Other Hacker News

Zoom is now Chrome only

Zoom has reportedly shifted its compatibility to be Chrome-only, potentially limiting user access on other browsers.

AI/ML arXiv cs.AI

Just A Rather Very Intelligent Spoken Agent

Researchers introduce JarvisBench and a modular prototype for always-on, spoken AI mediators that facilitate better long-horizon agent-user interaction.

AI/ML arXiv cs.AI

A Research Prototype for Closed-Loop Generative Design of Customized Foot Orthoses via Semantic-Physics Alignment

TANS-FO is a research prototype for automating the design of customized foot orthoses using a Text-Aligned Neural Surrogate and GNNs for real-time stress prediction.

AI/ML arXiv cs.AI

TopoTuner: Topological Finetuning of Large Language Models

TopoTuner is a new framework for selective fine-tuning of LLMs using topological persistence diagrams to identify which weights to freeze, reducing parameters and training time.

AI/ML arXiv cs.AI

Diversity-Oriented Fine-Tuning for Uncertainty-Based Hallucination Detection

This paper proposes diversity-oriented fine-tuning using SFT and DPO to make LLM hallucinations more detectable via semantic entropy.

AI/ML arXiv cs.AI

DS@GT ARC at eRisk 2026: Hybrid Multi-Agent LLM System with Structured Algorithmic Guidance for Conversational Depression Screening

A hybrid multi-agent system combining Gemma 27B with algorithmic guidance for depression screening outperformed a paid GPT-5-nano baseline at lower cost.

AI/ML arXiv cs.AI

Tractable Query Answering under Epistemic Confidentiality Policies in DL Ontologies (extended version)

The paper presents a new semantics based on minimal policy violation (MPV) for tractable query answering in DL ontologies under confidentiality policies.

AI/ML arXiv cs.AI

RECON: Benchmarking Agent Memory for Compositional Reasoning over Long Contexts

RECON is a new benchmark evaluating compositional reasoning over long contexts (50k-100k tokens) to test agent memory and reliability.

AI/ML arXiv cs.AI

Constraint-Anchored Reasoning Traces

CART is a neuro-symbolic framework that interleaves natural language reasoning with symbolic constraints to prevent 'error snowballing' in multimodal LLMs.

Software Engineering Hacker News

Reverse-engineering is cheap now

A discussion on the increasing accessibility and reduced cost of reverse-engineering tools and techniques.

Other Hacker News

The Rise of Grocery Tourism

An exploration of the trend where people travel specifically to visit unique or high-quality grocery stores.

Software Engineering Hacker News

GTFO VR Mod Postmortem

A technical postmortem detailing the development and challenges of creating a VR mod for the game GTFO.

AI/ML arXiv cs.AI

When to Plan: Learning to Select Between Reactive Control and Deliberative Planning

Research on meta-reasoning in AI agents to dynamically switch between fast reactive control and slower, more computationally expensive deliberative planning based on uncertainty.

AI/ML arXiv cs.AI

Berkeley and Heiserman as an Unexhausted Architecture for Embodied Machine Intelligence

A theoretical exploration of the works of Berkeley and Heiserman as a foundation for embodied machine intelligence and robotic cognition.

AI/ML arXiv cs.AI

SEER: Supervised Learning to Control Energetic Reasoning

Introduction of SEER, an ML-based oracle function used to optimize Constraint Programming by deciding when to execute complex propagators.

AI/ML arXiv cs.AI

Nonuniformity Principle in Human-AI Coworking

The proposal of the 'nonuniformity principle' to optimize the placement of human oversight stages in AI-driven multi-step workflows.

AI/ML arXiv cs.AI

From Modalities to Propositions: A Language-Centric Framework for Multimodal Intelligence

A language-centric framework that represents multimodal data (image, video, text) as a bag of atomic propositions for better interpretability and reasoning.

AI/ML arXiv cs.AI

Exact Network Surgery: Functional Invariance and Gradient Plasticity in Reactive Computational Graphs

Formalization of Exact Network Surgery for bit-exact insertion of residual blocks into live computational graphs without destroying learned functions.

AI/ML arXiv cs.AI

FST.ai 2.5: Explainable and Uncertainty-Aware AI for Olympic and Para-Taekwondo Decision Support, Athlete Digital Twins, and Federation-Scale Analytics

Presentation of FST.ai 2.5, an explainable AI framework for digital twins and decision support in Olympic and Para-Taekwondo.

Other Hacker News

Do we just want slaves?

A Hacker News discussion thread regarding the ethical and societal implications of AI and the concept of digital servitude.