AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

HARD-KV: Head-Adaptive Regularization for Decoding-time KV Compression

Yuxuan Yang, Feiyang Ren, Bowen Zeng, Dalin Zhang, Jinpeng Chen, Gang Chen, Huan Li

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: KV cache eviction and compression · 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
add HARD-KV regularized Top-p budgetingR-KV, Qwen3-8B · AIME24 (Top-k budget 4096)accuracy: 43.3% -> 57.9%Improved
add HARD-KVR-KV, Qwen3-8B · AIME25 (Top-k budget 4096)accuracy: 40.0% -> 56.7%Improved
add HARD-KVSnapKV, Qwen3-8B · AIME24 (Top-k budget 4096)accuracy: 57.7% -> 56.7%No clear effect
add HARD-KVSnapKV, Qwen3-8B · AIME25 (Top-k budget 4096)accuracy: 27.9% -> 42.3%Improved
add HARD-KVVanilla, Qwen3-8B · U-Math (Top-k budget 1024)accuracy: 9.8% -> 46.1%Improved
add HARD-KVVanilla, Qwen3-8B · AIME25 (Top-k budget 2048)accuracy: 16.7% -> 12.9%Hurt
add HARD-KV with Top-p 0.90 dynamic budgetSnapKV, Qwen3-8B · AIME24 throughputtokens/s: 301 -> 632Improved
add HARD-KV with Top-p 0.90 dynamic budgetVanilla, Qwen3-8B · AIME24 throughputtokens/s: 279 -> 732Improved
add HARD-KV with Top-p 0.90 dynamic budgetVanilla, Qwen3-8B · AIME25accuracy: 33.3% -> 20.7%Hurt
add CUDA Graph to base system (ablation)inference system · AIME24throughput (tokens/s): 346 -> 477Improved
add Top-k pruning on top of CUDA Graph (ablation)inference system · AIME24throughput (tokens/s): 477 -> 806Improved
increase Top-p budget from p70 to p90 (ablation)SnapKV, Qwen3-8B · AIME25accuracy: 36.6% -> 42.3%Improved

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