arXiv subjects

Yi Yang

Publications and source records attributed to Yi Yang.

At least 91 records · Page 5Linked to original sources

arXiv2026

Agentic Reinforcement Learning with Observation-Calibrated Self-Distillation

Large language model agents are commonly trained through reinforcement learning with sparse trajectory-level rewards, which offer limited guidance on how strongly individual tokens should be updated. On-Policy Self-Distillation (OPSD) addresses this by re-scoring generated tokens under a privileged replay view to obtain dense, token-level supervision. However, we identify a confounding issue: the resulting support may reflect both the privileged information contained in the replay view and score shifts induced by the replay scaffold, making it difficult to attribute the support specifically to that information. This issue is especially pronounced when future environment observations serve as privileged information, since replaying them requires reconstructing an extended scaffold that itself perturbs token scores. To resolve this confounding, we propose Observation-Calibrated Self-Distillation (OCSD), which contrasts two structurally matched replay views, Full and Observation-Ablated, differing only in whether the actual future observation is present, to derive an observation residual that discounts score changes shared by the replay scaffold. OCSD then applies this residual to modulate token-level GRPO updates at high-uncertainty steps, while preserving the trajectory-level update direction. Experiments on ALFWorld, WebShop, and Search-QA across three Qwen3 model scales show that OCSD consistently outperforms strong baselines. Diagnostic analyses further confirm that the calibrated residual aligns better with local environment feedback. Our code is publicly available at https://github.com/yiy1x/OCSD.

cs.LG↗

arXiv2026

Data-DPO: Direct Preference Optimization for Target Model Data Selection in LLM Post-Training

Data selection in supervised fine-tuning aims to select a small set of effective samples from large-scale candidate data, reducing training cost while preserving model performance. However, existing methods usually treat data value as a relatively static property, and pay limited attention to the compatibility between data and the capability distribution of the target model. To address this issue, we propose Data-DPO, a target model-oriented SFT data selection method. Data-DPO observes the local training feedback of the target model on different samples through one-step probing, transforms activation differences among samples into pairwise data preferences, and trains a lightweight reward model to learn target-model-aware data preferences. In the final selection stage, Data-DPO further combines target model preference, external quality scores, and marginal diversity to construct a more stable and effective training subset. Experimental results on Vision-Flan and LLaVA-CoT show that Data-DPO consistently outperforms existing data selection baselines under multiple data budgets and stably surpasses full data training performance.

cs.LG↗

arXiv2026

Hierarchical Data Selection via Manifold Coverage and Sparse Feature Coverage in LLM Post-training

As supervised fine-tuning data continues to scale, selecting high-value subsets from large candidate pools is crucial for reducing training cost and improving model performance. Existing methods often measure diversity directly in the original embedding space, where geometric metrics entangle dominant semantic directions, fine-grained supervision differences, and local noise. We address this limitation by formulating data selection as a coarse-to-fine hierarchical coverage problem and propose MASS. MASS learns low-dimensional principal manifold coordinates with a dense autoencoder for coarse semantic grouping, and then performs quality-aware sparse feature coverage within each group using a TopK sparse autoencoder. Experiments on Vision Flan and LLaVA-CoT show that MASS consistently outperforms strong data selection baselines across multiple budgets, and in several settings matches or surpasses full data training with only a small subset of data.

cs.LG↗

arXiv2026

FadeMem: Distance-Aware Memory Consolidation for Autoregressive Video Diffusion

Autoregressive video generators synthesize long videos by generating successive temporal segments, but their historical KV cache grows with video length. Existing bounded-cache methods reduce this cost with local windows, sink tokens, or compressed memory states, yet they usually assign fixed roles to different parts of the history. We propose FadeMem, a distance-aware KV memory consolidation mechanism that organizes historical KV blocks into a temporal hierarchy under a fixed cache budget. This design is motivated by frequency-dependent temporal decay: fine details decorrelate quickly, while coarse scene structure and identity remain useful over longer horizons. During generation, new history is inserted as fine-grained entries, while older adjacent entries are progressively merged under a power-law temporal allocation schedule, yielding a dense-near, sparse-far memory within one cache. Without architectural changes, FadeMem improves long-range consistency while largely preserving visual quality, and lightweight adaptation further enhances motion dynamics and visual fidelity. Using the same unified schedule under a fixed cache budget, FadeMem also remains effective over multi-minute and hour long video generation and reduces peak memory under a matched KV budget.

cs.CV↗

arXiv2026

Convex-Hull-Neighborhood Smooth Dual Generalization: Controlling Local Correction Propagation in Offline RL

Offline reinforcement learning (offline RL) can benefit from nearby out-of-distribution (OOD) actions, but estimation errors at these actions may be amplified by bootstrapping. Existing regularization and local-generalization methods control either the admissible OOD region or the influence of generalized targets, often through separate mechanisms. We propose Convex Hull Neighborhood Smooth Dual Generalization (CSDG), which expresses the Bellman backup as an in-sample value target plus a CHN-local correction. This formulation makes the generalized contribution explicit and separates it from the in-sample reference path. The correction is obtained by smoothing in-sample-oriented and OOD-oriented candidates sampled at different perturbation radii. A mixture coefficient lambda scales its contribution to each backup, while the recursive discount remains gamma. Under boundedness and fixed perturbation kernels, we derive an exact one-step correction identity, a time-varying iterate bound, and a fixed-point bound that depends only on the branch discrepancy at the fixed point. We further characterize the implicit policies induced by the idealized operators and give a conditional non-degradation criterion. The practical algorithm approximates these quantities using asymmetric bounded noise and expectile regression, without exact support classification or an additional pessimistic OOD penalty. Experiments on Gym-MuJoCo and AntMaze show strong aggregate performance and stable value estimation. Code is available at: https://github.com/YOUNG-fnxm/CSDG

cs.LG↗

arXiv2026

DAG-FM: A Foundation Model for Causal Discovery under Heterogeneous Causal Mechanisms

Causal discovery from observational tabular data remains fundamentally challenging, primarily due to the heterogeneity of underlying causal mechanisms and the high-dimensional combinatorial search space of Directed Acyclic Graphs (DAGs). In this paper, we propose \textbf{DAG-FM}, a novel foundation model architecture that amortizes causal discovery. Unlike direct matrix prediction, DAG-FM decomposes the causal discovery process into two auto-regressive stages using two specialized Transformer-based sub-modules: a leaf-node predictor and a parent-node predictor. To effectively model complex row-column interactions, we adopt a robust tabular interaction block to output feature-wise representations. Crucially, to handle diverse and unknown Functional Causal Model (FCM) assumptions in real-world scenarios, we introduce Mixture-of-Leaf-Experts (MoLE), allowing the model to dynamically route and adapt to identifiable mechanism families. Through an iterative inference algorithm, DAG-FM seamlessly extracts causal orderings and constructs valid DAGs. Extensive experiments demonstrate that DAG-FM achieves state-of-the-art performance on both synthetic benchmarks and complex real-world datasets, significantly outperforming traditional classical algorithms and recent foundation models in both accuracy and scalability.

cs.LG↗

arXiv2026

Agentic Graph Token Reasoning

Graphs model relational data throughout science and industry, from citation networks to product co-purchase graphs. Because the nodes of many such graphs carry rich text, a growing line of work applies large language models (LLMs) to graph analysis. The most graph-native of these methods use graph tokens: a graph encoder compresses a graph view, such as a node, its k-hop neighbourhood, or a cluster, into a short block of continuous tokens that jointly encodes node attributes and topology and is read directly by the model. Existing methods, however, use graph tokens in a static single-shot manner: they encode one predefined graph view before the model has even seen the target and never revise it, leaving the model's step-by-step reasoning ability unused. We introduce agentic graph token reasoning, which recasts graph tokenization as part of the reasoning process itself. At each step, the model chooses which graph view to encode and at what granularity; a graph encoder is invoked on demand to materialise the corresponding graph tokens; and the resulting block is spliced into the running context. The model thus reasons step by step in the graph token space, and the tokens it reads are trajectory-dependent. We realise this with a three-stage training pipeline: (i) self-supervised tasks that teach the model to read heterogeneous graph tokens, (ii) a token-robust trajectory stage with a graph-token consistency regulariser, and (iii) preference optimisation that rewards trajectories in which the graph-token evidence and the node-text evidence agree. Across evaluations spanning seven graph domains, our models outperform a broad set of baselines by a large margin and transfer zero-shot to unseen domains without any per-target fine-tuning. More broadly, this work pushes LLM-based graph analysis from static graph-token encoders towards a graph-native agent paradigm.

cs.LG↗

arXiv2026

TRACE: High-Fidelity 3D Scene Editing via Tangible Reconstruction and Geometry-Aligned Contextual Video Masking

Existing 3D Gaussian Splatting (3DGS) editing methods primarily focus on appearance modification and often struggle to support flexible geometry editing while preserving structural integrity and scene-consistent appearance. To address this limitation, we present TRACE, a mesh-guided 3DGS editing framework that automatically aligns explicit 3D geometry with Gaussian scenes and decouples Geometric Anchoring from Appearance Harmonization. First, Multi-view 3D-Anchor Synthesis, trained on our MV-TRACE dataset for scene-coherent object addition and modification, generates geometrically aligned editing anchors, while Tangible Geometry Alignment (TGA) performs coarse-to-fine mesh-scene registration. Then, Contextual Video Masking (CVM) integrates projected 3D anchors into an autoregressive video diffusion pipeline, harmonizing their appearance with the surrounding scene while maintaining multi-view consistency. We evaluate TRACE on eight held-out scenes across six editing categories. TRACE completes each edit in approximately 10 minutes on a single NVIDIA RTX Pro 6000 GPU. Extensive experiments demonstrate consistent improvements over existing methods in editing versatility, structural integrity, semantic alignment, multi-view consistency, and visual quality.

cs.CV↗

arXiv2026

OSCAgent: Accelerating the Discovery of Organic Solar Cells with LLM Agents

Organic solar cells (OSCs) hold great promise for sustainable energy, but discovering high-performance materials is time-consuming and costly. Existing molecular generation methods can aid the design of OSC molecules, but they are mostly confined to optimizing known backbones and lack effective use of domain-specific chemical knowledge, often leading to unrealistic candidates. In this paper, we introduce OSCAgent, a multi-agent framework for OSC molecular discovery that unifies retrieval-augmented design, molecular generation, and systematic evaluation into a continuously improving pipeline, without requiring additional human intervention. OSCAgent comprises three collaborative agents. The Planner retrieves knowledge from literature-curated molecules and prior candidates to guide design directions. The Generator proposes new OSC acceptors aligned with these plans. The Experimenter performs comprehensive evaluation of candidate molecules and provides feedback for refinement. Experiments show that OSCAgent produces chemically valid, synthetically accessible OSC molecules and achieves superior predicted performance compared to both traditional and large language model (LLM)-only baselines. Representative results demonstrate that some candidates achieve predicted efficiencies approaching 18\%. The code will be publicly available.

cs.CE↗

arXiv2026

Cross-Environment Diagnostics for New Physics in Proton-Proton and Heavy-Ion Collisions

Proton-proton and heavy-ion measurements are usually interpreted within separate theoretical and experimental frameworks, even when the same reconstructed final state is measured and the $pp$ result provides or constrains the elementary reference for nuclear observables. This separation creates a potential blind spot: an omitted contribution can be absorbed into production or signal-model parameters in $pp$ collisions and then be reinterpreted through independent nuclear-medium parameters in heavy-ion collisions. We propose cross-environment closure as a complementary search principle. Conventional parameters remain specific to each collision system, while one candidate new-physics contribution must remain coherent across systems and observables. Quarkonium is used as a controlled illustration rather than as the definition of the method. A near-degenerate dimuon component is shown to be absorbable into a retuned $pp$ spectrum and a single-peak mass fit, while the same $p_T$-dependent contribution propagates into representative $R_{pA}$, $R_{AA}$, and $v_2$ relations. These quantities are consistency observables, not assumed new-physics baselines. The numerical examples are stress tests, not claims about an allowed particle or an anomaly in current data. The main result is a transferable diagnostic principle, illustrated here with quarkonium rather than formulated as a universal statistical framework.

hep-ph↗

arXiv2026

Emission-Forecasting-Based Spatial-Temporal Carbon Response: A Multi-Agent Attention-Enhanced Deep Learning Framework

As a major contributor to carbon emissions, the decarbonization of power systems has garnered significant societal attention. Nodal carbon intensity (NCI), a critical factor in carbon-oriented demand response, has traditionally been determined through ex-post calculations. However, this ex-post approach introduces latency in low-carbon dispatch. To address this, this paper presents a proactive ex-ante spatial-temporal carbon response framework. At its core, we develop a novel deep learning-based hierarchical design, enhanced by a dual-stage attention mechanism and a large language model (LLM)-based multi-agent cooperation system, to accurately forecast day-ahead NCI. This design effectively mitigates the impact of renewable energy uncertainty and enhances predictive resilience. On the demand side, the framework proposes a spatial-temporal carbon scheduling model that integrates geographically dispatchable loads (GDLs), including mobile energy storage systems (MESSs) and distributed data centers (DDCs). Leveraging high-accuracy day-ahead NCI predictions, the framework can effectively reduce system emissions by quickly responding to carbon intensity fluctuations. The proposed framework is tested on the modified IEEE 33-bus system. According to the simulation results, the impacts of proposed framework on dispatching latency and emission outcomes are analyzed. The results demonstrate that under a one-hour reduction in carbon scheduling latency, the proposed model and methodology can achieve over 30% emission reduction. This research breaks through the limitations of passive carbon accounting, advancing toward proactive carbon management. It offers an intelligent solution that accelerates the transition to cleaner power systems while directly supporting sustainable production goals.

eess.SY↗

arXiv2026

JWST Spectra Indicate a Large Mass of Postshock Dust Formed by SN 2010jl

We present new JWST mid-IR spectra and ground-based optical spectra of the lingering source at the position of SN 2010jl, which was a relatively nearby superluminous Type IIn supernova having strong interaction with circumstellar material (CSM). Early-time data showed evidence of dust, interpreted as either pre-existing CSM dust, or as newly formed dust in the SN ejecta and post-shock region. At 13 yr post explosion, JWST reveals a strong MIR excess from warm dust, with broad features at 10-15 $μ$m. Our analysis reveals a minimum dust mass of $>$0.11 $M_{\odot}$, and a more likely value of 0.2 $M_{\odot}$ or more because the dust is optically thick. This is among the largest masses of SN-produced dust yet measured without far-IR/submm data, and greatly exceeds SN 2010jl's dust mass inferred around 2-3 yr post-explosion. Ground-based optical spectra confirm the presence of a young massive cluster at the SN position, and confirm that blueshifted line profiles persisting until the latest epochs arise from dust formed in the post-shock region. The warmest dust emitting in the MIR is likely to be the same post-shock dust causing the blueshift. The JWST spectrum also reveals silicate absorption, which may arise from cool SN ejecta dust along the line of sight to the receding shock. The large mass of post-shock dust in SN 2010jl suggests that strong CSM interaction promotes efficient dust production, where the new post-shock dust will survive. If strongly interacting SNe are common in the early Universe, this may contribute significantly to dust seen in infant galaxies.

astro-ph.HE↗

arXiv2026

ClinFusion: A Vision-Centric Multimodal LLM System for Holistic Medical Understanding

Multimodal large language models (MLLMs) hold immense potential to revolutionize clinical practice, yet deploying them in the medical domain is fundamentally a vision-centric challenge: models must absorb knowledge from heterogeneous 2D and 3D medical images, and evaluation protocols must align with radiologists' clinical practice and provide an accurate, fine-grained and factualness-driven assessment. In this paper, we introduce ClinFusion, a vision-centric MLLM designed for holistic medical understanding that systematically addresses these limitations. We propose a compositional and cascaded vision encoder architecture featuring a Cascade Spatial-Aware Locality Fusion operator that unifies diverse 2D and native 3D medical image understanding within a fused encoder. We further introduce a vision-grounded evaluation framework, including MedIF-Bench for instruction-following assessment and a region-of-interest-grounded method for clinically aligned and factualness-driven report generation evaluation. We show that ClinFusion sets a new state-of-the-art across a comprehensive suite of 2D and 3D multimodal medical benchmarks---spanning visual question answering, report generation, and instruction following---as well as textual medical tasks, outperforming leading open-source medical MLLMs (\textit{e.g.}, Hulu-Med, Lingshu) on 20 out of 24 benchmarks and demonstrating multimodal capabilities better than powerful proprietary models such as GPT-5.2 and Gemini-3-Flash on 13 out of 16 benchmarks, and can be further augmented with agentic tool use for retrieval-augmented and tool-assisted clinical workflows. A blinded evaluation by board-certified radiologists confirms that ClinFusion produces the highest-ranked reports, and validates our RoI-grounded metric as achieving the strongest correlation with expert judgment among all automatic evaluation metrics examined.

cs.CV↗

arXiv2026

SN 2022erq: A Superluminous Thermonuclear Supernova with Escalating Preexplosion Mass Loss

We present a photometric and spectroscopic study of the superluminous Type Ia supernova SN 2022erq. Its early spectra, dominated by iron-group elements with weak intermediate-mass features, might indicate highly efficient nuclear burning, broadly similar to that inferred for some overluminous SNe Ia. The rapid emergence and persistence of narrow Balmer emission lines superposed on this iron-rich spectrum provide clear evidence of long-lived interaction with a hydrogen-rich circumstellar medium (CSM), establishing SN 2022erq as a member of the rare Ia-CSM class. SN 2022erq reached a peak bolometric luminosity of about 8 x 10^43 erg/s and exhibited an exceptionally slow post-peak decline, indicating that its light curve is dominated by long-duration ejecta-CSM interaction. By combining H-alpha diagnostics with bolometric light-curve modeling, we reconstruct the pre-explosion mass-loss history of the progenitor. The mass-loss rate escalated by one order of magnitude over the final decades, rising from about 0.04 to about 0.6 solar masses per year. This surge produced a massive, extended CSM shell of about 3 solar masses out to about 3.5 x 10^16 cm. The young stellar environment (about 100 Myr) together with this substantial, extensive CSM points to a progenitor system consisting of a white dwarf and an intermediate-mass companion that underwent increasing mass loss prior to explosion.

astro-ph.HE↗

arXiv2026

Engineering two-body interaction for the Moore-Read State

Engineering interactions that stabilize non-Abelian fractional quantum Hall phases is a central challenge in strongly correlated topological matter and quantum simulation. We introduce a differentiable framework for inverse Hamiltonian design, in which Haldane pseudopotentials are optimized by gradient-based exact diagonalization to stabilize target fractional quantum Hall phases. In spherical geometry, the Haldane pseudopotentials are treated as variational parameters and optimized in a JAX-based exact-diagonalization framework. By directly maximizing the overlap between the many-body ground state and the Moore-Read state, we obtain a robust pseudopotential profile that has Pfaffian overlaps exceeding $99\%$ for systems up to $N_e=12$, substantially improving over conventional Coulomb interactions. Analyses of the neutral excitation spectrum and orbital entanglement spectrum further confirm that the optimized interaction stabilizes the Pfaffian topological phase. Our results demonstrate that essential features of the three-body Pfaffian parent Hamiltonian can be effectively encoded in a suitably designed two-body interaction. Furthermore, they identify a nearly universal exponentially decaying pseudopotential profile that stabilizes the Pfaffian phase and establishes a general framework toward engineering non-Abelian topological order in quantum simulation.

cond-mat.str-el↗

arXiv2026

Gravity with a Non-Dynamical Metric in a Yang-Mills Framework

General relativity assigns the spacetime metric two logically distinct roles: it defines physical measurements and it carries the propagating degrees of freedom of gravity. We investigate whether these roles must be assigned to the same field. We formulate a classical Yang-Mills framework in which the metric retains its operational role in defining distances, time intervals, causal structure, index contractions, and spacetime volume, but carries no kinetic term and no independent propagating degrees of freedom. The metric is nevertheless varied as an auxiliary field, producing an algebraic consistency condition rather than a propagation equation. The gravity dynamics are instead assigned to a local GL(4,R) Yang--Mills connection with sixteen gauge components. An explicit dictionary relates the Yang-Mills gauge potential to a connection-like geometric variable, under which the field strength is rewritten in curvature form. The distinctive feature of the construction is therefore not the use of GL(4,R) or a Yang-Mills-type gravitational action by themselves, but the variational hierarchy in which a physical measuring metric is non-propagating and algebraically constrained while the connection carries the dynamics. We clarify the relation of this framework to Poincaré gauge theory, metric-affine and Palatini formulations, pure-connection gravity, teleparallel approaches, metric quadratic gravity, and the historical Stephenson-Kilmister-Yang/Fairchild sector. In the restricted torsionless, metric-compatible sector, Ricci-flat geometries including the Schwarzschild metric remain admissible exact solutions, while additional branches are retained only as formal sectors requiring separate phenomenological assessment.

physics.gen-ph↗

arXiv2026

Supercurrent effect in a charge density wave intertwined superconductor

The energy-momentum (E-k) dispersion of quasiparticles constitutes a fundamental concept in condensed matter systems. The ability to modify the E-k dispersion, exemplified by supercurrent-induced Doppler shifts of Bogoliubov quasiparticle spectra in superconductors, enables manipulation of various emergent quantum properties. However, investigations into the supercurrent effect on superconductors intertwined with charge orders remain scarce. Here, we report that the Meissner current, generated by the diamagnetic response to an applied in-plane magnetic field, can tailor Bogoliubov quasiparticle excitations at the precursor charge density wave (CDW) vectors. Our scanning tunneling spectroscopic imaging reveals a field-driven symmetry breaking of CDW modulations, specifically a C3v-to-Cs transition, in superconducting NbSe2. Model calculations suggest that the observed anisotropy originates from a selective Doppler-shift-induced E-k dispersion reconstruction. Furthermore, altering the field direction enables on-demand tuning of anisotropic CDW modulations and visualization of their momentum-space distribution. These results highlight a novel mechanism for controlling emergent electronic phases through momentum-space engineering.

cond-mat.supr-con↗

arXiv2026

Characterization of the Scattered Light Noise in KAGRA Interferometer

In a gravitational wave (GW) telescope, non-stationary noise can prevent stable interferometer operation and limit the GW analyses. Among the non-stationary noise, we focus on the scattered light noise. It fluctuates across a wide frequency band within a short time. Therefore, the scattered light noise can mimic the GW signal, increase the false alarm rate, or bias the parameter estimation results. In this work, we propose a new analysis method, named f-cluster, to characterize these features without prior assumptions about the scattering source. The scattered light noise shows characteristic arch-shaped features in time-frequency maps. Therefore, the algorithm extracts the periodicity of the arch shapes from the spectrogram. The algorithm detects the occurrence times and periodicity of excess-power regions in the spectrogram to estimate the oscillation frequency of the scattering object. To demonstrate the algorithm, we applied this method to KAGRA data from April 1 to 30, 2025, and successfully identified a scattering frequency of $f_{sc} = 0.432$ Hz. Furthermore, we found a strong correlation between the occurrence rate of the scattered light noise and the amplitude of ground motion in the $0.3-1.0$ Hz band with a Pearson correlation coefficient of 0.82. This correlation showed that the occurrence of scattered light noise can be forecast from the seismic motion. We confirmed that our algorithm identifies the frequency of the scattered light noise and its occurrence time, without making assumptions about the scattering object, thereby contributing to the characterization of the scattered light noise in real data. Our algorithm provides information that leads to further investigation into the noise hunting to specify the source of scattered light noise.

gr-qc↗