AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

CARE: Covariance-Aware and Rank-Enhanced Decomposition for Enabling Multi-Head Latent Attention

Zhongzhu Zhou, Fengxiang Bie, Ziyan Chen, Zhenyu Zhang, Yibo Yang, Junxiong Wang, Ben Athiwaratkun, Xiaoxia Wu, Shuaiwen Leon Song

ICLR 2026 · Architectures and efficiency › Attention and transformers · approach: Sparse and structured attention · 1 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
replace weight-only SVD (Palu) with covariance-aware factorizationMLA conversion · One-shot, KV rank 512, Llama-3.1-8B (Alpaca)avg accuracy (%): 39.33 -> 62.33 (CARE-U)Improved
use covariance-aware factorization instead of Palu(SVD) at rank 128MLA conversion · One-shot, KV rank 128, Llama-3.1-8BWikiText2 perplexity: 3046.58 -> 353.74 (CARE-E)Improved
use adjusted-rank allocation instead of uniform rank (CARE-E vs CARE-U) at rank 128 (ablation)CARE · One-shot, KV rank 128, Llama-3.1-8Bavg accuracy (%): 36.72 -> 38.59Improved
use adjusted-rank allocation instead of uniform rank at rank 512 (ablation)CARE · One-shot, KV rank 512, Llama-3.1-8Bavg accuracy (%): 62.33 -> 57.82Hurt
use adjusted-rank allocation instead of uniform rank at rank 256 (ablation)CARE · One-shot, KV rank 256, Qwen3-4Bavg accuracy (%): 54.33 -> 51.23Hurt
covariance-aware init vs SVD-LLM V2 (activation-aware baseline) at rank 512CARE · One-shot, KV rank 512, Llama-3.1-8Bavg accuracy (%): 62.21 (SVD-LLM V2) vs 62.33 (CARE-U)No clear effect
replace TransMLA initialization with CARE(E) covariance-aware init, then heal 1B tokensTransMLA · Healed Llama-3.1-8B, MLA rank 512avg accuracy (%): 61.36 -> 62.59Improved
replace TransMLA initialization with CARE(E) init, heal 3B tokensTransMLA · Healed Llama-3.1-8B, MLA rank 512avg accuracy (%): 61.86 -> 63.27Improved
add healing fine-tune from 0B to 3B tokens (ablation)TransMLA + CARE(E) Init · Llama-3.1-8B, MLA rank 512avg accuracy (%): 57.04 -> 63.27Improved
more calibration samples beyond 512 or longer sequences (ablation)CARE · Llama-3.1-8B one-shot accuracyaccuracyNo clear effect
vary shrinkage coefficient alpha (ablation)CARE · one-shotaccuracyNo clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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