AI/ML arXiv cs.AI

The Maskability Index: Predicting Task-Objective Alignment in Pretrained Language Models

A research paper introducing the Maskability Index (MI) to predict how well task-objective alignment works with pretrained language models.

AI/ML arXiv cs.AI

Self-supervision drives representational convergence in medical foundation models more than clinical supervision

A research paper demonstrating that self-supervision, rather than clinical supervision, is the primary driver of representational convergence in medical foundation models.

AI/ML arXiv cs.AI

Sound Probabilistic Safety Bounds for Large Language Models

A research paper proposing a framework for calculating sound probabilistic safety bounds for large language models to prevent harmful outputs.

AI/ML arXiv cs.AI

Courteous Anticipation: Improving Long-Lived Task Planning in Persistent Shared Environments

A research paper presenting courteous anticipatory planning to improve long-lived task performance for robots sharing persistent environments.

Open Source arXiv cs.AI

Don't Trust the Label: License Laundering in AI Supply Chains

A study quantifying the phenomenon of 'license laundering' in AI supply chains, where license obligations are often lost during redistribution.

AI/ML arXiv cs.AI

Toward Reliable RGB-D Semantic Segmentation: Handling Missing Modalities via Condition Dropout

A research paper introducing Condition Dropout (ConD) to improve the robustness of RGB-D semantic segmentation models when sensor data is missing.

Other The Verge

The sci-fi movie that imagines AI isn’t so dystopian after all

A sci-fi film explores a future where AI is used to create robotic doubles of deceased loved ones.

AI/ML arXiv cs.AI

Active Inference as a Convex Markov Decision Process

Research proposes framing Active Inference as a convex Markov decision process to bridge it with modern reinforcement learning.

AI/ML arXiv cs.AI

Audio-Zero: Label-Free Self-Evolution for Fine-Grained Audio Reasoning

Audio-Zero is introduced as a label-free self-evolution framework for improving fine-grained reasoning in Large Audio Language Models.

AI/ML arXiv cs.AI

StreamHOI: Interaction-aware Temporal Memory Adaptation for Streaming HOI Video Generation

StreamHOI presents a low-latency streaming framework designed for long-duration human-object interaction video generation.

AI/ML arXiv cs.AI

The Quadrilateral Loss: Additivity as a Measurable Behavior of Dense Neural Networks

The paper introduces the quadrilateral loss to treat additivity in neural networks as a measurable, differentiable behavior.

AI/ML arXiv cs.AI

ELSAA: Efficient Low-Rank and Sparse Attention Approximation for Training Transformers

ELSAA is a proposed method for efficient Transformer training using a combination of low-rank and sparse attention approximations.

Cybersecurity arXiv cs.AI

Small, Free, and Effective: Orchestrating Open-Weight Small Language Models to Outperform Single LLM for Malware Analysis

A study investigates using orchestrated ensembles of small language models to perform malware analysis, outperforming single large models.

AI/ML arXiv cs.AI

DQAOA-GPT: AI-Accelerated Distributed Quantum Optimization for Combinatorial Problems

DQAOA-GPT combines distributed quantum optimization with GPT-based circuit generation to accelerate combinatorial problems.

AI/ML arXiv cs.AI

On the Systematic Challenges of Culturally Loaded Machine Translation: Dream of the Red Chamber as the Cultural Lens

Research investigates the challenges of translating culturally loaded expressions in large language models.

AI/ML arXiv cs.AI

Pushing the Frontier of Full-Song Generation: Hierarchical Autoregressive Planning Meets Flow-Matching Rendering

A unified framework for full-song generation is presented, utilizing hierarchical autoregressive planning and flow-matching rendering.

AI/ML Hacker News

Understanding the AI Economy

A Hacker News discussion exploring the economic dynamics and structures surrounding the artificial intelligence industry.

AI/ML Hacker News

Test-time training 3D reconstruction

A Hacker News discussion regarding the use of test-time training for 3D reconstruction tasks.

Other The Verge

The right-wing boomers protesting data centers have a lot in common with the left

Reports on protests against the expansion of AI data centers in Florida, highlighting a bipartisan opposition based on local concerns.

Other Ars Technica

White House report says Trump can usher in a "new golden age" of science

A critical look at a White House report suggesting a potential 'golden age' of science under a Trump administration.