AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Caracal: Causal Architecture via Spectral Mixing

BINGZHENG GAN, Tianyi Zhang, LI YUSU, Jing Huang, Wei Shi, Yangkai Ding, Tao Yu

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: State space and recurrent sequence models · 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 sliding window attention layers to MHF stack (ablation)Caracal Large · Avg accuracyavg accuracy (%): 48.22 -> 49.01Improved
add sliding window attention layers (ablation)Caracal 345M, ctx 4096 · Avg accuracy (Table 3)avg accuracy (%): 44.92 -> 46.62Improved
add pre-convolution (causal conv) in MHF (ablation)Caracal Large w/o SWA · Avg accuracyavg accuracy (%): 47.82 -> 48.22Improved
add explicit positional encodings (ablation)Caracal Large · Avg accuracyavg accuracy (%): 49.01 -> 48.94No clear effect
use two-stage gate (Linear+SiLU+group conv) instead of single linear projection (ablation)Caracal Large · Avg accuracyavg accuracy (%): 48.05 -> 49.01Improved
raise SWA density from 5:1 to 2:1 MHF:SWA (ablation)Caracal Small · Avg accuracyavg accuracy (%): 43.04 -> 43.35Improved
raise SWA density from 2:1 to 1:1 (ablation)Caracal Small · Avg accuracyavg accuracy (%): 43.35 -> 43.26Hurt
replace attention with MHF (Caracal) vs Llama on retrievalLarge models · SWDEaccuracy (%): 16.38 -> 8.82Hurt
replace attention with MHF (Caracal) vs Llama on LAMBADA perplexityLarge models · LAMBADAperplexity: 24.92 -> 29.39Hurt

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