FlexiSLM is the first spoken language model supporting dynamic and controllable frame rates on speech input and output, outperforming fixed-rate 7B models at high quality and enabling faster inference at lower rates like 6.25 Hz.
Tool reference
arXiv preprint arXiv:1912.06670 , year=
Tool reference. 80% of classified Pith citations use this work as a method, library, or software dependency, not as a substantive claim.
21
Pith papers citing it
Method reference
80% of classified citations
citation-role summary
dataset 3
background 1
method 1
citation-polarity summary
representative citing papers
HEALTHDIAL is a multilingual multi-parallel spoken dialogue dataset containing 1,500 dialogues per language grounded in WHO content, with recorded speech and speaker metadata across four languages.
ToxiAlert-Bench dataset and dual-head neural network detect toxic speech by distinguishing textual versus paralinguistic sources, reporting 21.1% Macro-F1 and 13% accuracy gains over baselines.
UMID infers membership in contrastive pre-training data using only text queries by performing latent inversion and comparing similarity and variability signals to synthetic gibberish references via unsupervised anomaly detection.
EnCodec is an end-to-end trained streaming neural audio codec that uses a single multiscale spectrogram discriminator and a gradient-normalizing loss balancer to achieve higher fidelity than prior methods at the same bitrates for 24 kHz mono and 48 kHz stereo audio.
BlasBench supplies an Irish-aware normalizer and scoring harness that enables reproducible ASR comparisons and exposes a 33-43 point generalization gap for fine-tuned models versus 7-10 points for massively multilingual ones.
Q2D2 uses 2D geometric grid projections to quantize feature pairs in neural audio codecs, yielding implicit codebooks that improve efficiency and utilization over RVQ, VQ, and FSQ while maintaining reconstruction quality.
StableToken introduces a multi-branch architecture with bit-wise voting to create noise-robust semantic speech tokens, achieving lower Unit Edit Distance and better SpeechLLM robustness than prior single-path tokenizers.
An empirical audit of one web-scraped ML training dataset reveals persistent PII after sanitization, which the authors combine with legal analysis to highlight privacy risks and advocate redefining 'publicly available' data for AI training.
SwitchCodec introduces Residual Experts Vector Quantization and a multi-tiered STFT discriminator to achieve PESQ 2.87 and ViSQOL 4.27 at 2.67 kbps while halving training time via post-training.
CosyVoice 3 achieves better content consistency, speaker similarity, and prosody naturalness in zero-shot multilingual speech synthesis by scaling data to one million hours, model size to 1.5 billion parameters, and introducing a supervised multi-task speech tokenizer plus a differentiable reward模型.
Seed-TTS models produce speech matching human naturalness and speaker similarity, with added controllability via self-distillation and reinforcement learning.
HARNESS introduces Arabic-centric speech foundation models that achieve high efficiency and performance through iterative self-distillation and PCA-based signal compression.
Incidental multilingualism from uneven web training makes LLMs unequal, brittle, and opaque across languages.
Combining LLM-based elderly-contextual paraphrasing with TTS synthesis using elderly speakers reduces word error rates in elderly ASR by up to 58% over standard Whisper baselines on English and Korean datasets.
The IQRA 2026 challenge on Arabic mispronunciation detection reports a 0.28 F1-score gain from new authentic human error data and diverse modeling approaches including self-supervised and audio-language models.
Kimi-Audio is an open-source audio foundation model that achieves state-of-the-art results on speech recognition, audio understanding, question answering, and conversation after pre-training on more than 13 million hours of speech, sound, and music data.
F5-TTS generates natural speech from text via flow matching on DiT with simple text padding, ConvNeXt refinement, and sway sampling, trained on 100K hours multilingual data.
Lightweight training strategies allow speech-aware LLMs to output accurate word timestamps alongside ASR transcripts while also improving recognition quality across datasets.
A survey that classifies non-intrusive ASR refinement methods into five categories, reviews domain adaptation and evaluation datasets, proposes standardized metrics, and identifies future research directions.
CNNs using MFCC features achieve 91.79% accuracy for keyword spotting in Hindi speech on a 40,000-sample dataset.
citing papers explorer
-
IQRA 2026: Interspeech Challenge on Automatic Pronunciation Assessment for Modern Standard Arabic (MSA)
The IQRA 2026 challenge on Arabic mispronunciation detection reports a 0.28 F1-score gain from new authentic human error data and diverse modeling approaches including self-supervised and audio-language models.