All Articles
16256 articles total
An Ontology for Machine Learning Interatomic Potentials
The MLIPs ontology provides a standardized OWL 2 DL framework to describe machine learning interatomic potentials, improving reproducibility in materials science.
Characterisation of Density-based FM generation methods in the context of Information Fusion
Research on density-based Fuzzy Measure (FM) generation methods to improve information fusion and provide confidence intervals for fusion results.
CAPT: A Multi-task Continuous Autoregressive Transformer enabling Cross-dataset and Cross-species Transfer for Calcium Population Dynamics
CAPT is a continuous autoregressive transformer designed for calcium population dynamics that generalizes across different datasets and species.
SeekJudge: A Practical Reward Framework for Reinforcement Learning in Computer-Use Agents
SeekJudge is a practical reward framework using specialized agents to provide accurate and efficient rewards for RL in computer-use agents.
TopoFE: topology-aware LLM-guided Automated Feature Engineering
TopoFE is a topology-aware evolutionary framework that uses LLMs to automate feature engineering for tabular data more efficiently.
Does every question mark deserve a Betteridge?
A discussion on the 'Betteridge's law of headlines', suggesting that any question mark in a headline implies a negative answer.
Hooray for the Sockets Interface
A technical appreciation of the Sockets interface and its role in network programming.
Beyond Greece and Rome
An exploration of intellectual and cultural history beyond the classical Greek and Roman traditions.
Chip stocks slide in US and Asia as AI jitters rattle investors
Chip stocks are declining in the US and Asia due to investor concerns regarding AI growth and sustainability.
ConsistencyGate: Preventing Memory Contamination in LLM Agents via Self-Consistency Admission Control
Introduces ConsistencyGate, a write-time admission control mechanism to prevent hallucinated facts from contaminating the external memory of LLM agents.
Reason Popper-ly: Patching In-Context Reasoning with Inductive Logic Programming
Presents Reason Popper-ly, a neurosymbolic framework using inductive logic programming to verify and correct in-context reasoning steps in LLMs.
Stress-testing large language model agents in a robotic chemistry laboratory
An evaluation of LLM agents in a physical robotic chemistry lab, revealing significant gaps in long-horizon planning and physical executability.
SymStep: Symbolic Step Verification for Logical Reasoning
Introduces SymStep, a symbolic step verification method that uses a constraint propagator to ensure logical consistency in LLM reasoning chains.
Structure over Depth: A Single-Block Spatio-Temporal Transformer for Multi-Entity Reasoning
Proposes a structured spatio-temporal transformer block that explicitly models interaction types to reduce the need for deep model stacks in multi-entity reasoning.
Compiler-Grounded Hierarchical Diagnosis for LLM-Based Triton Kernel Optimization
Presents a hierarchical diagnosis framework for optimizing Triton kernels on NPUs by linking runtime symptoms to compiler IR structure.
Multiple Mouse Cursors in Wayland
A discussion regarding the implementation of multiple mouse cursors within the Wayland display server protocol.
Coordinated Networking for On-Device Agent-Augmented Real-Time Communication
Researchers introduce HFS, a framework for on-device agent-augmented real-time communication that optimizes traffic between human video streams and AI agent data flows.
What Can Be Enforced? A Theory of Certified Runtime Safety for Tool-Using Agents
A theoretical study on certified runtime safety for AI agents that use tools, focusing on the limits of what can be enforced by runtime guardrails.
Physical AI Governance: From Theory to Practice Across Life Cycle
A comprehensive survey providing a governance framework and a five-stage lifecycle for 'Physical AI' systems (embodied AI) to ensure safety and trustworthiness.
How Well Can AI Generate Backlogs from App Mockups?
An evaluation of how well AI can generate software backlogs from app mockups, finding that hybrid prompting with architectural context improves precision.