AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Speech-Aware Long Context Pruning and Integration for Contextualized Automatic Speech Recognition

Yiming Rong, Yixin Zhang, Ziyi Wang, Deyang Jiang, Yunlong Zhao, Haoran Wu, Shiyu Zhou, Bo Xu

AAAI 2026 · Architectures and efficiency › Attention and transformers · approach: Sparse attention for long-context LLMs · 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 contextual keywords (5 slides) with SAP2-TPIQwen2-Audio · SlideSpeech L95WER (%): 10.76 -> 7.71Improved
use two-stage pruning (TPI) instead of prompt concatenation (ablation)SAP2 · SlideSpeech L95, 1 slideB-WER (%): 7.74 -> 4.71Improved
use two-stage pruning (TPI) instead of joint pruning-integration (ablation)SAP2 · SlideSpeech S95, 5 slidesB-WER (%): 9.62 -> 5.18Improved
use two-stage pruning (TPI) instead of joint pruning-integration (ablation)SAP2 · SlideSpeech L95keyword filtering F1: 20.45 -> 93.85Improved
extend context from 1 slide to 5 slides (ablation)SAP2-PC / SAP2-JPI · SlideSpeechWER: 7.93 -> 7.86No clear effect
extend context from 1 slide to 5 slidesSAP2-TPI · SlideSpeech L95WER (%): 7.92 -> 7.71Improved
add speech-driven attention pooling (ablation)Qwen2-Audio PC · SlideSpeech L95, 1 slideWER (%): 8.22 -> 7.93Improved
add speech-driven attention pooling (ablation)Qwen2-Audio TPI · SlideSpeech L95, 5 slidesWER (%): 7.90 -> 7.71Improved
add speech-driven attention pooling (ablation)Qwen2-Audio TPI Stage I · SlideSpeech L95 5 slidestraining time (s): 51003.8 -> 38267.9Improved
add speech-driven attention pooling (ablation)Qwen2-Audio TPI Stage I · SlideSpeech 15-slide testinference time (s): 1960 -> 1564Improved
increase pooling window size from 2 to 8 (ablation)SAP2 · SlideSpeech L95 (5 slides)WERHurt
replace CTC-assisted LLM ASR with SAP2-TPILibriSpeech · test-other N=100B-WER (%): 8.02 -> 6.55Improved

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