Posters
2024
Parameter Efficient Adaptation for Image Restoration with Heterogeneous Mixture-of-Experts
·3390 words·16 mins·
loading
·
loading
Computer Vision
Image Restoration
π’ Tsinghua University
AdaptIR: A novel parameter-efficient method for generalized image restoration using a heterogeneous Mixture-of-Experts (MoE) architecture that achieves superior performance and generalization.
Parameter Disparities Dissection for Backdoor Defense in Heterogeneous Federated Learning
·1504 words·8 mins·
loading
·
loading
Machine Learning
Federated Learning
π’ Wuhan University
FDCR defends against backdoor attacks in heterogeneous federated learning by identifying malicious clients via Fisher Information-based parameter importance discrepancies and rescaling crucial paramet…
Parameter Competition Balancing for Model Merging
·3629 words·18 mins·
loading
·
loading
AI Generated
Natural Language Processing
Large Language Models
π’ Harbin Institute of Technology
PCB-MERGING: A training-free model merging technique boosts performance by intelligently balancing parameter competition across multiple tasks.
Parallelizing Model-based Reinforcement Learning Over the Sequence Length
·2553 words·12 mins·
loading
·
loading
Machine Learning
Reinforcement Learning
π’ Zhejiang University
PaMoRL framework boosts model-based reinforcement learning speed by parallelizing model and policy learning stages over sequence length, maintaining high sample efficiency.
Parallelizing Linear Transformers with the Delta Rule over Sequence Length
·1639 words·8 mins·
loading
·
loading
Natural Language Processing
Large Language Models
π’ MIT
DeltaNet, a linear transformer boosting associative recall, now trains efficiently via a novel algorithm, scaling to large language models and outperforming existing linear baselines.
ParallelEdits: Efficient Multi-Aspect Text-Driven Image Editing with Attention Grouping
·3634 words·18 mins·
loading
·
loading
AI Generated
Multimodal Learning
Vision-Language Models
π’ University of Buffalo
ParallelEdits efficiently edits multiple image aspects simultaneously, guided by text prompts, surpassing sequential methods in speed and accuracy.
Paralinguistics-Aware Speech-Empowered Large Language Models for Natural Conversation
·2883 words·14 mins·
loading
·
loading
AI Generated
Natural Language Processing
Dialogue Systems
π’ Seoul National University
Unified Spoken Dialog Model (USDM) directly generates coherent spoken responses with natural prosody, surpassing cascaded baselines and enhancing natural conversation in speech-enabled LLMs.
Pandora's Box: Towards Building Universal Attackers against Real-World Large Vision-Language Models
·2651 words·13 mins·
loading
·
loading
Multimodal Learning
Vision-Language Models
π’ Peking University
Researchers developed a universal adversarial patch to fool real-world large vision-language models (LVLMs) across multiple tasks, without needing access to internal model details.
Panacea: Pareto Alignment via Preference Adaptation for LLMs
·2565 words·13 mins·
loading
·
loading
Natural Language Processing
Large Language Models
π’ Peking University
Panacea: a novel LLM alignment method achieving Pareto optimality via online preference adaptation using a single model.
PaGoDA: Progressive Growing of a One-Step Generator from a Low-Resolution Diffusion Teacher
·2966 words·14 mins·
loading
·
loading
Computer Vision
Image Generation
π’ Stanford University
PaGoDA: Train high-resolution image generators efficiently by progressively growing a one-step generator from a low-resolution diffusion model. This innovative pipeline drastically cuts training cost…
PageRank Bandits for Link Prediction
·2009 words·10 mins·
loading
·
loading
Machine Learning
Deep Learning
π’ University of Illinois Urbana-Champaign
PageRank Bandits (PRB) revolutionizes link prediction by framing it as a sequential decision-making problem, thus enabling the system to adapt to evolving data. Combining contextual bandits with PageR…
PaDeLLM-NER: Parallel Decoding in Large Language Models for Named Entity Recognition
·2356 words·12 mins·
loading
·
loading
Natural Language Processing
Named Entity Recognition
π’ ByteDance
PaDeLLM-NER massively accelerates LLM-based NER inference by up to 10x, enabling near real-time performance without accuracy loss.
PaCE: Parsimonious Concept Engineering for Large Language Models
·2526 words·12 mins·
loading
·
loading
Natural Language Processing
Large Language Models
π’ 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…
PACE: Pacing Operator Learning to Accurate Optical Field Simulation for Complicated Photonic Devices
·2611 words·13 mins·
loading
·
loading
Machine Learning
Deep Learning
π’ University of Texas at Austin
PACE, a novel neural operator, achieves unprecedented accuracy and speed in optical field simulation for complex photonic devices, surpassing existing methods by significantly reducing errors and boos…
PAC-Bayes-Chernoff bounds for unbounded losses
·358 words·2 mins·
loading
·
loading
AI Theory
Generalization
π’ Basque Center for Applied Mathematics (BCAM)
New PAC-Bayes oracle bound extends CramΓ©r-Chernoff to unbounded losses, enabling exact parameter optimization and richer assumptions for tighter generalization bounds.
P$^2$C$^2$Net: PDE-Preserved Coarse Correction Network for efficient prediction of spatiotemporal dynamics
·3055 words·15 mins·
loading
·
loading
AI Generated
Machine Learning
Deep Learning
π’ Gaoling School of Artificial Intelligence, Renmin University of China
P2C2Net: A physics-encoded neural network efficiently predicts complex spatiotemporal dynamics using coarse grids and limited training data, achieving state-of-the-art results.
OxonFair: A Flexible Toolkit for Algorithmic Fairness
·3793 words·18 mins·
loading
·
loading
AI Theory
Fairness
π’ University of Oxford
OxonFair: a new open-source toolkit for enforcing fairness in binary classification, supporting NLP, Computer Vision, and tabular data, optimizing any fairness metric, and minimizing performance degra…
OwMatch: Conditional Self-Labeling with Consistency for Open-world Semi-Supervised Learning
·2493 words·12 mins·
loading
·
loading
Machine Learning
Semi-Supervised Learning
π’ Hong Kong Polytechnic University
OwMatch: a novel framework conquering open-world semi-supervised learning challenges by combining conditional self-labeling and consistency for substantially enhanced accuracy across known and unknown…
Overfitting Behaviour of Gaussian Kernel Ridgeless Regression: Varying Bandwidth or Dimensionality
·1931 words·10 mins·
loading
·
loading
AI Generated
AI Theory
Generalization
π’ University of Chicago
Ridgeless regression, surprisingly, generalizes well even with noisy data if dimension scales sub-polynomially with sample size.
Overcoming the Sim-to-Real Gap: Leveraging Simulation to Learn to Explore for Real-World RL
·2419 words·12 mins·
loading
·
loading
Machine Learning
Reinforcement Learning
π’ University of California, Berkeley
Leveraging simulation for real-world RL is often hampered by the sim-to-real gap. This paper shows that instead of directly transferring policies, transferring exploratory policies from simulation d…