Skip to main content

🏢 Johns Hopkins University

Wild-GS: Real-Time Novel View Synthesis from Unconstrained Photo Collections
·1766 words·9 mins· loading · loading
Computer Vision 3D Vision 🏢 Johns Hopkins University
Wild-GS achieves real-time novel view synthesis from unconstrained photos by efficiently adapting 3D Gaussian Splatting, significantly improving speed and quality over existing methods.
Where does In-context Learning \ Happen in Large Language Models?
·2289 words·11 mins· loading · loading
Natural Language Processing Large Language Models 🏢 Johns Hopkins University
LLMs learn tasks via in-context learning, but the task recognition location is unknown. This paper reveals that LLMs transition from task recognition to task performance at specific layers, enabling s…
Testing Semantic Importance via Betting
·4904 words·24 mins· loading · loading
AI Generated Multimodal Learning Vision-Language Models 🏢 Johns Hopkins University
This work presents statistically grounded methods to rank semantic concept importance in black-box models, using conditional independence testing for both global and local interpretations.
Stability and Generalization of Adversarial Training for Shallow Neural Networks with Smooth Activation
·201 words·1 min· loading · loading
AI Generated AI Theory Robustness 🏢 Johns Hopkins University
This paper provides novel theoretical guarantees for adversarial training of shallow neural networks, improving generalization bounds via early stopping and Moreau’s envelope smoothing.
Sample Complexity of Algorithm Selection Using Neural Networks and Its Applications to Branch-and-Cut
·428 words·3 mins· loading · loading
AI Theory Optimization 🏢 Johns Hopkins University
Neural networks enhance algorithm selection in branch-and-cut, significantly reducing tree sizes and improving efficiency for mixed-integer optimization, as proven by rigorous theoretical bounds and e…
ReGS: Reference-based Controllable Scene Stylization with Gaussian Splatting
·1952 words·10 mins· loading · loading
Computer Vision 3D Vision 🏢 Johns Hopkins University
ReGS: Real-time reference-based 3D scene stylization using Gaussian Splatting for high-fidelity texture editing and free-view navigation.
Proximal Causal Inference With Text Data
·4077 words·20 mins· loading · loading
AI Theory Causality 🏢 Johns Hopkins University
Unmeasured confounders hinder causal inference; this paper introduces a novel method using two pre-treatment text instances and zero-shot models to infer proxies for unobserved confounders, enabling p…
PaCE: Parsimonious Concept Engineering for Large Language Models
·2526 words·12 mins· loading · loading
Natural Language Processing Large Language Models 🏢 Johns Hopkins University
PaCE, a novel activation engineering framework, efficiently aligns LLMs by removing undesirable concepts from activations using sparse coding, achieving state-of-the-art performance while preserving l…
On the Stability and Generalization of Meta-Learning
·1358 words·7 mins· loading · loading
Machine Learning Meta Learning 🏢 Johns Hopkins University
This paper introduces uniform meta-stability for meta-learning, providing tighter generalization bounds for convex and weakly-convex problems, addressing computational limitations of existing algorith…
Off-Dynamics Reinforcement Learning via Domain Adaptation and Reward Augmented Imitation
·6706 words·32 mins· loading · loading
AI Generated Machine Learning Reinforcement Learning 🏢 Johns Hopkins University
DARAIL, a novel algorithm, tackles off-dynamics reinforcement learning by combining reward modification with imitation learning to transfer a learned policy from a source to a target domain. This app…
Neural Embeddings Rank: Aligning 3D latent dynamics with movements
·2698 words·13 mins· loading · loading
Machine Learning Deep Learning 🏢 Johns Hopkins University
Neural Embeddings Rank (NER) aligns 3D latent neural dynamics with movements, enabling cross-session decoding and revealing consistent neural dynamics across brain areas.
LP-3DGS: Learning to Prune 3D Gaussian Splatting
·2308 words·11 mins· loading · loading
Computer Vision 3D Vision 🏢 Johns Hopkins University
LP-3DGS learns to optimally prune 3D Gaussian splatting, achieving significant efficiency gains without compromising rendering quality via a trainable binary mask and the Gumbel-Sigmoid method.
Learning to Reason via Program Generation, Emulation, and Search
·1757 words·9 mins· loading · loading
Natural Language Processing Large Language Models 🏢 Johns Hopkins University
Language models excel at generating programs for algorithmic tasks, but struggle with soft reasoning. COGEX leverages pseudo-programs and program emulation to tackle these tasks, while COTACS searches…
Learning in Markov Games with Adaptive Adversaries: Policy Regret, Fundamental Barriers, and Efficient Algorithms
·349 words·2 mins· loading · loading
AI Generated Machine Learning Reinforcement Learning 🏢 Johns Hopkins University
Learning against adaptive adversaries in Markov games is hard, but this paper shows how to achieve low policy regret with efficient algorithms by introducing a new notion of consistent adaptive advers…
Learning Cut Generating Functions for Integer Programming
·1721 words·9 mins· loading · loading
AI Generated AI Theory Optimization 🏢 Johns Hopkins University
This research develops data-driven methods for selecting optimal cut generating functions in integer programming, providing theoretical guarantees and empirical improvements over existing techniques.
HDR-GS: Efficient High Dynamic Range Novel View Synthesis at 1000x Speed via Gaussian Splatting
·1800 words·9 mins· loading · loading
Computer Vision 3D Vision 🏢 Johns Hopkins University
HDR-GS: 1000x faster HDR novel view synthesis via Gaussian splatting!
Federated Black-Box Adaptation for Semantic Segmentation
·3023 words·15 mins· loading · loading
AI Generated Computer Vision Image Segmentation 🏢 Johns Hopkins University
BlackFed: Privacy-preserving federated semantic segmentation using zero/first-order optimization, avoiding gradient/weight sharing!
Efficient Large Multi-modal Models via Visual Context Compression
·2910 words·14 mins· loading · loading
AI Generated Natural Language Processing Large Language Models 🏢 Johns Hopkins University
LLaVolta significantly boosts multi-modal LLMs by using visual context compression, achieving substantial training cost reduction and enhanced inference efficiency without performance loss.
DiffNorm: Self-Supervised Normalization for Non-autoregressive Speech-to-speech Translation
·2806 words·14 mins· loading · loading
AI Generated Natural Language Processing Machine Translation 🏢 Johns Hopkins University
DIFFNORM boosts non-autoregressive speech-to-speech translation by normalizing speech data with a diffusion model and classifier-free guidance, achieving significant quality improvements.
Compositional Generalization Across Distributional Shifts with Sparse Tree Operations
·2516 words·12 mins· loading · loading
Machine Translation 🏢 Johns Hopkins University
Sparse Differentiable Tree Machine (sDTM) improves compositional generalization in neural networks by efficiently representing tree structures in vector space, enabling simultaneous symbolic and neura…