PulseAugur
EN
LIVE 10:35:44

New methods enhance safety and efficiency in autoregressive image generation

By PulseAugur Editorial · [7 sources] ·

Researchers have developed a novel method for enhancing the safety of autoregressive image generation models by iteratively improving their codebooks. This approach leverages the model's own understanding to identify and eliminate harmful image-text mappings, then fine-tunes the codebook using safe examples to maintain generation quality. Separately, another study introduces ScalingAR, a test-time scaling framework designed to improve autoregressive image generation efficiency and robustness by using token entropy as a confidence signal, leading to significant performance gains without requiring early decoding or external reward models. AI

IMPACT These advancements aim to make AI image generation safer and more efficient, potentially leading to wider adoption in various applications.

RANK_REASON The cluster contains two research papers detailing new methods for autoregressive image generation.

Read on Hugging Face Daily Papers →

AI-generated summary · Google Gemini · from 7 sources. How we write summaries →

New methods enhance safety and efficiency in autoregressive image generation

COVERAGE [7]

  1. arXiv cs.AI TIER_1 English(EN) · Jiayi Xu, Di He, Guolin Ke ·

    Parallel Rollout Approximation for Pixel-Space Autoregressive Image Generation

    arXiv:2606.27978v1 Announce Type: cross Abstract: Pixel-space continuous-token autoregressive (AR) generation directly models images as sequences of raw pixel patches, avoiding discrete tokenization or a separately pretrained tokenizer. However, it faces coupled challenges: high-…

  2. arXiv cs.AI TIER_1 English(EN) · Guolin Ke ·

    Parallel Rollout Approximation for Pixel-Space Autoregressive Image Generation

    Pixel-space continuous-token autoregressive (AR) generation directly models images as sequences of raw pixel patches, avoiding discrete tokenization or a separately pretrained tokenizer. However, it faces coupled challenges: high-dimensional patch generation causes large single-s…

  3. arXiv cs.AI TIER_1 English(EN) · Yunqi Xue, Zhijiang Li, Philip Torr, Jindong Gu ·

    Safe Autoregressive Image Generation with Iterative Self-Improving Codebooks

    arXiv:2606.27147v1 Announce Type: cross Abstract: Unlike diffusion-based models that operate in continuous latent spaces, autoregressive unified multimodal models produce images by sequentially predicting discretized visual tokens. These tokens are derived from a codebook that ma…

  4. Hugging Face Daily Papers TIER_1 English(EN) ·

    Parallel Rollout Approximation for Pixel-Space Autoregressive Image Generation

    Parallel Rollout Approximation (PRA) addresses limitations in pixel-space autoregressive image generation by using low-dimensional intermediate states and parallel training to improve quality and efficiency.

  5. Hugging Face Daily Papers TIER_1 English(EN) ·

    Safe Autoregressive Image Generation with Iterative Self-Improving Codebooks

    Unlike diffusion-based models that operate in continuous latent spaces, autoregressive unified multimodal models produce images by sequentially predicting discretized visual tokens. These tokens are derived from a codebook that maps embeddings to quantized visual patterns. The la…

  6. arXiv cs.CV TIER_1 English(EN) · Jindong Gu ·

    Safe Autoregressive Image Generation with Iterative Self-Improving Codebooks

    Unlike diffusion-based models that operate in continuous latent spaces, autoregressive unified multimodal models produce images by sequentially predicting discretized visual tokens. These tokens are derived from a codebook that maps embeddings to quantized visual patterns. The la…

  7. arXiv cs.CV TIER_1 English(EN) · Harold Haodong Chen, Xianfeng Wu, Wen-Jie Shu, Rongjin Guo, Disen Lan, Harry Yang, Ying-Cong Chen ·

    ScalingAR: Scaling Confidence for Autoregressive Image Generation

    arXiv:2509.26376v3 Announce Type: replace Abstract: Test-time strategies have shown remarkable success in improving large language models, but their application to next-token prediction (NTP) autoregressive (AR) image generation remains largely underexplored. Existing test-time s…

RELATED ENTITIES

RELATED TOPICS