Questions · Research & Expert Networks

What new AI research papers appeared this week?

This week, five new AI research papers were published across arXiv's AI Infrastructure, Inference & Ops, Model Efficiency & Engineering, and RAG, Search & Knowledge Systems categories. On June 8, "FP8 is All You Need (Part 1): Debunking Hardware FP64 as the HPC Holy Grail" appeared in AI Infrastructure, Inference & Ops. The following day, June 9, saw three new papers: "A 65-nm Privacy-Preserving Neuromorphic Encoder With 7.13-nJ Efficiency, 2.38-Mb/mm^2 Item-Memory Density, and Federated Learning Support" and "APEX4: Efficient Pure W4A4 LLM Inference via Intra-SM Compute Rebalancing" were published in AI Infrastructure, Inference & Ops, while "How Small Can You Go? LoRA Fine-Tuning 270M-8B Models for Merchant Information Extraction in Financial Transactions" was released in Model Efficiency & Engineering. Finally, on June 10, "ConvMemory v2: A Recall-Preserving Top-10 Evidence Reranker for Conversational Memory Retrieval" was published in RAG, Search & Knowledge Systems. Synthesized from 1148 manifests produced by 10 monitored preprint sources in the last 7 days, including arXiv.

Answer updated Jun 11, 2026 00:00 UTC · rebuilt twice daily from the rolling 168-hour window

A 65-nm Privacy-Preserving Neuromorphic Encoder With 7.13-nJ Efficiency, 2.38-Mb/mm^2 Item-Memory Density, and Federated Learning Support

arXiv AI Infrastructure, Inference & Ops · 2026-06-09 · 12 claims · manifest 1780984721674944811 source →

APEX4: Efficient Pure W4A4 LLM Inference via Intra-SM Compute Rebalancing

arXiv AI Infrastructure, Inference & Ops · 2026-06-09 · 15 claims · manifest 1780984721689651805 source →

FP8 is All You Need (Part 1): Debunking Hardware FP64 as the HPC Holy Grail

arXiv AI Infrastructure, Inference & Ops · 2026-06-08 · 13 claims · manifest 1780897792400354862 source →

How Small Can You Go? LoRA Fine-Tuning 270M-8B Models for Merchant Information Extraction in Financial Transactions

arXiv Model Efficiency & Engineering · 2026-06-09 · 12 claims · manifest 1780981552911232187 source →

ConvMemory v2: A Recall-Preserving Top-10 Evidence Reranker for Conversational Memory Retrieval

arXiv RAG, Search & Knowledge Systems · 2026-06-10 · 10 claims · manifest 1781080456553565194 source →