AI/ML Hacker News

Show HN: Frugon – Find which LLM calls a cheaper model could handle (local, MIT)

Frugon is a tool designed to identify which LLM calls can be handled by cheaper models to optimize costs.

AI/ML arXiv cs.AI

UBEP: Re-architecting Expert Parallelism Communication Library for Production Superpods

Introduces UBEP, a communication library that optimizes Mixture-of-Experts (MoE) model communication on high-bandwidth superpods.

AI/ML arXiv cs.AI

Spider 2.0-AIFunc: Extending Real-World Text-to-SQL to AI-Native SQL Workflows

Presents Spider 2.0-AIFunc, a benchmark for evaluating LLMs' ability to generate AI-native SQL workflows on the Snowflake platform.

AI/ML arXiv cs.AI

VendorBench-100: A Unified Cross-Paradigm Benchmark for Deepfake Image Detection

Introduces VendorBench-100, a unified benchmark to evaluate deepfake image detection across commercial APIs, VLMs, and open-source detectors.

AI/ML arXiv cs.AI

Designing Maintainable Hybrid Generative Systems: A Quantum-Inspired Approach to Automated Music Harmony Generation

Proposes a hybrid generative system using a quantum-inspired approach for automated music harmony generation from melodies.

AI/ML arXiv cs.AI

UI2App: Benchmarking Visual Interaction Inference in Executable Web Application Generation

Introduces UI2App, a benchmark for evaluating the ability of LLMs to infer interaction behavior from UI screenshots for web app generation.

AI/ML arXiv cs.AI

Token-Based Dual-view Fusion and Adaptation of Large Vision Models for Breast Cancer Classification

Proposes a token-centric dual-view learning framework to improve breast cancer classification from mammography using vision transformers.

AI/ML arXiv cs.AI

Estimating Uncertainty from Reasoning: A Large-Scale Study of Multi- and Crosslingual MCQA Performance in LLMs

A large-scale study on uncertainty estimation in LLMs across 22 languages, highlighting the role of reasoning language in performance.

AI/ML Hacker News

We made Grok 4.5, GPT-5.5, and Claude build the same apps

A comparison of how different top-tier LLMs (Grok 4.5, GPT-5.5, Claude) perform when tasked with building the same applications.

Cybersecurity Hacker News

Remote Attestation

A discussion on Remote Attestation, a security mechanism used to verify the integrity of a remote system's software and hardware.

Software Engineering Hacker News

Unicode's transliteration rules are Turing-complete

A theoretical exploration showing that Unicode's transliteration rules are computationally powerful enough to be Turing-complete.

AI/ML Hacker News

MIRA: Multiplayer Interactive World Models Trained on Rocket League

MIRA introduces multiplayer interactive world models trained using Rocket League data to simulate and predict game environments.

AI/ML arXiv cs.AI

LLM Agents for Deliberative Collaboration: A Study on Joint Decision Making Under Partial Observability

A study on LLM agents collaborating in joint decision-making tasks under partial observability, highlighting the challenges of deliberation.

AI/ML arXiv cs.AI

LongCrafter: Towards Diverse Long-Context Understanding via Evidence-Graph-Guided Instruction Synthesis

LongCrafter is a framework for synthesizing diverse long-context SFT data using evidence graphs to improve LLM understanding and mitigate the 'lost in the middle' problem.

AI/ML arXiv cs.AI

X-FEMR: A Token-level Explainable Approach for Electronic Health Records Foundation Models using Transformer-based Models

X-FEMR introduces a token-level explainability approach for Electronic Health Records foundation models using a Transformer-based surrogate model.

AI/ML arXiv cs.AI

Improving LLM-Generated Process Model Quality Through Reinforcement Learning: The Role of Reward Function Design

Research on improving the quality of LLM-generated BPMN process models using Reinforcement Learning and optimized reward function design.

AI/ML arXiv cs.AI

TriA Pipeline: A Large-Scale Automatic Audio Annotation Pipeline For Audio Classification In Specific Scenarios

The TriA Pipeline is an automatic audio annotation system designed to create high-quality training data for audio classification in domestic environments.

Other Hacker News

John Deere owners will get the right to repair equipment under FTC settlement

John Deere will provide equipment owners with the right to repair following a settlement with the FTC.

Software Engineering Hacker News

Beyond Git: Real-Time Version Control for Godot – Lilith Duncan – GodotCon 2026 [video]

Lilith Duncan discusses a real-time version control system designed specifically for the Godot game engine.

AI/ML arXiv cs.AI

EcoVision: AI-Powered Drone Imaging for Salt Marsh Vegetation Monitoring and Dominance Mapping

EcoVision is an AI-powered pipeline using transformers and ConvNeXt for high-resolution monitoring of salt marsh vegetation via drones.