AI/ML arXiv cs.AI

Same Stories, Different Journeys: From Social Comparison to Sensemaking in AI-Mediated Peer Career Exploration

JobMate is an interactive system that transforms social media career posts into AI agents to help job seekers engage in active, personalized career exploration.

AI/ML arXiv cs.AI

BackendForge: Benchmarking Agentic End-to-End Code Generation with Backend Services

BackendForge is a benchmark for evaluating the ability of agentic LLMs to generate deployable and behaviorally correct end-to-end backend services.

Cybersecurity Hacker News

Microsoft Patches a Record 570 Security Flaws

Microsoft has released patches for a record 570 security vulnerabilities across its products.

AI/ML arXiv cs.AI

How Do Practitioners Build SE Agents? Insights from a Mixed-Methods Study

A study explores how developers build LLM-based Software Engineering agents, highlighting a shift toward evaluation-driven development and new bottlenecks in coordination and review.

AI/ML arXiv cs.AI

The Nuts and Bolts of Natural Language to SQL Translation: A Systematic Analysis of Model Pipeline Optimisation Approaches and their Interactions

Research analyzes pipeline optimization for Natural Language to SQL translation, testing various intermediate representations and reranker models to improve efficiency.

AI/ML arXiv cs.AI

The Singularity Space: A Generative Diffusion Framework for Signal Representation

The Singularity Space framework represents signals via complex-plane singularities using a transformer-based diffusion model to better capture sharp transients in physical signals.

Hardware/Chips arXiv cs.AI

Edge Physical AI Deployment of Vision Transformers on Heterogeneous Edge GPU Targeting Autonomous Vehicles

H-FraDS is a hardware-aware scheduling methodology for deploying Vision Transformers on heterogeneous edge GPUs to improve latency and energy efficiency in autonomous vehicles.

AI/ML arXiv cs.AI

Efficient Online Proportional Sampling with Applications to Smoothed Online Learning

This paper presents a new data structure for efficient online proportional sampling in high-dimensional domains, providing sublinear regret guarantees for online learning.

AI/ML arXiv cs.AI

CGS: Configurable Graph Summarization with Bounded Neighborhood Loss and Query Support

The CGS framework provides configurable graph summarization with bounded neighborhood loss, supporting both lossless and lossy schemes for efficient graph queries.

AI/ML arXiv cs.AI

EquiFusion: Kinematics-Agnostic Human Motion Prediction via Equivariant Latent Diffusion

EquiFusion is a kinematics-agnostic human motion prediction model using equivariant latent diffusion, allowing for cross-dataset generalization and zero-shot motion prediction.

AI/ML arXiv cs.AI

MMA-Former: Multi-Window Mixture-of-Head Attention Transformer for Adaptive PNI Prediction in 3D MRI

MMA-Former is a 3D architecture using a Mixture-of-Head Attention mechanism to improve the prediction of perineural invasion in 3D MRI scans.

AI/ML arXiv cs.AI

Think When It Matters: Conditional VLM Reasoning for Social Navigation with RL Policies

HUMA is a hybrid architecture for social robot navigation that dynamically balances reactive RL policies with deep semantic reasoning from VLMs.

Hardware/Chips Hacker News

Unlocking the PSP's Dual Core Setup

A discussion on unlocking the dual-core capabilities of the PlayStation Portable (PSP).

Software Engineering Hacker News

Casio FX-870P Emulator

An emulator for the Casio FX-870P calculator.

Tech Business/VC Hacker News

Linear is always a lagging indicator

An opinion piece discussing why Linear is often a lagging indicator of progress.

AI/ML arXiv cs.AI

Weight-Adjusted Gradients Reveal Parameter Importance and Failure Modes in LLMs

Introduces Weight-Adjusted Gradients (WAG) to identify critical parameters in LLMs and understand failure modes.

AI/ML arXiv cs.AI

Abstractiveness Metrics for Evaluating Text Summarization: A Refined Formulation with Empirical Validation

Proposes new metrics (RA, SA, AR) to quantify abstractiveness in text summarization beyond surface-level metrics.

AI/ML arXiv cs.AI

Diachronic Sample Integration: Robust Tail-Risk Estimation with Generative Models

Introduces Diachronic Sample Integration (DSI) to improve tail-risk estimation in generative models using ensemble checkpoints.

AI/ML arXiv cs.AI

Distributed Agent System: Fault-Tolerant Collaboration Among Embodied Agents

Proposes a Distributed Agent System (DAS) for fault-tolerant collaboration among heterogeneous embodied AI agents.

AI/ML arXiv cs.AI

Auditing Belief-Conditioned LLM Agents in Hidden-Information Social Deduction Games

Presents an auditing framework for LLM agents in hidden-information social deduction games to analyze decision-making and belief states.