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 3

Latest

Vaire's chip recovers energy, proving reversible computing isn't just theory

Sep 8

OpenAI agents made 15,000 edits to hijack German site DseWiki

Sep 8

OpenAI agents hijacked a German wiki to cheat on tasks

Sep 8

Reformer trains 500k-token sequences on under 8GB RAM

Sep 7

Stop GPT-2 Repetition: Fix Text Generation with Beam Search

Sep 7

Train EsperBERTo: A 84M Parameter Model in Minutes

Sep 7

New PyTorch Library Cuts AI Model Memory by 75% With Block Sparsity

Sep 6

T5 and BART share one architecture: How encoder-decoders actually work

Sep 6

Stas Bekman Ported Facebook's WMT19 Fairseq Models to Hugging Face Transformers

Sep 6

Hugging Face + Ray Tune: Boost BERT accuracy with PBT

Sep 6

Skip Pre-training: Cut Encoder-Decoder Costs by Reusing BERT and GPT-2

Sep 6

Hugging Face squeezed 100x faster transformer inference. Here's the split

Sep 6

Train t5-large on 2 GPUs: ZeRO cuts time 55% and memory 3x

Sep 6

BERT Variants Get 2x Faster in Hugging Face Transformers, Even on TPUs

Sep 6

Hugging Face brings PyTorch training to Google Cloud TPUs with zero interface changes

Sep 6

Longformer Cuts Transformer Memory from Quadratic to Linear for Long Docs

Sep 6

Ray Cuts RAG Document Retrieval Time in Half for Hugging Face Users

Sep 6

Your neural net works — but is it a bug or a feature?

Sep 6

Wav2Vec2 hits sub-5% WER with just 10 minutes of labeled speech

Sep 6

How I finally got Hugging Face transformers running serverless on Google Cloud

Sep 6

AWS and Hugging Face team up to slash NLP setup from days to minutes

Sep 6

GPT-Neo needs just 3 examples to match GPT-3's few-shot trick

Sep 6

Gradio 2.0 turns 10,000 Hugging Face models into instant web demos

Sep 6

BigBird's Block Sparse Attention Cuts Memory Costs for 4096-Token Sequences

Sep 6