2026-06-09 AI 리서치 브리핑

Neural network pruning aims to compress models for efficient deployment, yet two fundamental challenges remain. First, many methods rely on per-component importance scores, selecting filters or neurons independently and ignoring redundancy: the retained set may include multiple components capturing similar discriminative patterns while missing others entirely. Second, determining per-layer pruning ratios typically requires manual, architecture-specific tuning with no principled stopping criterion. We propose CoSeP (Complementary Separability Pruning) to address both issues. Rather than scoring components in isolation, CoSeP represents each component by its class-separability profile across all class pairs, computed via Jeffries--Matusita distances. This defines a separability space in which nearby components are potentially redundant and distant components capture complementary information. CoSeP selects a compact set of representatives in this space: components are grouped via k-medoids clustering, candidate subset sizes are evaluated using the Mean Simplified Silhouette, and a knee-detection criterion automatically determines how many components to retain. Across CIFAR-10, CIFAR-100, and ImageNet-1K, on ResNet, VGG, MobileNet, and DenseNet architectures, CoSeP matches or improves accuracy while reducing FLOPs, with measured wall-clock inference-time reductions of up to 20%. For example, it achieves a +0.66% top-1 accuracy gain with 2.30x FLOPs reduction on ResNet-50/ImageNet-1K, and a 0.37% gain with 2.59x FLOPs reduction on VGG-16/CIFAR-10. These results demonstrate that modeling complementarity in class-separability space provides an effective and principled approach to pruning.

Your UnEmbedding Matrix is Secretly a Feature Lens for Text Embeddings에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

SoCRATES: Towards Reliable Automated Evaluation of Proactive LLM Mediation across Domains and Socio-cognitive Variations에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

GENEB: Why Genomic Models Are Hard to Compare에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

MMAE: A Massive Multitask Audio Editing Benchmark에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

AnchorWorld: Embodied Egocentric World Simulation with View-based Evolution Customization에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

Grading knee osteoarthritis (KOA) on plain radiographs is poorly reproducible across readers. A single-grade disagreement on the Kellgren-Lawrence (KL) scale can alter surgical management or redirect a patient from conservative therapy to intra-articular injection. Meanwhile, deep learning models that outperform human readers often offer no explanation for their decisions. We present Knee-xRAI, a pipeline that decomposes the grading process by mimicking clinical radiological workflows. It independently measures joint space narrowing (JSN), osteophytes, and subchondral sclerosis, then combines these findings into an explainable KL grade. Specifically, a U-Net++ architecture quantifies JSN via contour segmentation, an SE-ResNet-50 multi-task network grades osteophytes per anatomical site on the OARSI scale, and a hybrid texture-CNN detects binary sclerosis. This pipeline yields a 50-dimensional feature vector evaluated via an XGBoost-SHAP classifier (Path A, audit) and a ConvNeXt hybrid predictor (Path B, deployed). On 8,260 OAI-derived radiographs, the JSN module achieved a Dice score of 0.8909 and an mJSW ICC of 0.8674. Path A reached a QWK of 0.6294 and an AUC of 0.8046, confirming the structured feature vector carries substantial diagnostic signal. Path B achieved a QWK of 0.8436 and an AUC of 0.9017. SHAP analysis identifies JSN as the dominant feature, with osteophytes adding a consistent increment and sclerosis contributing marginally. Removing JSN evidence collapses KL3-KL4 recall while early grades remain intact, aligning with the KL diagnostic criteria. Knee-xRAI grounds every prediction in an auditable chain of measured radiographic findings, providing clinical transparency at the point of care.

Data pruning (DP), as an oft-stated strategy to alleviate heavy training burdens, reduces the volume of training samples according to a well-defined pruning method while striving for near-lossless performance. However, existing approaches, which commonly select highly informative samples, can lead to biased gradient estimation compared to full-dataset training. Furthermore, the analysis of this bias and its impact on final performance remains ambiguous. To address these challenges, we propose OrderDP, a plug-and-play framework that aims to obtain stable, unbiased, and near-lossless training acceleration with theoretical guarantees. Specifically, OrderDP first randomly selects a subset and then chooses the top-$q$ samples, where unbiasedness is established with respect to a surrogate loss. This ensures that OrderDP conducts unbiased training in terms of the surrogate objective. We further establish convergence and generalization analyses, elucidating how OrderDP affects optimal performance and enables well-controlled acceleration while ensuring guaranteed final performance. Empirically, we evaluate OrderDP against comprehensive baselines on CIFAR-10, CIFAR-100, and ImageNet-1K, demonstrating competitive accuracy, stable convergence, and exact control -- all with a simpler design and faster runtime, while reducing training cost by over 40%. Delivering both strong performance and computational efficiency, our method serves as a robust and easily adaptable tool for data-efficient learning. The code is publicly available at https://github.com/shengze-xu/OrderDP.

While Omni-modal Large Language Models (OLLMs) have demonstrated impressive capabilities in jointly processing audio and visual streams, their ability to strictly adhere to complex, multi-faceted user instructions remains largely unexplored. Existing benchmarks primarily focus on holistic video understanding or text-only instruction following, failing to capture the intricate interplay between modalities and user constraints. To bridge this gap, we introduce OmniCap-IF, the first comprehensive benchmark specifically designed to evaluate instruction-following capabilities in omni-modal captioning. OmniCap-IF incorporates a systematic framework that assesses captions on two dimensions: format correctness and content correctness. Our benchmark encompasses 50 distinct constraint types across pure visual, pure audio, and audio-visual modalities, while integrating Temporal Grounding to assess spatio-temporal precision. Extensive evaluations of prominent models on 1,920 high-quality samples reveal significant performance disparities. Furthermore, our analysis uncovers a critical "format-content tradeoff", demonstrating that increasing formatting complexity directly degrades models' omni-modal reasoning abilities. Finally, to advance the field, we curate a 54K instruction-tuning dataset, OmniCap-IF-54K and present OmniCaptioner-IF, which achieves notable improvements in both complex instruction adherence and general omni-modal captioning performance.

Unified and scalable Transformers have recently achieved remarkable success in modeling diverse phenomena traditionally associated with computer graphics, such as 3D visual effects, rendering processes, and motion in videos. In this work, we take a step further by investigating whether modern Transformer techniques can tackle the challenging task of cloth simulation. To this end, we present ClothTransformer, a framework that reformulates cloth simulation as autoregressive sequence modeling in a learned latent space. Existing neural cloth simulators are largely specialized to single scenarios, intrinsically coupled to the mesh discretization, and lack robust collision handling. Our approach addresses these limitations through three contributions: (1) a unified Transformer architecture that handles diverse scenarios -- body-driven garments, robotic manipulation, and free-fall collisions -- under a single model and achieves approximately $4$--$9{\times}$ lower error than prior state-of-the-art methods across all scenarios; (2) a scalable latent-space formulation that compresses arbitrary-resolution meshes into a fixed-size set of latent tokens, making temporal dynamics computation independent of mesh resolution; and (3) a diverse-scenario high-fidelity penetration-free dataset of ${\sim}$493.4k frames spanning all three settings, which enables a differentiable Continuous Collision Detection (CCD) module to suppress penetration artifacts. Project Page: https://yucrazing.github.io/clothtransformer/