AI Pulse by Inblix
Sunday, September 20, 2026 AutoCurated · Updated 11:15 UTC

AI Research

Cutting-edge AI research papers, breakthroughs, and academic developments.

1153 articles · page 29

Latest

AWS drops an open-source SDK that makes robots learn from Hugging Face in 5 lines of code

Jun 17

GLM-5.2 crushes open-source coding leaders, lands within 4 points of Claude Opus 4.8

Jun 17

Microsoft, Google, and Hugging Face team up to kill hardcoded AI tool lists

Jun 17

Data centers adapt to grid demands with AI-powered flexibility

Jun 16

First Brain Implant Power User and South Korea's AI Obsession

Jun 16

PyTorch's nn.Linear already runs a fused kernel—here's why torch.compile adds nothing for one layer

Jun 11

I turned a coding agent loose on 2 Hugging Face URLs and it built a 3D Paris

Jun 9

Trackio Slashes CI Time 30% by Ditching GitHub Runners for Hugging Face GPUs

Jun 9

Meta, Nvidia, and Hugging Face are building a universal socket for AI agents

Jun 8

NVIDIA's new AI safety model can now enforce custom rules, not just a one-size-fits-all policy

Jun 4

Hugging Face rebuilt its CLI for AI agents, cutting token use by 6×

Jun 4

DPO crushed text degeneration by 59% in OCR models — SFT never could

Jun 3

Reachy Mini can now search the web and check weather with one command

Jun 3

Holo3.1 ships quantized agents that run locally on your laptop

Jun 2

JetBrains drops Mellum2, a 12B open model that runs 2x faster by only waking a fifth of its brain

Jun 1

IBM's agents use 30× fewer tokens than pure LLMs by giving AI a map, not just a brain

Jun 1

PyTorch profiling isn't just for experts: Here's how to read your first trace

May 29

Async RL's dirty secret: 98% of weights don't change between steps, and TRL now exploits that

May 27

Reachy Mini cuts the cloud entirely with a local AI voice pipeline

May 27

AI's Tower of Babel: Why 'Harness' and 'Scaffold' Mean Everything and Nothing

May 25

OlmoEarth v1.1 slashes satellite AI compute by 3x by killing redundant tokens

May 19

Six new open-source AI rerankers just dropped, hitting state-of-the-art accuracy at every size

May 19

PaddleOCR 3.5 now runs on a Transformers backend, cutting a major friction point for RAG devs

May 18

IBM's 97M-param Granite model crushes sub-100M multilingual retrieval at 60.3

May 14