From the source

[ICLR 2022] Part 1: Efficient Transformer

Research paper

Image published with the post by LG AI Research — Image from lgresearch.ai

LG AI Research introduces three papers presented at ICLR 2022 on efficient Transformer implementation: Pyraformer, cosFormer, and PoNet, which aim to reduce the computational complexity of self-attention.

From the source
We introduce some of the papers presented by LG AI Research at the online ICLR 2022 conference held in April.
lgresearch.ai

Other stories

LG AI Research· 2 OctResearch paper

[ICML2026] Application of Mechanistic Interpretability: Understanding Graph Transformers (TokenGT)

Applying mechanistic interpretability to graph transformers for the first time.

LG AI Research· 22 SepResearch paper

[ICML2026] MCFlow, a New Integrated Model for AI-Generated Inorganic Materials

Announced: 18 Jul 2022
Detected: 28 Aug 2026 22:20 UTC (+1501d after)

Visit source ↗lgresearch.ai

LG AI Research· 21 Sep

LG Aimers 9th: Forecasting Pitch Control with AI

Hugging Face· 2 AugResearch paper

Nyströmformer: Approximating self-attention in linear time and memory via the Nyström method

Worldwide Repls, part 1: The Control Plane
Replit separates control and data planes for global infrastructure
Replit· 8 JulInfrastructure
Exploring text to image models
Replicate· 18 Jul