Deep Minds and Shallow Probes
arXiv:2605.11448v1 Announce Type: new Abstract: Neural representations are not unique objects. Even when two systems realize the same downstream computation, their hidden coordinates may differ by rep
Knowledge catalogue
arXiv:2605.11448v1 Announce Type: new Abstract: Neural representations are not unique objects. Even when two systems realize the same downstream computation, their hidden coordinates may differ by rep
arXiv:2605.11475v1 Announce Type: new Abstract: We propose a deep probabilistic unfolding model to address the classical quantized compressive sensing problem that leverages an unfolding framework to
arXiv:2605.11388v1 Announce Type: new Abstract: Humans intuitively solve complex problems by flexibly shifting among reasoning modes: they plan, execute, revise intermediate goals, resolve ambiguity t
arXiv:2605.12261v1 Announce Type: new Abstract: Many real-world tasks involve delayed effects, where the outcomes of actions emerge after varying time lags. Existing delay-aware reinforcement learning
arXiv:2605.11908v1 Announce Type: new Abstract: Softmax policy gradient converges at O(1/t), but its transient behavior near sub-optimal corners of the simplex can be exponentially slow. The bottlenec
arXiv:2312.02549v2 Announce Type: replace-cross Abstract: Temporal Language Grounding seeks to localize video moments that semantically correspond to a natural language query. Recent advances employ t
arXiv:2603.24652v3 Announce Type: replace Abstract: Network pruning, which removes less important parameters or architectures, is often expected to improve efficiency while preserving performance. How
arXiv:2605.11265v1 Announce Type: new Abstract: Dense prediction tasks in surgical computer vision, such as segmentation and surgical zone prediction, can provide valuable guidance for laparoscopic an
arXiv:2605.11109v1 Announce Type: cross Abstract: Self-supervised learning (SSL) has emerged as a promising approach to seismic data denoising as it does not require clean reference data. In this work
arXiv:2605.12138v1 Announce Type: cross Abstract: Generating realistic and user-preferred advertisements is a key challenge in e-commerce. Existing approaches utilize multiple independent models drive
arXiv:2510.27055v2 Announce Type: replace Abstract: We present Contamination Detection via Context (CoDeC), a practical and accurate method to detect and quantify training data contamination in large
arXiv:2507.12002v2 Announce Type: replace Abstract: Social interactions play a crucial role in shaping human behavior, relationships, and societies. It encompasses various forms of communication, such
arXiv:2605.12394v1 Announce Type: new Abstract: Training Neural Networks (NNs) without overfitting is difficult; detecting that overfitting is difficult as well. We present a novel Random Matrix Theor
arXiv:2605.12182v1 Announce Type: new Abstract: Dexterous teleoperation via Mixed Reality (MR)-based interfaces offers a scalable paradigm for transferring human manipulation skills to dexterous robot
arXiv:2605.11430v1 Announce Type: new Abstract: Diabetic Retinopathy (DR) is an art and science of recording and classifying the retinal images of a diabetic patient. DR classification deals with clas
arXiv:2304.09479v5 Announce Type: replace Abstract: We introduce a novel approach to single-view face relighting in the wild, addressing challenges such as global illumination and cast shadows. A comm
arXiv:2510.06719v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by grounding them in external knowledge. However, its application i
arXiv:2605.11601v1 Announce Type: new Abstract: Autoregressive language models are widely used for text evaluation, however, their left-to-right factorization introduces positional bias, i.e., early t
arXiv:2605.11758v1 Announce Type: cross Abstract: Unsupervised segmentation of pulmonary pathologies in CT remains an open challenge due to the absence of annotated multi pathology cohorts and the fai
arXiv:2602.06462v3 Announce Type: replace Abstract: Masked diffusion language models generate text through iterative masked-token filling, but terminal-only rewards on final completions provide coarse
arXiv:2502.20209v3 Announce Type: replace Abstract: In this paper, a novel dataset is introduced, designed to assess student attention within in-person classroom settings. This dataset encompasses RGB
arXiv:2411.00471v3 Announce Type: replace-cross Abstract: This paper introduces Dirichlet process mixtures of block g priors for model selection and prediction in linear models. These priors are exten
arXiv:2605.11005v1 Announce Type: new Abstract: Mixture-of-experts (MoE) architectures enable trillion-parameter LLMs with sparsely activated experts. Expert parallelism (EP) is a widely adopted MoE t
arXiv:2605.12379v1 Announce Type: new Abstract: Many reinforcement learning (RL) tasks have discrete action spaces, but most generative policy methods based on diffusion and flow matching are designed
arXiv:2605.12122v1 Announce Type: cross Abstract: Unlearning specific concepts in text-to-image diffusion models has become increasingly important for preventing undesirable content generation. Among
arXiv:2509.09838v2 Announce Type: replace Abstract: While Soft Actor-Critic (SAC) is highly effective in continuous control, its discrete counterpart (DSAC) performs poorly on challenging discrete-act
arXiv:2605.11210v1 Announce Type: new Abstract: We present a framework for distributed Pose Graph Optimization (PGO) by formulating the problem as a second-order continuous-time dynamical system evolv
arXiv:2602.15006v2 Announce Type: replace-cross Abstract: Gaussian Processes (GPs) are a powerful tool for probabilistic modeling, but their performance is often constrained in complex, large-scale re
arXiv:2509.26469v3 Announce Type: replace Abstract: Vector quantization is common in deep models, yet its hard assignments block gradients and hinder end-to-end training. We propose DiVeQ, which treat
arXiv:2605.12178v1 Announce Type: cross Abstract: World models enable agents to anticipate the effects of their actions by internalizing environment dynamics. In enterprise systems, however, these dyn
arXiv:2605.12055v1 Announce Type: new Abstract: Large Language Models (LLMs) achieve strong linguistic performance, yet their internal mechanisms for producing these predictions remain unclear. We inv
arXiv:2512.03199v2 Announce Type: replace Abstract: Biometric facial recognition models often demonstrate significant decreases in accuracy when processing real-world images, often characterized by po
arXiv:2510.04265v4 Announce Type: replace-cross Abstract: Pass@k is widely used to report the reasoning performance of LLMs, but it often produces unstable and potentially misleading rankings, especia
arXiv:2605.11683v1 Announce Type: new Abstract: Vision Transformers (ViTs) incur significant computational overhead due to the quadratic complexity of self-attention relative to the token sequence len
arXiv:2601.22334v2 Announce Type: replace Abstract: Differentially private stochastic gradient descent (DP-SGD) is the gold standard for training machine learning models with formal differential priva
arXiv:2605.11750v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models are often brittle in fine-grained manipulation, where minor action errors during the critical phases can rapidly e
arXiv:2505.18780v3 Announce Type: replace-cross Abstract: Achieving versatile humanoid locomotion with a single policy presents a critical scalability challenge. Prevailing methods often rely on disti
arXiv:2605.12183v1 Announce Type: new Abstract: Drifting Models have emerged as a new paradigm for one-step generative modeling, achieving strong image quality without iterative inference. The premise
arXiv:2605.11467v1 Announce Type: new Abstract: Reasoning models post-hoc rationalize answers they have already committed to internally, producing chains of *reasoning theater*: deliberative-looking s
arXiv:2506.17501v3 Announce Type: replace-cross Abstract: Following successful large-vessel recanalization via endovascular thrombectomy (EVT) for acute ischemic stroke (AIS), some patients experience
arXiv:2605.11569v1 Announce Type: cross Abstract: Accurate short-term demand forecasting is crucial to airline revenue management, yet most existing systems fail to meet this need because current mode
arXiv:2605.11567v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models predominantly adopt action chunking, i.e., predicting and committing to a short horizon of consecutive low-level act
arXiv:2605.11369v1 Announce Type: new Abstract: Generating physically plausible dynamic motions of human-object interaction (HOI) remains challenging, mainly due to existing HOI datasets limited to st
arXiv:2605.10993v1 Announce Type: new Abstract: Memory capacity is a critical factor determining the performance of Vision-Language-Action (VLA) models in long-horizon manipulation tasks. Existing mem
arXiv:2605.12140v1 Announce Type: new Abstract: Myocardial point tracking (MPT) has recently emerged as a promising direction for motion estimation in echocardiography, driven by advances in general-p
arXiv:2605.12196v1 Announce Type: new Abstract: Accurate ultra-short-term wind power forecasting is critical for grid dispatch and reserve management, yet remains challenging due to the non-stationary
arXiv:2605.12002v1 Announce Type: new Abstract: Text-guided inpainting has made image forgery increasingly realistic, challenging both SID and IFL. However, existing methods often struggle to point ou
arXiv:2605.11480v1 Announce Type: new Abstract: Reward fine-tuning has become a common approach for aligning pretrained diffusion and flow models with human preferences in text-to-image generation. Am
arXiv:2506.13163v3 Announce Type: replace Abstract: We study the Logistic Contextual Slate Bandit problem, where, at each round, an agent selects a slate of N items from an exponentially large set (of
arXiv:2605.12019v1 Announce Type: new Abstract: Human Activity Recognition (HAR) is a core task in pervasive computing systems, where models must operate under strict computational constraints while r
arXiv:2605.11526v1 Announce Type: cross Abstract: Training a deep neural network with the outputs of selected layers satisfying linear constraints is required in many contemporary data-driven applicat
arXiv:2510.09333v2 Announce Type: replace-cross Abstract: Evaluating generative models is challenging because standard metrics often fail to reflect human preferences. Human evaluations are more relia
arXiv:2605.11582v1 Announce Type: new Abstract: Large language models (LLMs) have shown remarkable potential in advertising scenarios such as ad creative generation and targeted advertising. However,
arXiv:2605.11019v1 Announce Type: new Abstract: Although large language models rely on chain-of-thought for complex reasoning, the overthinking phenomenon severely degrades inference efficiency. Exist
arXiv:2602.09725v3 Announce Type: replace-cross Abstract: Remote KV cache reuse fetches KV cache for identical contexts from remote storage, avoiding recomputation, accelerating LLM inference. While i
arXiv:2605.12297v1 Announce Type: new Abstract: Egocentric 3D hand pose estimation and gesture recognition are essential for immersive augmented/virtual reality, human-computer interaction, and roboti
arXiv:2605.12498v1 Announce Type: new Abstract: Reconstructing the absolute 3D pose and shape of the hands from the user's viewpoint using a single head-mounted camera is crucial for practical egocent
arXiv:2605.12335v1 Announce Type: cross Abstract: Electronic Health Records (EHR) contain rich longitudinal patient information and are widely used in predictive modeling applications. However, effect
arXiv:2605.12491v1 Announce Type: new Abstract: Vision Transformers (ViTs) achieve strong data-driven scaling by leveraging all-to-all self-attention. However, this flexibility incurs a computational
arXiv:2605.12079v1 Announce Type: new Abstract: Human-in-the-loop Bayesian optimization (HITL BO) methods utilize human expertise to improve the sample-efficiency of BO. Most HITL BO methods assume th