AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Equivalence of Context and Parameter Updates in Modern Transformer Blocks

Adrian Goldwaser, Michael Munn, Javier Gonzalvo, Benoit Dherin

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Transformer expressivity and learnability · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use numerically stable update via RMSNorm inversion instead of naive update in bfloat16 (ablation)Implicit weight updates, Gemma 3 · Weather forecast prompt, bfloat16token prediction agreement: 87.5% -> 100%Improved
use float32 instead of bfloat16 precision (ablation)Implicit weight updates, Gemma 3 · Weather forecast promptlogit difference and token match: float32 maintains perfect token matching; bfloat16 87.5% agreementImproved
update only the final layer instead of several layers (ablation)Implicit weight updates, Gemma 3 · Starting layer depth ablationgeneration fidelityHurt
start update at earlier layers (spread across more layers) (ablation)Implicit weight updates, Gemma 3 · Starting layer depth ablationupdate normImproved
use scaling update variant instead of naive update (ablation)Implicit weight updates, Gemma 3 · Scaling update variant analysisTVD and token match in low precisionImproved
scaling update instead of stable update (ablation)Implicit weight updates, Gemma 3 · Scaling update variant analysisupdate normNo clear effect
apply updates to Falcon (no post-norm) instead of GemmaImplicit weight updates · Falcon-7B, bfloat16token matchingImproved
include an image in the context (Gemma 3 4B)Implicit weight updates · Multimodal contextlogit matchNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy