MatFormer: Nested Transformer for Elastic Inference · HackerPapers discussions

MatFormer: Nested Transformer for Elastic Inference — mentioned by janalsncm

Sun, 28 Jun 2026 02:15:56 GMT

They also shipped Gemma models with their new Matformer architecture which allows for dynamic computation. https://arxiv.org/pdf/2310.07707v2

Paper and discussion history. A mention is not an endorsement.

MatFormer: Nested Transformer for Elastic Inference — mentioned by andy12_

Tue, 20 May 2025 20:00:01 GMT

I think that it's a poorly named reference to this paper [1] that they mention in the blogpost. If I had to give it another more descriptive name, I would probably name it "Per-Layer Embedding Dimensionality" [1] https://arxiv.org/pdf/2310.07707

Paper and discussion history. A mention is not an endorsement.

MatFormer: Nested Transformer for Elastic Inference — mentioned by byefruit

Thu, 25 Jan 2024 18:53:22 GMT

Probably something similar to https://arxiv.org/abs/2310.07707

Paper and discussion history. A mention is not an endorsement.

MatFormer: Nested Transformer for Elastic Inference — HN submission

Tue, 17 Oct 2023 01:44:39 GMT

MatFormer: Nested Transformer for Elastic Inference

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.