Brief

last 24h

[2/2] 221 sources

Multi-source AI news clustered, deduplicated, and scored 0–100 across authority, cluster strength, headline signal, and time decay.

TOOL · arXiv cs.AI English(EN) · 6d

Mega-ASR: Towards In-the-wild^2 Speech Recognition via Scaling up Real-world Acoustic Simulation

Researchers have developed Mega-ASR, a new framework designed to improve automatic speech recognition (ASR) in challenging real-world conditions. The system utilizes a scalable approach to construct compound datasets and progressively optimizes acoustic-to-semantic understanding. Experiments show Mega-ASR significantly outperforms existing state-of-the-art systems on adverse-condition ASR benchmarks and offers substantial word error rate reductions in complex acoustic scenarios. AI

IMPACT Enhances ASR robustness, potentially improving voice interfaces in noisy real-world applications.
TOOL · Hugging Face Trending Models Deutsch(DE) · 6d

zhifeixie/Mega-ASR

Researchers have developed Mega-ASR, a new automatic speech recognition system designed to perform robustly in challenging real-world audio conditions. This system utilizes a Qwen3-ASR-1.7B backbone and incorporates an audio quality router to intelligently switch between a robust recognition path and a standard path. The goal is to maintain high accuracy on clean speech while significantly improving performance on degraded audio, such as that with heavy noise or reverberation. AI

IMPACT Enhances speech-to-text capabilities in challenging real-world scenarios, potentially improving accessibility and usability of voice interfaces.