Tech Business/VC Hacker News

Elastic lays off 7% of employees

Elastic has laid off 7% of its workforce.

Software Engineering Hacker News

Robotics Teams Are Rebuilding the Data Stack from Scratch

Robotics teams are re-evaluating and rebuilding their data stacks to better handle the unique requirements of robotic data.

Tech Business/VC TechCrunch

Cerebras stock plunges after earnings as CEO says margin outlook was misunderstood

Cerebras stock declined following an earnings report where the CEO claimed gross margin outlooks were misunderstood by investors.

AI/ML arXiv cs.AI

Deep Learning Approaches for 3D Medical Scene Completion: From Geometric Modeling to Generative Paradigms

A systematic review of 3D medical scene completion from 2016 to 2026, covering the evolution from voxel semantic completion to generative diffusion and Gaussian splatting.

AI/ML arXiv cs.AI

Co-occurring associated retained concepts in Diffusion Unlearning

Introduction of ReCARE, a framework to prevent the accidental erasure of benign co-occurring concepts when unlearning harmful content in diffusion models.

AI/ML arXiv cs.AI

MMed-Bench-IR: A Heterogeneous Benchmark for Multilingual Medical Information Retrieval

MMed-Bench-IR is a new heterogeneous benchmark for evaluating multilingual medical information retrieval, highlighting significant performance drops in non-English languages.

AI/ML arXiv cs.AI

Inclusive Interactive Collisions for Multi-View Consistent Compositional 3D Generation

I2C-3D is a novel optimization-based method for generating multi-view consistent compositional 3D assets with physically plausible interactions.

AI/ML arXiv cs.AI

AutoSpec: Safety Rule Evolution for LLM Agents via Inductive Logic Programming

AutoSpec uses inductive logic programming and counterexample-guided inductive synthesis to automatically evolve interpretable safety rules for LLM agents.

AI/ML arXiv cs.AI

Social Structure Matters in 3D Human-Human Interaction Generation

The Solo-to-Social framework uses an LLM planner and a motion executor to generate 3D human-human interaction motions based on social structure.

AI/ML arXiv cs.AI

SURGELLM: Rethinking Multi-Task Evaluation through Task-Aware Feature Gating with Class-Balanced Normalization

SURGELLM is a unified transformer framework for multi-task NLP evaluation featuring surgical feature gating and instance-weighted normalization to improve F1 scores.

Tech Business/VC TechCrunch

AI was supposed to kill engineering jobs, but new data suggests they’re the most resilient

New data suggests that software engineers are more resilient to AI-driven layoffs than previously thought, actually representing a larger share of new hires.

AI/ML VentureBeat

Mistral launches OCR 4, turning document extraction into a full enterprise AI play

Mistral AI released OCR 4, a document intelligence model providing structured representations with bounding boxes and classification, focusing on enterprise sovereignty and self-hosting.

AI/ML arXiv cs.AI

A Benchmark for Hallucination Detection in VLMs for Gastrointestinal Endoscopy

Researchers introduced a benchmark for hallucination detection in VLMs for gastrointestinal endoscopy, finding that white-box hidden-state access (ReXTrust) provides the best detection.

AI/ML arXiv cs.AI

DTT-BSR+: A Generative-Regression Cascade for Music Source Restoration

DTT-BSR+ is proposed as a two-stage generative-regression cascade for music source restoration, improving signal reconstruction and semantic consistency.

AI/ML arXiv cs.AI

Metis: Bridging Text and Code Memory for Self-Evolving Agents

Metis is a self-evolving agent system that uses a dual-representation memory (text and code) to improve task accuracy and execution efficiency on the AppWorld benchmark.

AI/ML arXiv cs.AI

Breaking Shortcut Learning for Cross-Trial EEG-Guided Target Speech Extraction via Two-Stage Training

TRUST-TSE is a two-stage framework designed to prevent shortcut learning in EEG-guided target speech extraction, improving cross-trial generalization.

AI/ML arXiv cs.AI

A P\={a}ninian Foundation for Indic Language Processing

This paper proposes a Paninian computational architecture as a unifying framework for Indic language processing to increase data efficiency and transferability.

AI/ML arXiv cs.AI

Lightweight Transformer Models for On-Device Fault Detection: A Benchmark Study on Resource-Constrained Deployment

A benchmark study on lightweight Transformer models (like TinyBERT) for on-device fault detection shows that traditional ML often outperforms them in latency and size.

AI/ML arXiv cs.AI

Agon: An Autonomous Large-Scale Omnidisciplinary Research System Built on Prompt Economy

Agon is an autonomous research orchestrator built on 'Prompt Economy' that automates claim validation and scales research production.

AI/ML arXiv cs.AI

Zero-Shot Test-Time Canonicalization using Out-of-Distribution Scoring

The authors propose a zero-shot test-time canonicalization method using OOD scoring to make vision models robust against affine transformations.