Tag
This paper proposes a non-autoregressive CTC-based approach for speech-to-text diacritic restoration in Arabic, incorporating hard constraints during decoding to improve efficiency and reduce error rates.
FreyaTTS is a compact, tokenizer-free Turkish-first text-to-speech model based on a non-autoregressive conditional flow-matching Diffusion Transformer, achieving state-of-the-art performance with a fraction of the parameters of larger systems and released under Apache-2.0.
iLLaDA is an 8B parameter masked diffusion language model with fully bidirectional attention, trained from scratch on 12T tokens. It shows broad improvements over LLaDA and remains competitive with Qwen2.5 7B on several benchmarks. The model and code are open-sourced.
A systematic experimental analysis evaluating eight state-of-the-art Diffusion Language Models across multiple benchmarks, analyzing trade-offs between generation quality and computational efficiency.
Proposes a non-autoregressive scoring method for punctuation restoration in streaming ASR that preserves the input transcript and outperforms prompt-based and fine-tuned baselines under a limited lookahead budget.
Researchers propose a training-free method called Suffix-Anchored Confidence Modulation to improve confidence-based decoding in diffusion language models by addressing issues with EOT tokens and premature decoding.
This paper introduces a diffusion language model that treats text as a continuous process over binary bitstreams, using entropy-gated stochastic sampling to close the performance gap with autoregressive models. It achieves state-of-the-art results on LM1B and OWT benchmarks while reducing memory footprint.
Meta's FAIR team released the code for Flowception, a CVPR 2026 paper presenting a non-autoregressive video generation framework that interleaves frame insertion with continuous denoising to reduce error accumulation and computational cost.
Cola DLM is a hierarchical latent diffusion language model that uses text-to-latent mapping and conditional decoding to achieve efficient, non-autoregressive text generation.
CRoCoDiL proposes a continuous and robust conditioned diffusion approach for language that shifts masked diffusion models into a continuous semantic space, achieving superior generation quality and 10x faster sampling speeds compared to discrete methods like LLaDA.