arXiv subjects

Guang Yang

Publications and source records attributed to Guang Yang.

At least 55 records · Page 3Linked to original sources

arXiv2026

Less Data, Better Alignment: Data-Centric Multi-Evaluator Agreement for Preference Optimization

Research on preference optimization often varies the training objective while holding the data fixed. We instead ask whether a small, high-confidence set of on-policy responses can provide a reliable learning signal. Our method, DMAPO (Data-centric Multi-evaluator Agreement for Preference Optimization), generates candidate responses from the target policy, evaluates helpfulness, factuality, and conciseness with rubric-specialized evaluators, applies a process-critic correction, and retains only high-consensus desirable or undesirable examples. This procedure accepts 1,871 of 54,236 Mistral-7B candidates (3.45%). KTO trained on this set reaches 7.50 on MT-Bench, 95.5% length-controlled win rate against a text-davinci-003 reference, and 57.3% IFEval prompt accuracy. Independent pairwise evaluation also favors DMAPO over SimPO: GPT-4o yields a net win rate of 23.3 points on 129 held-out prompts and 24.0 points on 200 out-of-distribution LMSYS-Chat prompts; Claude Opus 4.7 yields 24.1 points on the held-out set. Changing the evaluator model or rubric alters the selected examples but has little effect on downstream performance. A second-backbone study yields a similar 3.41% acceptance rate, although its performance gains are more modest. Across these experiments, consensus filtering offers a data-efficient route to preference optimization for general instructions, at the cost of additional curation compute and dependence on evaluator judgments.

cs.AI↗

arXiv2026

GLID: Gated Local Intrinsic Dimension Repairs the Blind Spots of Face-Forgery Detectors

Fine-tuned foundation-model detectors dominate face-forgery benchmarks, yet they stay blind to generator families absent from training. We present GLID, a detector that repairs this blind spot with geometry instead of data. GLID treats the patch tokens of a single image as a sample from a manifold and estimates their local intrinsic dimension (LID) at several depths of a frozen vision transformer. This 12-dimensional, training-free signal enters a fine-tuned detector through a confidence gate whose strength is calibrated purely in-distribution. On a 16-axis cross-generator benchmark, GLID reaches 0.805 mean AUC, first among retrained state-of-the-art baselines and never significantly behind the strongest of them on any axis. It lifts the generation axes by +0.084 AUC while moving reenactment by only -0.005. Two empirical laws explain the design. First, forged faces bend the token manifold at family-specific depths: GAN artifacts peak at the last layer, diffusion artifacts peak mid-network, and the pattern survives four backbones, three dimension estimators, and non-face imagery. Second, fine-tuning absorbs auxiliary gains exactly where training data covers: injecting 1% target-family images erases a +0.100 gain, so geometric signals matter precisely where data is unavailable. The deterministic signal also cuts the cross-seed spread of accuracy 5.5x. Code, preregistered analysis gates, and per-image scores accompany the paper.

cs.CR↗

arXiv2026

Thinking in Video: Can Video Generators Really Reason About the Real World?

Recent advances in world models and video generation have given rise to an emerging reasoning paradigm that leverages video generative models to simulate, predict, and reason about real-world dynamics. We redefine this paradigm as Thinking in Video, where video is not merely an output artifact but a medium for constructing, extending, and verifying causal thought. However, this promise remains unverified: convincing rollouts may reflect memorized appearances rather than causal understanding, while existing metrics separate perceptual fidelity from semantic logic. To evaluate whether video generators support such reasoning, we introduce the Causal-Generative Dual-Judge (CGDJ), auditing World Model Consistency from two perspectives. Explicit Causal Perception tests whether a generator reads a video scenario as a reasoning problem through spatio-temporal flattened visual question answering, while Implicit Generative Perception-Prediction Gap evaluates whether it renders the causal consequence as a consistent future video. Applying CGDJ to representative open- and closed-source generators reveals a clear Perception-Prediction Gap: open-source models produce plausible dynamics despite near-zero explicit causal perception, whereas advanced closed-source systems show stronger but still limited alignment between reasoning and generation. Further analysis exposes audio-visual misalignment, where models verbalize correct causal logic more reliably than they render it, challenging the "world simulator" narrative.

cs.CV↗

arXiv2026

General and scalable vapor etching and transformation platform for two-dimensional materials

Two-dimensional (2D) nanomaterials derived from non-van der Waals (non-vdW) solids offer exceptional physicochemical properties, yet their synthesis is impeded by intrinsic covalent/metallic bonding and high surface reactivity of the precursors. Here, we report a general vapor-phase etching and transformation platform for producing a library of 36 2D carbides, nitrides, and carbonitrides, exhibiting electrical conductivities spanning six orders of magnitude. Using reactive vapors like hydrogen chloride, we selectively remove A-layers from MAX phases to yield well-defined layers (MXenes), including previously inaccessible semiconducting Hf2CTx. By varying the reactive vapor environment, MXenes can be engineered at X-site and surface-termination site and even be transformed into non-vdW layers such as 2D MAX phases. This general and scalable vapor-phase platform reframes 2D material synthesis, opening new avenues for various applications.

cond-mat.mtrl-sci↗

arXiv2026

Motion-Conditioned Multi-View Fusion for Myocardial Infarction Localization from Echocardiography

Myocardial infarction (MI) remains a leading cause of mortality worldwide. Echocardiography (Echo) is a widely available modality for MI assessment, where regional wall motion abnormality is a key indicator. Prior learning based methods for myocardial motion analysis often use handcrafted descriptors or densely supervised estimation, but the need for extensive annotation limits applicability. Foundation models have recently improved vision-based Echo analysis; however, most methods operate on single views and segment-level localization remains unreliable under view-dependent ambiguity, especially in apical views. To address this, we propose MCF-Net, a novel motion-guided multi-view fusion framework that fuses myocardial motion cues with foundation model representations to localize infarction. Visual features are extracted using EchoPrime, a pretrained Echo foundation model shared across dual views. Cardiac motion is modeled with extremely sparse supervision: a single annotated template frame is transferred across videos to initialize point tracking, avoiding dense labels. Motion-derived segment-aware soft masks provide coarse spatial priors that selectively enhance features for challenging myocardial segments. A motion-conditioned fusion mechanism then integrates motion and vision across views, refining predictions without overriding strong appearance cues. On segment-level MI localization, MCF-Net achieves 72.4\% F1 and 84.9\% accuracy, outperforming state-of-the-art motion-only, vision-only, and fusion baselines.

cs.CV↗

arXiv2026

Comprehensive study of timing resolution in plastic scintillator detectors with wavelength-shifting fiber and silicon photomultiplier readout

We present a comprehensive study of the timing resolution achievable in plastic scintillator detectors read out through wavelength-shifting (WLS) fibers coupled to silicon photomultipliers (SiPMs), combining a semi-analytical framework, toy Monte Carlo validation, and full Geant4 optical photon simulation. The analytical model traces the complete photon detection chain: scintillation emission, WLS fiber re-emission, optical transit time dispersion, SiPM single-photon time resolution, and electronics quantization. It expresses the timing resolution $\sigt$ as a function of the detected photoelectron yield $\Npe$, scintillator decay constants ($\taur$, $\taud$), WLS re-emission time ($\tauwls$), fiber numerical aperture, detector geometry, and readout electronics parameters. The analytical predictions are validated at two levels. First, toy Monte Carlo simulations ($2\times 10^5$ events per parameter point across 80 grid points spanning 8 fiber types and $\Npe$ from 5 to 200) achieve analytical-to-MC agreement of $0.9997 \pm 0.0015$. Second, full Geant4 optical photon simulations track the entire scintillation, wavelength-shifting, and fiber transport chain in realistic detector geometries, confirming the analytical timing predictions and providing first-principles photoelectron yield calibration. A comprehensive parameter scan covering 7 scintillator materials, 8 WLS fiber types, 5 SiPM models, 5 electronics configurations, 3 readout topologies, and 3 boundary conditions produces quantitative design maps and lookup tables for detector optimization.

physics.ins-det↗

arXiv2026

LEGATO 2: Toward Multimodal Sheet Music Recognition and Understanding

We propose a novel pipeline, Legato 2, for extracting symbolic notation and semantic knowledge from images of sheet music. Legato 2 features the first large-scale neural model for optical music recognition (OMR) to operate sequentially on a system-by-system basis, following the horizontal lines of notation as they are read on the page, rather than treating the page as an undifferentiated image, enabling better scaling to arbitrarily long inputs. It is also the first OMR model capable of generating symbolic transcriptions that include embedded textual content, such as titles and annotations. The pipeline combines system-level segmentation with an autoregressive vision-LM to capture both local notation details and score structure. Across multiple datasets, Legato 2 consistently outperforms prior state of the art. We also show that symbolic transcriptions complement visual inputs for frontier language models, improving their interpretation of dense musical documents. Legato 2 establishes new state-of-the-art performance in both OMR and downstream sheet music understanding.

cs.CV↗

arXiv2026

Self-Supervised Implicit CEST Reconstruction via Physics-Informed Lorentz Encoding

Multi-Pool Chemical Exchange Saturation Transfer (CEST) MRI provides valuable metabolic information but is clinically limited by long acquisition times. Although sparse sampling reduces scanning time, reconstructing high-resolution Z-spectra from limited data remains an ill-posed inverse problem. Conventional interpolation and generic Implicit Neural Rep-resentations (INRs) often lack physical constraints, leading to spectral artifacts and physically invalid signals. To address this, we propose Lorentz Encoding (LE), a physics-informed framework that formulates CEST reconstruction as a self-supervised reconstruction task via implicit continuous coordinate learning. Unlike generic positional encodings, LE regularizes the continuous spectral mapping by projecting sparse coordinates into a physically constrained space governed by a combination of parametric Lorentzian profiles with learnable basis functions. This mechanism effectively reduces noise and enforces consistency with physical models. Experiments on in vivo human brain data demonstrate that LE significantly outperforms state-of-the-art methods. Specifically, under a 39-point sampling strategy, LE achieves a PSNR of 57.58 dB and an SSIM of 0.9994. Furthermore, the learned physics-informed encodings form a continuous, geometrically ordered trajectory in the latent space, ensuring accurate quantitative metabo-lite mapping (APT, NOE, MT).

cs.LG↗

arXiv2026

The MIRI Early Obscured-AGN Wide Survey (MEOW): A Population of Hidden AGN at $z \gtrsim 5$ Revealed by JWST/MIRI Imaging

We present the MIRI Early Obscured-AGN Wide Survey (MEOW), a JWST/MIRI imaging survey designed to detect dust-obscured active galactic nuclei (AGN) across cosmic time, with a particular focus on the high-redshift universe at $z \gtrsim 5$. MEOW observes the GOODS-N and GOODS-S fields with 43 pointings covering 95 arcmin$^2$ with the F1000W and F2100W filters, reaching depths of 0.5 and 3.6 $μ$Jy ($5σ$), respectively. Using spectral energy distribution (SED) modeling combining MEOW photometry with archival HST, JWST/NIRCam, and SCUBA-2 data, we identify a sample of 16 MIRI-selected AGN at $z = 4.5$--$7.2$ (12 spectroscopically confirmed), spanning bolometric luminosities of $L_{\rm bol} = 10^{44.6}$--$10^{46.4}$~erg~s$^{-1}$. Twelve of the 16 AGN are newly identified in this work, including at least five narrow-line AGN representing the obscured population to which broad-line spectroscopic searches are insensitive. Two broad-line AGN exhibit markedly different mid-infrared emission properties, consistent with one being a little red dot (LRD) and the other either a typical AGN or an LRD with unusually strong hot-dust emission. The MIRI-selected AGN bolometric luminosity function at $z = 4.5$--$6$ yields number densities comparable to those of broad-line AGN and LRDs, suggesting that obscured AGN contribute significantly to the total AGN census at these epochs. The narrow-line AGN reside in diverse host environments, with evidence for both circumnuclear and host-galaxy-scale obscuration, pointing to multiple physical mechanisms at work. These results establish JWST/MIRI imaging as an indispensable component of a multi-faceted approach to a complete census of early supermassive black hole growth.

astro-ph.GA↗

arXiv2026

Reconsidering Overthinking: Penalizing Internal and External Redundancy in CoT Reasoning

Large reasoning models (LRMs) often exhibit overthinking, producing verbose Chain-of-Thought (CoT) traces that increase inference cost and obscure the underlying reasoning process. Existing CoT compression methods mainly rely on global length rewards, which conflate necessary intermediate reasoning with redundant text and may therefore compromise reasoning fidelity. This paper revisits overthinking from a semantic-efficiency perspective and decomposes CoT redundancy into two distinct forms: internal redundancy, defined as informational stagnation before the first correct answer, and external redundancy, defined as superfluous continuation after the first correct answer. Based on this decomposition, we propose a dual-penalty reinforcement learning framework that separately optimizes reasoning progress and termination behavior. Specifically, a sliding-window semantic similarity metric penalizes low-progress reasoning segments, while a normalized external-redundancy metric discourages post-answer continuation. Experiments on GSM8K, MATH500, and AIME24 across different model scales show that our method reduces average reasoning length by 41.3% on the 1.5B model and 40.1% on the 7B model, while preserving competitive accuracy and achieving the best overall accuracy-efficiency score among evaluated baselines. The learned compression behavior further transfers to out-of-domain reasoning tasks, including GPQA and LiveCodeBench. More importantly, our analysis reveals a clear asymmetry between the two redundancy types: external redundancy can be largely removed with little performance loss, whereas internal redundancy compression follows a sensitive accuracy-efficiency trade-off. These results suggest that effective CoT compression should optimize semantic efficiency rather than sequence length alone, offering a principled route toward more concise, efficient, and interpretable LRMs.

cs.AI↗

arXiv2026

Learning from Acquisition: Metadata-driven Multimodal Pre-training for Cardiac MRI

Cardiac magnetic resonance imaging (CMR) routinely records structured acquisition metadata, yet most CMR foundation models rely primarily on image-only pre-training and leave this naturally available source of weak semantic supervision largely underexplored. We propose MetaCLIP-CMR, a metadata-driven framework based on Contrastive Language--Image Pre-training (CLIP), which converts imaging modality, anatomical view, scanner vendor, field strength, and scanner model into textual supervision for CMR representation learning. The pretrained image encoder is evaluated on imaging modality classification, cine view classification, and cardiac segmentation. MetaCLIP-CMR achieves 86.8% modality accuracy and 86.5% cine view accuracy, clearly outperforming ImageNet and masked reconstruction initialisations. For downstream cardiac segmentation, MetaCLIP-CMR consistently obtains the highest Dice score across the evaluated ACDC and M&Ms cine short-axis (SAX) settings under both full-data and 20% fine-tuning regimes. Compared with recent image-focused large-scale CMR pre-training models, MetaCLIP-CMR achieves comparable ACDC segmentation performance, while requiring less than 1% of their pre-training image scale. These results suggest that metadata learning offers a natural and easy-to-use strategy for transforming routinely recorded acquisition information into effective supervision for foundation-level CMR representation learning, highlighting the promise of metadata-driven multimodal pre-training.

cs.CV↗

arXiv2026

Yuvion VL: A Multimodal Foundation Model for Adversarial Content and AI Safety

General-purpose models often struggle to reliably identify and understand real-world multimodal risks, largely due to the inherent multimodal adversarial nature of content and AI safety. We present Yuvion VL, a family of multimodal large language models purpose-built for content and AI safety, with both instruction-tuned and reasoning-oriented variants. Yuvion VL addresses this gap by treating safety as an inherently adversarial and multimodal problem and designing the entire pipeline around adversarial robustness. For data construction, we develop an automated pipeline integrating adversarial-aware data synthesis with multi-stage quality control, producing large-scale, high-quality multimodal samples augmented with domain knowledge and reasoning annotations. For training, we adopt a three-stage pipeline that includes continued pretraining for risk-concept cross-modal alignment, instruct post-training for production-grade safety tasks, and reasoning post-training for enhanced interpretability and performance in complex tasks. We further introduce Confuse-then-Contrast Fine-Tuning, a contrastive framework that mines model-specific confusions and constructs multi-image contrastive groups to enforce explicit discrimination of fine-grained visual-semantic elements, enabling the model to distinguish between visually similar cases with different safety implications in adversarial safety tasks. To support rigorous evaluation, we further introduce Yuvion VL RiskEval (YVRE), a collection of benchmarks covering diverse open and internal evaluations, with a focus on content and AI safety, adversarial robustness, and real-world capability requirements. Experiments show that Yuvion VL-32B achieves industry-leading safety performance, surpassing comparably sized open-source models and best closed-source commercial models, while maintaining comparable general capabilities.

cs.CV↗

arXiv2026

Yuvion LLM: An Adversarially-Aware Large Language Model for Content And AI Safety

As large language models are increasingly deployed in real-world systems, safety failures can still lead to harmful outputs and dangerous misuse. We argue that the essence of safety is adversarial: many failures arise not from natural inputs alone, but from strategic attempts to evade model policies and safeguards. However, existing general-purpose model development largely overlook this adversarial nature, and often remain insufficient for realistic safety scenarios involving planning, tool use, and multi-step reasoning, causing measured safety performance to overestimate real deployment robustness. To address this gap, we present Yuvion LLM, a large language model built for adversarially robust content safety and broader AI safety. Yuvion LLM treats adversarial robustness and agentic capability as first-class objectives. Its pipeline combines adversarially aware data construction, knowledge-enhanced continued pretraining, and policy-grounded multi-task safety post-training, including risk-aware supervised fine-tuning and reinforcement learning-based policy optimization, together with safety-aware agentic reinforcement learning for tool use and multi-step reasoning in complex safety scenarios. We further introduce the Yuvion LLM RiskEval (YLRE), a collection of 93 benchmarks across four evaluation categories, covering diverse open and internal evaluations with a focus on safety, adversarial robustness, and real-world capability requirements. Across these evaluations, Yuvion LLM demonstrates clear advantages on safety-focused benchmarks and particularly strong robustness under adversarial conditions, while maintaining solid overall capability. Notably, Yuvion-8B outperforms most state-of-the-art baselines, including substantially larger models such as GPT-5.4 and Qwen3-MAX, on several safety tasks.

cs.CL↗

arXiv2026

HiPath: Hierarchical Vision-Language Alignment for Structured Pathology Report Prediction

Pathology reports are structured, multi-granular documents encoding diagnostic conclusions, histological grades, and ancillary test results across one or more anatomical sites; yet existing pathology vision-language models (VLMs) reduce this output to a flat label or free-form text. We present HiPath, a lightweight VLM framework built on frozen UNI2 and Qwen3 backbones that treats structured report prediction as its primary training objective. Three trainable modules totalling 15M parameters address complementary aspects of the problem: a Hierarchical Patch Aggregator (HiPA) for multi-image visual encoding, Hierarchical Contrastive Learning (HiCL) for cross-modal alignment via optimal transport, and Slot-based Masked Diagnosis Prediction (Slot-MDP) for structured diagnosis generation. Trained on 749K real-world Chinese pathology cases from three hospitals, HiPath achieves 68.9% strict and 74.7% clinically acceptable accuracy with a 97.3% safety rate, outperforming all baselines under the same frozen backbone. Cross-hospital evaluation confirms generalisation with only a 3.4pp drop in strict accuracy while maintaining 97.1% safety.

cs.CV↗

arXiv2026

C^2GR: Coupled Comprehensive Generative Replay for a Continually Learnable Universal Segmentation Model

Universal segmentation models exhibit significant potential for diverse tasks involving different imaging modalities and segmentation objectives. Task-Incremental Learning provides a privacy-preserving approach to continually evolve a universal model on tasks from sequentially-arriving medical departments. However, training the model solely on the incoming task induces forgetting on past tasks, since consecutive tasks exhibit concurrent shifts in image appearance and segmentation objective. To address this problem, we propose a novel Coupled Comprehensive Generative Replay (C^2GR) framework that simultaneously synthesizes image-mask pairs of previous tasks to mitigate forgetting under concurrent appearance and objective shifts. This requires preserving image-mask correspondence for structure-realistic generation and bridging asynchronous optimization of the generator and segmentor for segmentation-oriented generation. Specifically, we propose a Bayesian Joint Diffusion (BJD) method that formulates the correspondence as conditional distributions optimized via conditional denoising. Furthermore, we develop a Relation-aware Unified Prompt Synchronization (RUPS) scheme to simultaneously modulate the generator and segmentor via a shared task-relation-aware prompt for synchronizing their optimization. Experiments on 20 tasks spanning diverse modalities and objectives demonstrate that C^2GR exhibits only a 2.44% drop in overall performance compared to joint training with all task data, effectively alleviating forgetting from the concurrent shifts. Our code will be made publicly available at https://github.com/mar-cry/C2GR.

cs.CV↗

arXiv2026

SegDINO: Introducing Multi-Scale Structure into DINO for Efficient Medical Image Segmentation

Self-supervised DINO models provide strong transferable visual representations, yet applying them directly to image segmentation remains challenging. Existing approaches commonly rely on heavy decoders with complex upsampling, introducing substantial parameter and computational overhead. We observe that introducing scale into DINO features is far more critical than increasing decoder capacity. In this work, we present SegDINO, an efficient segmentation framework that integrates a DINOv3 backbone with lightweight scale modeling. SegDINO introduces Token Pyramid Adaptation (TPA) to reorganize intermediate DINO features into a pseudo multi-scale hierarchy, and Scale-Aware Decoding (SAD) for efficient intra-scale refinement and top-down multi-scale propagation. We further curate PanCT, a new CT dataset containing 284 patients with expert-annotated pancreatic tumors, to assess SegDINO's ability to handle difficult small-lesion cases. Extensive experiments on PanCT and three public benchmarks demonstrate that SegDINO achieves state-of-the-art results with high efficiency. The code is available at https://github.com/script-Yang/segdino_v2.

cs.CV↗

arXiv2026

MEGA and SMILES Find Fewer Dusty Galaxies than Expected at Cosmic Noon

We present infrared (IR) luminsosity functions (LFs) and resulting star formation rate densities using the JWST Mid-infrared Instrument (MIRI) observations from the MIRI EGS Galaxy and AGN (MEGA) survey and Systematic MIRI Legacy Extragalactic Survey (SMILES). JWST allows us to perform a robust analysis on the faint end of the IR LF beyond the local universe. We directly measure the 7.7$μ$m polycyclic aromatic hydrocarbon (PAH) feature using either F1000W, F1500W, or F2100W photometry. This results in a sample of 634 galaxies across the two surveys covering an area of 105 arcmin$^2$ ($\sim$70 in the EGS and $\sim35$ in the GOODS-S/HUDF fields) and spanning $0.2<z<2$. We convert the 7.7$μ$m PAH luminosity to total IR luminosity, resulting in LFs that are two orders of magnitude fainter than previous studies. In contrast to previous extrapolations based on shallower observations, we find a strong flattening in the faint end of the LF with an average slope of $α\sim0.147$. This indicates that less luminous galaxies do not have as much dust obscured star formation as predicted. We measure the star formation rate density (SFRD) by integrating our new IR LFs and find a slightly lower SFRD in all redshift bins than previous studies made with ALMA, Herschel, and Spitzer. We also measure the contribution to the SFRD as a function of luminosity and confirm that LIRGs and ULIRGs remain the dominant contributors to the dust-obscured star formation at $z\sim1-2$.

astro-ph.GA↗

arXiv2026

TimeROME-DLM: Temporal Causal Tracing and Low-Rank Inference-Time Knowledge Editing for Masked Diffusion Language Models

Masked diffusion language models (MDLMs) such as LLaDA now rival autoregressive (AR) LLMs, but every existing knowledge-editing and unlearning method (ROME, MEMIT, etc.) targets AR transformers and either makes assumptions that fail under iterative denoising, or requires gradient updates whose backward-pass activations cost tens of GB of extra VRAM and which collapse MDLMs at standard learning rates. We introduce TimeROME-DLM, the first training-free, gradient-free, inference-time knowledge-editing framework for MDLMs. It couples two components: a Temporal Indirect Effect (TIE) causal-tracing protocol that identifies, for each fact, the coordinate whose intervention most strongly drives the object prediction at later denoising steps; and a closed-form, low-rank residual edit memory that aggregates subject keys and target deltas across all forget facts and applies a single ridge-regularised update at that coordinate at every diffusion forward, with sparsification to limit utility spillover. Backbone weights stay frozen; only three hyperparameters (alpha, lambda, q) are tuned on a small validation split. On TOFU forget01 with TOFU-finetuned LLaDA-8B-Base, TimeROME-DLM cuts forget-set log-probability by roughly 83 nats. The same configuration transfers to LLaDA-8B-Instruct, Dream-7B, MMaDA-8B, DiffuLLaMA-7B, and LLaDA-MoE-1.4B. It keeps retain-set log-probability nearly flat (within ~1 nat at the utility-safe operating point) across 50 sequentially inserted facts, delivers a four- to fourteen-fold wall-clock speedup with zero additional VRAM over the strongest converged training-time baseline, and scales sub-linearly to 400 facts. TimeROME-DLM closes the locate-then-edit gap between AR LLMs and MDLMs at a fraction of the computational cost.

cs.LG↗