All Articles
16186 articles total
Rashomon Alignment
Introduces Rashomon Alignment (RA), a geometrical measure to assess functional similarity between ML models regardless of specific data distributions.
From Deterministic to Generative Deep Learning for Urban Air Quality Reconstruction from Sparse Observations
A diffusion-based generative framework for reconstructing urban air quality from sparse observations, outperforming deterministic deep learning models.
Tools Are Not Islands: Set-Level Tool Retrieval for LLM Agents via Query-Conditioned Hyperedge Prediction
Presents HYSET, a hyperedge-based set-level tool retrieval system that evaluates the joint utility of candidate tool sets for LLM agents.
Shared Voxel-Map-Based Cooperative Indoor UAV Guidance with a Multi-Agent Soft Actor-Critic Controller
A cooperative indoor UAV guidance framework using a shared voxel-map world model and a multi-agent Soft Actor-Critic controller.
Image Quality Dependent Degradation for AI Systems
A design strategy for automated driving systems to maintain safety in poor image quality by adjusting confidence thresholds using normalizing flows.
SpectONet: A Physics-Guided Spectral Deep Operator Network for Euler-Bernoulli Beam Dynamics
Introduces SpectONet, a physics-guided spectral deep operator network for efficient and accurate Euler-Bernoulli beam vibration analysis.
Lowering the implementation barrier of neutral-atom quantum computing with agentic workflows
An agentic workflow designed to automate the translation of theoretical quantum protocols into experiments on neutral-atom quantum processors.
Sell Your AI Skills?
A Hacker News community discussion regarding the monetization and sale of AI-related skills in the current job market.
Contrastive Representation Learning of Longitudinal Disease Trajectories on Temporal Graphs
Proposes a contrastive representation learning framework using temporal graphs and graph neural networks to model disease trajectories from longitudinal clinical data.
A Human-in-the-Loop Corpus for LLM-Based Simplification of Scientific Summaries
Introduces a human-in-the-loop workflow and a new corpus for simplifying complex scientific summaries using LLMs to make research more accessible to non-specialists.
Construction-Driven Injection: Linguistically-Grounded Edit-Based Code-Mixing Fingerprints for Large Language Models
Presents a unified framework for injecting linguistically-grounded, code-mixing fingerprints into LLMs to enable black-box ownership verification and prevent unauthorized redistribution.
F(AI)2R: Who Did What, and Who Checked? Verifiable AI Provenance as an Executable Skill
Introduces F(AI)2R and 'aiprov', a PROV-O extension for creating machine-readable, verifiable provenance graphs for AI-assisted research artefacts.
OmniPhys: Knowledge-Graph-Driven Benchmarking and Collective Optimization for Physical Commonsense in Text-to-Image Generation
Introduces OmniPhys, a benchmark for testing physical commonsense in text-to-image models, and OmniPrompt, a framework to optimize physical consistency in generative AI.
KQFuzz: Knowledge-Guided Fuzzing for Quantum Libraries via Large Language Models
Presents KQFuzz, an LLM-based fuzzer that uses codebase knowledge to find bugs in quantum computing libraries like Qiskit, PennyLane, and Cirq.
Why Public Service AI Governance Frameworks Risk Failing in the Age of General-Purpose AI: Lessons from Policing
Discusses the failure risks of current AI governance frameworks when applied to general-purpose AI, specifically within the context of public policing.
MyMentorLLM: A psychotherapy GenAI environment with multimodal voice/text patients, trainees and experts for deliberate practice
Describes MyMentorLLM, a multimodal simulation environment using GenAI to provide deliberate practice for psychotherapists training in CBT.
DynaBridge: Dynamic Summary-Guided Cross-Task Multimodal Fusion for DASS-Structured Mental Health Assessment
Introduces DynaBridge, a multimodal fusion framework that uses LLM-generated summaries to assess mental health risks based on DASS-21 psychometric structures.
LLM Honeypot
A discussion on Hacker News regarding the creation of a honeypot designed specifically for Large Language Models.
Argus-Unified: Towards A Compact and Economical Unified Model for Image Understanding and Generation
Introduction of Argus-Unified, a compact multimodal model that unifies image understanding and generation with significantly lower compute and data requirements.
Multi-Scale Structural Features for Continual, Comprehensible Visual Recognition in a Developmental Learning Framework
A new visual feature representation for a developmental learning framework that enables continual, human-interpretable visual recognition without destructive forgetting.