Blog
Daily notes on new models, LLM releases, agent frameworks and AI research, written from the sources we follow and delivered as a newsletter every day.
Get the daily issue
Every new post of the day, in one email. Confirmation required.
LLMs1 min read
arXiv:2609.13151v1 Announce Type: new Abstract: Leading multilingual speech recognition models like Whisper transcribe diverse, low-resource languages without language-specific training but are computationally expensive to deploy. Token ...
From arXiv cs.CL
LLMs1 min read
arXiv:2609.13445v1 Announce Type: new Abstract: Speech-to-speech LLMs like Moshi, and its derivative PersonaPlex, can listen and speak concurrently through full-duplex generation. However, they can begin speaking inappropriately during p...
From arXiv cs.CL
How this blog is made
Each feed entry becomes one request to OpenSmartRoute: the router picks a model with a cost-weighted objective, the editorial-writer skill is layered on the prompt, and the outcome trains the learners - the same pipeline available to every workspace.
Open any post to see which target answered, its confidence, the alternatives and what the request cost. Run the same pipeline yourself: register feeds in the operator console, map a small model under Providers, or call POST /api/v1/route with execute: true.
LLMs1 min read
Researchers developed a scalable method using acoustic masking to quantify the contribution of individual consonants to word intelligibility. The study, employing ASR models across English, Spanish, German, and Czech, identified phoneme frequency and functional load as key factors influencing consonant disruption rates.
From arXiv cs.CL
LLMs1 min read
arXiv:2609.10896v1 Announce Type: new Abstract: Speech-based automatic detection of Alzheimer's disease (AD) provides a non-invasive and scalable approach to early cognitive screening. AD affects both lexical-semantic organization and sp...
From arXiv cs.CL
LLMs1 min read
SEA-SpeechBench is a new large-scale benchmark evaluating speech understanding across 11 Southeast Asian languages. Initial evaluations of leading models reveal significant performance gaps, particularly in temporal understanding and low-resource language prompting.
From arXiv cs.CL
LLMs1 min read
This research traces the evolution of algorithmic outputs, identifying three key mutations – speech as data, speech as engagement, and generative text – and their associated legal and social consequences. It provides a framework for understanding these changes and their impact on freedom of expression and informational privacy.
From arXiv cs.CL
LLMs1 min read
StreamAlign enables real-time streaming speech-text alignment through a framework combining RNN-Transducer alignment and word-level ASR guidance. This results in lower latency and improved performance on speech continuation tasks compared to existing methods.
From arXiv cs.CL
LLMs1 min read
BuzzASR is a collection of 102 fine-tuned Whisper models for ASR, achieving significant improvements in CER compared to Whisper-large-v3 across many languages. The project releases models, code, and tokenizer replacement strategies for enhanced compression.
From arXiv cs.CL
LLMs1 min read
X2-NativeCursor is a lightweight observer for incremental-text streaming TTS that tracks progress using native speech tokens, enabling synchronized highlighting and dialogue updates. It achieves significantly lower tracking error compared to waveform-based alignment, improving real-time performance.
From arXiv cs.CL
LLMs1 min read
A new dataset, TamilEOT, and two audio-only detectors are released for semantic end-of-turn detection in Tamil telephone speech. The models achieve accuracy improvements of 13.41% on a held-out test set, with inference times under 150ms.
From arXiv cs.CL
Research1 min read
A training-free method using low-rank projection of decoder activations significantly lowers hallucination rates in Whisper during inference, with minimal impact on speech recognition accuracy.
From arXiv cs.AI
LLMs1 min read
TRILOGUE is a large-scale trilingual benchmark for spoken dialogue fact-checking, including audio, transcripts, and turn-level labels across English, Russian, and Kazakh.
From arXiv cs.CL
LLMs1 min read
arXiv:2609.04404v1 Announce Type: new Abstract: ASR systems sometimes produce fluent text that is unrelated to the speech they receive. We view these hallucinations as one possible consequence of a broader grounding failure, in which the...
From arXiv cs.CL
LLMs1 min read
The Open ASR Leaderboard has added a new model for Swahili, marking the first language from the Global South. This expansion provides a new evaluation point for Automatic Speech Recognition models.
From Hugging Face blog
Research1 min read
Gemini 3.5 Transcribe offers enhanced speech-to-text transcription capabilities, aiming to improve accuracy and intelligence in transcription tasks.
From Google DeepMind blog
LLMs1 min read
Hugging Face introduced the ASRA benchmark to evaluate and compare the optimization of speech recognition models. The benchmark focuses on measuring the impact of various optimization techniques on model performance.
From Hugging Face blog
LLMs1 min read
Hugging Face and NVIDIA Magpie TTS offer open-weights for building low-latency multilingual voice agents. Engineers gain full deployment control and optimized inference performance for real-time voice applications.
From Hugging Face blog
Posts are drafted from public feeds by models OpenSmartRoute routes to - the same router, skill and metering customers use - and always link to the original source. Corrections: support.
