AI/ML arXiv cs.AI

AutoJourn: Multi-Perspective Summarisation, Bias Detection and Bias Neutralisation for LLM-Generated News in Automated Journalism

AutoJourn is a system for multi-perspective news generation, bias detection, and neutralization using LLMs to promote responsible automated journalism.

AI/ML arXiv cs.AI

SWITi: Quantifying and Reducing Tiling Artifacts with Sliding Window Inner Tiling

SWITi is a test-time method designed to reduce tiling artifacts in large image predictions for neural networks, particularly in biomedical data.

AI/ML arXiv cs.AI

MedDDC-Eval: Diagnosis-Decoupled Evaluation of Multi-Turn Medical Consultation Agents

MedDDC-Eval introduces a diagnosis-decoupled testbed for evaluating multi-turn medical consultation agents to separate history quality from diagnosis generation.

AI/ML Hacker News

Are AI Labs Pelicanmaxxing?

A Hacker News community discussion regarding whether AI labs are 'Pelicanmaxxing', likely referring to a specific strategy or trend in model development or resource acquisition.

Hardware/Chips TechCrunch

Science Corporation’s vision-restoring chip wins EU approval

Science Corporation's vision-restoring chip has received EU approval, moving closer to commercial viability in the medical hardware space.

Tech Business/VC TechCrunch

Yope raises $12.3M to build a private social network without algorithms or ads

Yope has raised $12.3 million to create a private, ad-free social network focused on small communities rather than algorithmic feeds.

Tech Business/VC TechCrunch

Menlo Ventures’ Matt Murphy explains why Anthropic is winning (and it’s not the model)

Menlo Ventures' Matt Murphy discusses the rapid revenue growth of Anthropic, attributing its success to factors beyond just the underlying model.

AI/ML arXiv cs.AI

Public perceptions of AI-driven decision-making in healthcare: A structural equation modeling approach

A study using structural equation modeling to analyze public perceptions of AI-driven decision-making in healthcare, emphasizing trust in clinicians over technology.

AI/ML arXiv cs.AI

Circuit Claims Depend on What Is Extracted and How It Is Compared

Research challenging the determinism of circuit extraction in AI, arguing that reported circuits depend heavily on extraction methods and comparison criteria.

AI/ML arXiv cs.AI

Functional Equivalence and Geometric Diversity in Neural Network Approximations: An Empirical Characterization

An empirical study on the functional equivalence and geometric diversity of neural network approximations, proposing a model selection criterion based on parsimony.

AI/ML arXiv cs.AI

Dual Adversarial Fine-tuning for Enhancing Robustness of Large Vision Language Model

Introduction of a dual adversarial fine-tuning framework to enhance the robustness of Large Vision-Language Models against adversarial attacks.

AI/ML arXiv cs.AI

SFGA: A Statistics-First Gating Architecture with Adjudicative Escalation for Trustworthy SFT Data Procurement

Presentation of SFGA, a statistics-first gating architecture designed for cost-aware and trustworthy procurement of supervised fine-tuning (SFT) data.

AI/ML arXiv cs.AI

Variational meta-learning inference for low dimensional neural system identification

A probabilistic extension of the manifold meta-learning framework using Variational Inference for better system identification in low-data regimes.

AI/ML Hacker News

Terrence Tao's ChatGPT Conversation about the Jacobian Conjecture Counterexample

A discussion featuring Terrence Tao using ChatGPT to explore a potential counterexample to the Jacobian Conjecture.

AI/ML Hacker News

GigaToken: ~1000x faster Language model tokenization

Introduction of GigaToken, a tokenization method for language models that claims to be approximately 1000x faster.

Hardware/Chips Hacker News

Nobody knows what a used GPU cluster is worth

An analysis of the volatility and uncertainty regarding the valuation of used GPU clusters in the current AI hardware market.

Hardware/Chips Hacker News

When Is NVLink Worth It?

A technical exploration of when the performance gains of NVLink justify its cost and implementation in GPU clusters.

Tech Business/VC TechCrunch

Monday.com lays off hundreds to focuses on AI

Monday.com is laying off approximately 630 employees (20% of staff) to refocus resources toward its AI Work Platform.

AI/ML arXiv cs.AI

ABOPD: Antibody CDR Design via On-Policy Distillation

ABOPD is a new antibody design framework using on-policy distillation to improve structural recovery in protein generative models.

Cybersecurity arXiv cs.AI

Data Leakage Prevention in Agentic Applications via Preemptive Hardening

A new pre-deployment pipeline for scanning and hardening agentic LLM applications to prevent data leakage and prompt injection.