19
Validity of LLMs as data annotators: AMALIA on authority
A study evaluates the validity of Portugal's AMALIA language model when used as an automated data annotator for identifying moral authority.
Top stories in Research
19
Multi-Modal, Multi-Environment Machine Teaching for Robust Reward Learning
This paper explores optimal teaching strategies for inverse reinforcement learning to help autonomous agents learn robust reward functions across multiple environments and modalities.
19
Using AI-based Learning Assistants in Higher Education: A Large-Scale Descriptive Analysis
A large-scale study analyzes the usage patterns of an AI-based learning assistant among over 77,000 higher education students in distance learning programs.
19
Autonomous heterogeneous catalyst discovery with a self-evolving multi-agent digital twin
Researchers developed CatDT, a self-evolving multi-agent digital twin system designed to autonomously simulate and discover heterogeneous catalysts.
19
Training and Evaluating Diffusion Policies with Long Context Lengths
The study benchmarks the performance of robotic diffusion policies as their observation context length is increased, addressing memory limitations in dexterous manipulation.
19
On the Role of Conversational Timing in Synthetic Training Data for ASR
This research explores how controlling conversational timing parameters in synthetic training data affects the performance of automatic speech recognition systems.
19
IFAR: Multi-Perspective and Multi-Level Causal Discovery with LLMs
To improve the abductive reasoning capabilities of language models, researchers created the DeepAbduction dataset and proposed a multi-perspective causal discovery method called IFAR.
19
StateLinFormer: Stateful Training Enhancing Long-term Memory in Navigation
StateLinFormer is a linear-attention navigation model designed to improve long-term memory and persistent adaptation across extended interactions.
19
LLT: Local Linear Transformer for PDE Operator Learning
The Local Linear Transformer is proposed to improve partial differential equation operator learning by addressing the quadratic scaling and lack of local bias in standard attention mechanisms.
19
Deep Neural Networks as Discrete Dynamical Systems: Implications for Physics-Informed Learning
This study analyzes the relationship between feed-forward deep neural networks and discrete dynamical systems to offer new insights into physics-informed machine learning.
19
Answer Set Programming Energised! End-to-End Neurosymbolic Reasoning and Learning with ASP and Energy Based Models
This paper presents a neurosymbolic framework that integrates answer set programming with energy-based models to enable joint optimization in continuous latent spaces.
19
Profiling in PyTorch (Part 3): Attention is all you profile
Hugging Face has released the third part of its technical guide on profiling performance within the PyTorch framework, focusing on attention mechanisms.
19
World working together ‘as one ecosystem’ on RISC-V chip research: executive
An Alibaba executive emphasized that global collaboration on the open-source RISC-V chip architecture remains strong despite ongoing geopolitical tensions between the US and China.
18
Swapping Faces, Saving Features: A Dual-Purpose Pipeline for Pedestrian Privacy in ITS
A new face-swapping pipeline protects pedestrian privacy in autonomous vehicle training datasets while preserving critical behavioral features needed for trajectory prediction.
18
TMI: Text-to-Image Meets Image-to-Image for Complementary Data Synthesis to Boost Long-Tailed Instance Segmentation
Researchers designed a hybrid data synthesis pipeline that combines text-to-image and image-to-image generation to improve instance segmentation for rare object categories.
18
Workflow as Knowledge: Semantic Persistence for LLM-Mediated Workflows
This paper introduces a language-independent conceptual model that uses symbolic forms and object identity to manage semantic persistence in large language model workflows.
18
Nigeria Machinery: A Low-Resource Industrial Dataset with a Domain-Grounded Reasoning Layer
A new industrial dataset covering machinery usage and failures in Nigeria has been released to support quantitative analysis and language model training for African economies.
18
Pose-to-Biomechanics: Bridging 3D Human Pose Estimation and Biomechanical Attribute Prediction
The proposed BioModule framework aims to bridge the gap between 3D human pose estimation and the prediction of biomechanical attributes for clinical and sports applications.
18
APIVOT: Adaptive Planning with Interleaved Vision-Language Thoughts
The proposed APIVOT system combines linguistic and visual reasoning to help robots plan and execute complex, long-horizon tasks while respecting physical constraints.
18
Curvature-Weighted Capacity Allocation: A Minimum Description Length Framework for Layer-Adaptive Large Language Model Optimization
A new curvature-aware optimization framework uses minimum description length principles to guide layer-adaptive capacity allocation and pruning in large language models.
18
GenDA: Generative Data Assimilation on Complex Urban Areas via Classifier-Free Diffusion Guidance
GenDA is a generative data assimilation framework that uses graph-based diffusion models to reconstruct high-resolution urban wind fields from sparse sensor data.
18
Correcting Visual Blur Induced by Attention Distraction to Reduce Hallucinations: Algorithm and Theory
This paper proposes an algorithm and theoretical framework to mitigate object hallucinations in multimodal large language models by addressing attention distraction and visual blur.
18
CoCo-Fed: A Unified Framework for Memory- and Communication-Efficient Federated Learning at the Wireless Edge
CoCo-Fed is a federated learning framework designed to reduce memory usage and communication bandwidth constraints when deploying large neural networks at the wireless edge.
18
Time-to-Collision Based Dynamic Obstacle Avoidance Using Pretrained Vision Models for Robots in Unstructured Environments
The paper proposes a data-efficient method that uses pretrained vision models to help mobile robots avoid dynamic obstacles in unstructured real-world environments.