6.02\%} improvement over the state-of-the-art distillation baselines. Moreover, models distilled with our method show remarkable training efficiency (e.g., surpassing vanilla fine-tuning with \textbf{10-25\%} training data) and strong generalization to out-of-distribution tasks. Implementation is released at https://github.com/Zhen-Tan-dmml/ExGRPO.git.

cs.CL↗

arXiv2026

Varying core-envelope coupling efficiency identified from stellar rotation--activity relation

Core-envelope coupling provides a reasonable explanation of the spin-down stalling of stars in open clusters, which was not predicted by classical gyrochronology. However, it remains an open question whether the coupling efficiency is constant or variable. M dwarfs, possessing thicker convective envelopes and thus longer coupling timescales than other late-type stars, are ideal objects for this investigation. In this work, based on the $R_{\rm{HK}}^{'}$ measurements from LAMOST and DESI spectra, we construct new rotation--activity relations for M dwarfs. Unlike the traditional picture, we suggest that the new relation consists of three distinct regimes of fast, intermediate, and slow rotation, closely matching the three sequences of gyrochronology, namely the ``Convective'' sequence, ``Gap'', and ``Interface'' sequence. Our study reveals, for the first time, a variable activity decay rate in the intermediate-rotation regime (i.e., the ``Gap'' region). This implies a varying core-envelope coupling efficiency, peaking towards the end of this region. It also coincides with the well-known stage of stalled stellar spin-down.

astro-ph.SR↗

arXiv2026

Retarded Stellar Dynamo in Tidally Deformed M Dwarfs

Current studies of stellar dynamos primarily focus on spherical stars, leaving their behavior in distorted stars largely unexplored. We utilize stars of varying distortions to examine the relation between stellar cycle periods ($P_{\rm cyc}$) and rotational periods ($P_{\rm rot}$), which are closely linked to dynamo processes. By analyzing a sample of tidally distorted M dwarfs in cataclysmic variables, we identify an anti-correlation between $P_{\rm cyc}$ and $P_{\rm rot}$, in contrast to the lack of such a relation in single M dwarfs. This means that stars with greater deformation have longer cycle periods, suggesting variations in dynamo behavior under non-spherical geometries. Our numerical simulations further reveal that, the thermal convection weakens in highly distorted stars, and subsequently, the differential rotation is also reduced. These effects may lengthen the conversion timescale between poloidal and toroidal magnetic fields, potentially explaining the newly discovered $P_{\rm cyc}$-$P_{\rm rot}$ relation in cataclysmic variables.

astro-ph.SR↗

arXiv2026

RegTrack: Simplicity Beneath Complexity in Robust Multi-Modal 3D Multi-Object Tracking

Existing 3D multi-object tracking (MOT) methods often sacrifice efficiency and generalizability for robustness, largely relying on complex association metrics derived from multi-modal architectures and class-specific motion priors. Challenging the rooted belief that greater complexity necessarily yields greater robustness, we propose a robust, efficient, and generalizable method for multi-modal 3D MOT, dubbed RegTrack. Inspired by Yang-Mills gauge theory, RegTrack is built upon a unified tri-cue encoder (UTEnc), comprising three tightly coupled components: a local-global point cloud encoder (LG-PEnc), a mixture-of-experts-based geometry encoder (MoE-GEnc), and an image encoder from a well-pretrained visual-language model. LG-PEnc efficiently encodes the spatial and structural information of point clouds to produce foundational representations for each object, whose pairwise similarities serve as the sole association metric. MoE-GEnc seamlessly interacts with LG-PEnc to model inter-object geometric relationships across frames, adaptively compensating for inter-frame object motion without relying on any class-specific priors. The image encoder is kept frozen and is used exclusively during training to provide a well-pretrained representation space. Point cloud representations are aligned to this space to supervise the motion compensation process, encouraging representation invariance across frames for the same object while enhancing discriminability among different objects. Through this formulation, RegTrack attains robust, efficient, and generalizable inference using only point cloud inputs, requiring just 2.6M parameters. Extensive experiments on KITTI and nuScenes show that RegTrack outperforms its thirty-five competitors.

cs.CV↗

arXiv2026

RewardMap: Tackling Sparse Rewards in Fine-grained Visual Reasoning via Multi-Stage Reinforcement Learning

Fine-grained visual reasoning remains a core challenge for multimodal large language models (MLLMs). The recently introduced ReasonMap highlights this gap by showing that even advanced MLLMs struggle with spatial reasoning in structured and information-rich settings such as transit maps, a task of clear practical and scientific importance. However, standard reinforcement learning (RL) on such tasks is impeded by sparse rewards and unstable optimization. To address this, we first construct ReasonMap-Plus, an extended dataset that introduces dense reward signals through Visual Question Answering (VQA) tasks, enabling effective cold-start training of fine-grained visual understanding skills. Next, we propose RewardMap, a multi-stage RL framework designed to improve both visual understanding and reasoning capabilities of MLLMs. RewardMap incorporates two key designs. First, we introduce a difficulty-aware reward design that incorporates detail rewards, directly tackling the sparse rewards while providing richer supervision. Second, we propose a multi-stage RL scheme that bootstraps training from simple perception to complex reasoning tasks, offering a more effective cold-start strategy than conventional Supervised Fine-Tuning (SFT). Experiments on ReasonMap and ReasonMap-Plus demonstrate that each component of RewardMap contributes to consistent performance gains, while their combination yields the best results. Moreover, models trained with RewardMap achieve an average improvement of 3.47% across 6 benchmarks spanning spatial reasoning, fine-grained visual reasoning, and general tasks beyond transit maps, underscoring enhanced visual understanding and reasoning capabilities.

cs.CV↗

arXiv2026

CM2: Reinforcement Learning with Checklist Rewards for Multi-Turn and Multi-Step Agentic Tool Use

AI agents are increasingly used to solve real-world tasks by reasoning over multi-turn user interactions and invoking external tools. However, applying reinforcement learning to such settings remains difficult: realistic objectives often lack verifiable rewards and instead emphasize open-ended behaviors; moreover, RL for multi-turn, multi-step agentic tool use is still underexplored; and building and maintaining executable tool environments is costly, limiting scale and coverage. We propose CM2, an RL framework that replaces verifiable outcome rewards with checklist rewards. CM2 decomposes each turn's intended behavior into fine-grained binary criteria with explicit evidence grounding and structured metadata, turning open-ended judging into more stable classification-style decisions. To balance stability and informativeness, our method adopts a strategy of sparse reward assignment but dense evaluation criteria. Training is performed in a scalable LLM-simulated tool environment, avoiding heavy engineering for large tool sets. Experiments show that CM2 consistently improves over supervised fine-tuning. Starting from an 8B Base model and training on an 8k-example RL dataset, CM2 improves over the SFT counterpart by 8 points on tau^-Bench, by 10 points on BFCL-V4, and by 12 points on ToolSandbox. The results match or even outperform similarly sized open-source baselines, including the judging model. CM2 thus provides a scalable recipe for optimizing multi-turn, multi-step tool-using agents without relying on verifiable rewards. Code provided by the open-source community: https://github.com/namezhenzhang/CM2-RLCR-Tool-Agent.

cs.AI↗

arXiv2026

Privacy-Concealing Cooperative Perception for BEV Scene Segmentation

Cooperative perception systems for autonomous driving aim to overcome the limited perception range of a single vehicle by communicating with adjacent agents to share sensing information. While this improves perception performance, these systems also face a significant privacy-leakage issue, as sensitive visual content can potentially be reconstructed from the shared data. In this paper, we propose a novel Privacy-Concealing Cooperation (PCC) framework for Bird's Eye View (BEV) semantic segmentation. Based on commonly shared BEV features, we design a hiding network to prevent an image reconstruction network from recovering the input images from the shared features. An adversarial learning mechanism is employed to train the network, where the hiding network works to conceal the visual clues in the BEV features while the reconstruction network attempts to uncover these clues. To maintain segmentation performance, the perception network is integrated with the hiding network and optimized end-to-end. The experimental results demonstrate that the proposed PCC framework effectively degrades the quality of the reconstructed images with minimal impact on segmentation performance, providing privacy protection for cooperating vehicles. The source code will be made publicly available upon publication.

cs.CV↗

arXiv2026

BrainTAP: Brain Disorder Prediction with Adaptive Distill and Selective Prior Integration

Predicting clinical outcomes from brain networks in large-scale neuroimaging cohorts such as the Adolescent Brain Cognitive Development (ABCD) study requires effectively integrating functional connectivity (FC) and structural connectivity (SC) while incorporating expert neurobiological knowledge. However, existing multimodal fusion approaches are shallow or over-homogenize the inherently heterogeneous characteristics of FC and SC, while expert-defined anatomical priors are underutilized with static integration. To address these limitations, we propose Brain Transformer with Adaptive Mutual-Distill and Selective Prior Fusion (BrainTAP). We introduce Adaptive Mutual Distill (AMD), which enables layer-wise information exchange between modalities through learnable distill-intact ratios, preserving modality-specific signals while capturing cross-modal synergies. We further develop Selective Prior Fusion (SPF), which integrates expert-defined anatomical priors in an adaptive way. Evaluated on the ABCD dataset for predicting attention-related disorders, BrainTAP achieves superior performance over state-of-the-art baselines, demonstrating its effectiveness for brain disorder prediction.

cs.CE↗

arXiv2026

How Much Reasoning Do Retrieval-Augmented Models Add beyond LLMs? A Benchmarking Framework for Multi-Hop Inference over Hybrid Knowledge

Large language models (LLMs) continue to struggle with knowledge-intensive questions that require up-to-date information and multi-hop reasoning. Augmenting LLMs with hybrid external knowledge, such as unstructured text and structured knowledge graphs, offers a promising alternative to costly continual pretraining. As such, reliable evaluation of their retrieval and reasoning capabilities becomes critical. However, many existing benchmarks increasingly overlap with LLM pretraining data, which means answers or supporting knowledge may already be encoded in model parameters, making it difficult to distinguish genuine retrieval and reasoning from parametric recall. We introduce HybridRAG-Bench, a framework for constructing benchmarks to evaluate retrieval-intensive, multi-hop reasoning over hybrid knowledge. HybridRAG-Bench automatically couples unstructured text and structured knowledge graph representations derived from recent scientific literature on arXiv, and generates knowledge-intensive question-answer pairs grounded in explicit reasoning paths. The framework supports flexible domain and time-frame selection, enabling contamination-aware and customizable evaluation as models and knowledge evolve. Experiments across three domains (artificial intelligence, governance and policy, and bioinformatics) demonstrate that HybridRAG-Bench rewards genuine retrieval and reasoning rather than parametric recall, offering a principled testbed for evaluating hybrid knowledge-augmented reasoning systems. We release our code and data at github.com/junhongmit/HybridRAG-Bench.

cs.LG↗

arXiv2026

The Quest for Efficient Reasoning: A Data-Centric Benchmark to CoT Distillation

Data-centric distillation, including data augmentation, selection, and mixing, offers a promising path to creating smaller, more efficient student Large Language Models (LLMs) that retain strong reasoning abilities. However, there still lacks a comprehensive benchmark to systematically assess the effect of each distillation approach. This paper introduces DC-CoT, the first data-centric benchmark that investigates data manipulation in chain-of-thought (CoT) distillation from method, model and data perspectives. Utilizing various teacher models (e.g., o4-mini, Gemini-Pro, Claude-3.5) and student architectures (e.g., 3B, 7B parameters), we rigorously evaluate the impact of these data manipulations on student model performance across multiple reasoning datasets, with a focus on in-distribution (IID) and out-of-distribution (OOD) generalization, and cross-domain transfer. Our findings aim to provide actionable insights and establish best practices for optimizing CoT distillation through data-centric techniques, ultimately facilitating the development of more accessible and capable reasoning models. The codebase can be accessed at https://github.com/UNITES-Lab/Distillation-Bench

cs.AI↗

arXiv2026

MedRoute: RL-Based Dynamic Specialist Routing in Multi-Agent Medical Diagnosis

Medical diagnosis using Large Multimodal Models (LMMs) has gained increasing attention due to capability of these models in providing precise diagnoses. These models generally combine medical questions with visual inputs to generate diagnoses or treatments. However, they are often overly general and unsuitable under the wide range of medical conditions in real-world healthcare. In clinical practice, diagnosis is performed by multiple specialists, each contributing domain-specific expertise. To emulate this process, a potential solution is to deploy a dynamic multi-agent LMM framework, where each agent functions as a medical specialist. Current approaches in this emerging area, typically relying on static or predefined selection of various specialists, cannot be adapted to the changing practical scenario. In this paper, we propose MedRoute, a flexible and dynamic multi-agent framework that comprises of a collaborative system of specialist LMM agents. Furthermore, we add a General Practitioner with an RL-trained router for dynamic specialist selection, and a Moderator that produces the final decision. In this way, our framework closely mirrors real clinical workflows. Extensive evaluations on text and image-based medical datasets demonstrate improved diagnostic accuracy, outperforming the state-of-the-art baselines. Our work lays a strong foundation for future research. Code and models are available at https://github.com/UCF-CRCV/MedRoute/.

eess.IV↗

arXiv2026

Physics-Driven Local-Whole Elastic Deformation Modeling for Point Cloud Representation Learning

Existing point cloud representation learning methods primarily rely on data-driven strategies to extract geometric information from large amounts of scattered data. However, most methods focus solely on the spatial distribution features of point clouds while overlooking the relationship between local information and the whole structure, which limits the accuracy of point cloud representation. Local information reflect the fine-grained variations of an object, while the whole structure is determined by the interaction and combination of these local features, collectively defining the object's shape. In real-world, objects undergo deformation under external forces, and this deformation gradually affects the whole structure through the propagation of forces from local regions, thereby altering the object's geometric features. Therefore, appropriately introducing a physics-driven mechanism to capture the topological relationships between local parts and the whole object can effectively mitigate for the limitations of data-driven point cloud methods in structural modeling, and enhance the generalization and interpretability of point cloud representations for downstream tasks such as understanding and recognition. Inspired by this, we incorporate a physics-driven mechanism into the data-driven method to learn fine-grained features in point clouds and model the structural relationship between local regions and the whole shape. Specifically, we design a dual-task encoder-decoder framework that combines the geometric modeling capability of data-driven implicit fields with physics-driven elastic deformation. Through the integration of physics-based loss functions, the framework is guided to predict localized deformation and explicitly capture the correspondence between local structural changes and whole shape variations.

cs.CV↗

arXiv2026

GenMRP: A Generative Multi-Route Planning Framework for Efficient and Personalized Real-Time Industrial Navigation

Existing industrial-scale navigation applications contend with massive road networks, typically employing two main categories of approaches for route planning. The first relies on precomputed road costs for optimal routing and heuristic algorithms for generating alternatives, while the second, generative methods, has recently gained significant attention. However, the former struggles with personalization and route diversity, while the latter fails to meet the efficiency requirements of large-scale real-time scenarios. To address these limitations, we propose GenMRP, a generative framework for multi-route planning. To ensure generation efficiency, GenMRP first introduces a skeleton-to-capillary approach that dynamically constructs a relevant sub-network significantly smaller than the full road network. Within this sub-network, routes are generated iteratively. The first iteration identifies the optimal route, while the subsequent ones generate alternatives that balance quality and diversity using the newly proposed correctional boosting approach. Each iteration incorporates road features, user historical sequences, and previously generated routes into a Link Cost Model to update road costs, followed by route generation using the Dijkstra algorithm. Extensive experiments show that GenMRP achieves state-of-the-art performance with high efficiency in both offline and online environments. To facilitate further research, we have publicly released the training and evaluation dataset. GenMRP has been fully deployed in a real-world navigation app, demonstrating its effectiveness and benefits.

cs.RO↗

arXiv2026

RE-TRAC: REcursive TRAjectory Compression for Deep Search Agents

LLM-based deep research agents are largely built on the ReAct framework. This linear design makes it difficult to revisit earlier states, branch into alternative search directions, or maintain global awareness under long contexts, often leading to local optima, redundant exploration, and inefficient search. We propose Re-TRAC, an agentic framework that performs cross-trajectory exploration by generating a structured state representation after each trajectory to summarize evidence, uncertainties, failures, and future plans, and conditioning subsequent trajectories on this state representation. This enables iterative reflection and globally informed planning, reframing research as a progressive process. Empirical results show that Re-TRAC consistently outperforms ReAct by 15-20% on BrowseComp with frontier LLMs. For smaller models, we introduce Re-TRAC-aware supervised fine-tuning, achieving state-of-the-art performance at comparable scales. Notably, Re-TRAC shows a monotonic reduction in tool calls and token usage across rounds, indicating progressively targeted exploration driven by cross-trajectory reflection rather than redundant search.

cs.CL↗

arXiv2026

Dynamical Mass Loss at the End of TP-AGB stars

The thermally pulsating asymptotic giant branch (TP-AGB) phase plays a key role in the evolution of low- to intermediate-mass stars, driving mass loss that influences their final stages and contributes to galactic chemical enrichment. However, the mechanisms behind mass loss, particularly at the end of AGB, are still not well understood. We aim to investigate the relationship between stellar parameters and envelope dynamics during the TP-AGB phase, evaluating whether dynamical instabilities in the envelope can act as a possible mass-loss mechanism. We use hydrodynamics method in MESA to simulate the dynamical pulsations and resulting mass loss during the TP-AGB phase of a star evolved from a 1.5 Msun zero-age main sequence. Our simulations reproduce the dynamical pulsation behavior of stars during the TP-AGB phase, demonstrating that the envelope mass is a key factor governing pulsational properties. As the envelope mass decreases, both the pulsation period and radial amplitude increase, consistent with observational trends. For 1.5 Msun model, once the envelope mass declines to approximately 0.25 Msun, the model enters a regime of violent pulsations, potentially ejecting the remaining envelope within a few hundred years. We suggest that the instability can act as the dominant mass-loss mechanism in the end of the TP-AGB phase, marking a rapid transitional stage toward the post-AGB phase.

astro-ph.SR↗

arXiv2026

Evolution of Stellar Activity and Habitable Zone II: Ca H&K Emissions of Late-type Dwarfs

Stellar chromospheric activity serves as a valuable proxy for estimating stellar ages, though its applicable range and accurate functional form are still debated. In this study, utilizing the LAMOST spectra we compiled a catalog of open cluster members and field stars to investigate $R_{\rm{HK}}^{'}$--age relations across various spectral types. We find that a linear model, specifically a Skumanich-type relation, can best describe the overall decline of chromospheric activity with age, with the slope varying across different spectral types. However, we also identify variations in the decay rate along the main sequence, which call for more accurate follow-up investigation. Finally, we find that lower-metallicity stars exhibit enhanced activity for F-, G-, and K-type stars, whereas no clear metallicity dependence is observed for M dwarfs.

astro-ph.SR↗

arXiv2026

Metacognitive Self-Correction for Multi-Agent System via Prototype-Guided Next-Execution Reconstruction

Large Language Model based multi-agent systems (MAS) excel at collaborative problem solving but remain brittle to cascading errors: a single faulty step can propagate across agents and disrupt the trajectory. In this paper, we present MASC, a metacognitive framework that endows MAS with real-time, unsupervised, step-level error detection and self-correction. MASC rethinks detection as history-conditioned anomaly scoring via two complementary designs: (1) Next-Execution Reconstruction, which predicts the embedding of the next step from the query and interaction history to capture causal consistency, and (2) Prototype-Guided Enhancement, which learns a prototype prior over normal-step embeddings and uses it to stabilize reconstruction and anomaly scoring under sparse context (e.g., early steps). When an anomaly step is flagged, MASC triggers a correction agent to revise the acting agent's output before information flows downstream. On the Who&When benchmark, MASC consistently outperforms all baselines, improving step-level error detection by up to 8.47% AUC-ROC ; When plugged into diverse MAS frameworks, it delivers consistent end-to-end gains across architectures, confirming that our metacognitive monitoring and targeted correction can mitigate error propagation with minimal overhead.

cs.AI↗

arXiv2026

Forging Spatial Intelligence: A Roadmap of Multi-Modal Data Pre-Training for Autonomous Systems

The rapid advancement of autonomous systems, including self-driving vehicles and drones, has intensified the need to forge true Spatial Intelligence from multi-modal onboard sensor data. While foundation models excel in single-modal contexts, integrating their capabilities across diverse sensors like cameras and LiDAR to create a unified understanding remains a formidable challenge. This paper presents a comprehensive framework for multi-modal pre-training, identifying the core set of techniques driving progress toward this goal. We dissect the interplay between foundational sensor characteristics and learning strategies, evaluating the role of platform-specific datasets in enabling these advancements. Our central contribution is the formulation of a unified taxonomy for pre-training paradigms: ranging from single-modality baselines to sophisticated unified frameworks that learn holistic representations for advanced tasks like 3D object detection and semantic occupancy prediction. Furthermore, we investigate the integration of textual inputs and occupancy representations to facilitate open-world perception and planning. Finally, we identify critical bottlenecks, such as computational efficiency and model scalability, and propose a roadmap toward general-purpose multi-modal foundation models capable of achieving robust Spatial Intelligence for real-world deployment.

cs.CV↗