PRA approximates sequential rollout training in parallel for pixel-space AR models via intermediate states and a pixel decoder, achieving FID 2.58 (135M params) and 1.94 (511M params) on ImageNet-1K 256x256, new SOTA among pixel-space AR models.
hub Mixed citations
Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation
Mixed citation behavior. Most common role is background (47%).
83
Pith papers citing it
Background
47% of classified citations
abstract
We introduce LlamaGen, a new family of image generation models that apply original ``next-token prediction'' paradigm of large language models to visual generation domain. It is an affirmative answer to whether vanilla autoregressive models, e.g., Llama, without inductive biases on visual signals can achieve state-of-the-art image generation performance if scaling properly. We reexamine design spaces of image tokenizers, scalability properties of image generation models, and their training data quality. The outcome of this exploration consists of: (1) An image tokenizer with downsample ratio of 16, reconstruction quality of 0.94 rFID and codebook usage of 97% on ImageNet benchmark. (2) A series of class-conditional image generation models ranging from 111M to 3.1B parameters, achieving 2.18 FID on ImageNet 256x256 benchmarks, outperforming the popular diffusion models such as LDM, DiT. (3) A text-conditional image generation model with 775M parameters, from two-stage training on LAION-COCO and high aesthetics quality images, demonstrating competitive performance of visual quality and text alignment. (4) We verify the effectiveness of LLM serving frameworks in optimizing the inference speed of image generation models and achieve 326% - 414% speedup. We release all models and codes to facilitate open-source community of visual generation and multimodal foundation models.
hub tools
citation-role summary
background 10
baseline 6
method 3
citation-polarity summary
claims ledger
- abstract We introduce LlamaGen, a new family of image generation models that apply original ``next-token prediction'' paradigm of large language models to visual generation domain. It is an affirmative answer to whether vanilla autoregressive models, e.g., Llama, without inductive biases on visual signals can achieve state-of-the-art image generation performance if scaling properly. We reexamine design spaces of image tokenizers, scalability properties of image generation models, and their training data quality. The outcome of this exploration consists of: (1) An image tokenizer with downsample ratio o
co-cited works
representative citing papers
LD-Pruning applies latent discrepancy to prune tokens and adaptively skip unconditional branches in VAR models for up to 2.35x faster inference with preserved quality.
Mind-Omni unifies seven brain-vision-language tasks in one discrete-diffusion framework with a brain tokenizer and a new BQA dataset, claiming SOTA multi-task performance competitive with larger single-task models.
RankE co-evolves AR policy and decoder via alternating ranking optimization, improving both FID and CLIP scores on LlamaGen-XL and Janus-Pro where policy-only RL degrades FID.
HeadKV compresses KV cache for autoregressive image generation via head-aware budget allocation, early head-type identification from consistent patterns, and stratified token eviction.
ToBAC is the first backdoor attack on unified autoregressive models, using data or model poisoning to make triggers elicit cross-modal malicious behavior in text and image generation.
Uni-AdGen uses a unified autoregressive framework with foreground perception, instruction tuning, and coarse-to-fine preference modules to generate personalized image-text ads from noisy user behaviors, outperforming baselines on a new PAd1M dataset.
ExtraVAR enables resolution extrapolation in visual autoregressive models by stage-aware RoPE remapping and entropy-driven attention scaling, suppressing repetition and detail loss.
NTM models each generative reverse step as a conditional normalizing flow with a hybrid shallow-deep architecture, enabling exact-likelihood training and strong four-step sampling performance on text-to-image tasks.
Masked Logit Nudging aligns visual autoregressive model logits with source token maps under target prompts inside cross-attention masks, delivering top image editing results on PIE benchmarks and strong reconstructions on COCO and OpenImages while running faster than diffusion approaches.
Delta tokens compress VFM feature differences into single tokens, enabling a lightweight generative world model that predicts diverse futures with far lower compute than existing approaches.
Drift-AR achieves 3.8-5.5x speedup in AR-diffusion image models by using entropy to enable entropy-informed speculative decoding and single-step (1-NFE) anti-symmetric drifting decoding.
SOLACE improves text-to-image generation by using intrinsic self-confidence rewards from noise reconstruction accuracy during reinforcement learning post-training without external supervision.
GAR-Font is a global-aware autoregressive framework for multimodal few-shot font generation that adds global tokenization, a language-style adapter, and post-refinement to improve style coherence over patch-based methods.
VVS accelerates visual AR image generation by partially skipping verifications in speculative decoding, achieving 2.8x fewer target forward passes while preserving competitive quality.
Derives exact guidance transition rates for discrete flow matching models that require only one model evaluation per sampling step and unify prior approximation-based methods.
Phoneme-guided autoregressive framework for talking-head animation that reduces inter-frame flicker via causal keyframe generation and timestamp-aware interpolation, outperforming diffusion baselines on FVD and a new BG-Flicker metric.
PacTure uses view packing and next-scale autoregressive prediction to generate consistent multi-view PBR textures faster than prior sequential or cross-attention methods.
OAR distills specialized generation orders from any-order AR models via self-distillation, improving FID from 2.39 to 2.17 on ImageNet 256x256 while preserving multi-task flexibility.
T2I-FactualBench is a new three-tier benchmark for factuality of knowledge-intensive concepts in T2I models, using multi-round VQA evaluation to show SOTA models need improvement.
Janus decouples visual encoding into task-specific pathways inside a single autoregressive transformer to unify multimodal understanding and generation while outperforming earlier unified models.
VPG is a training-free inference-time guidance technique that improves autoregressive image and video generation by contrasting model outputs under generated versus corrupted prefixes to strengthen next-step support for the prefix.
CVQ replaces patch-wise vector quantization with channel-wise quantization of feature maps, enabling a next-channel autoregressive model that reports 100% codebook utilization and text-to-image scores of DPG 86.7 and GenEval 0.79.
MIND integrates discrete patch tokenization into diffusion score functions via soft top-k aggregation and dual-branch embeddings, achieving FID 1.95 on ImageNet-256 with 715M parameters while outperforming larger baselines.
citing papers explorer
-
ExtraVAR: Stage-Aware RoPE Remapping for Resolution Extrapolation in Visual Autoregressive Models
ExtraVAR enables resolution extrapolation in visual autoregressive models by stage-aware RoPE remapping and entropy-driven attention scaling, suppressing repetition and detail loss.
-
VibeToken: Scaling 1D Image Tokenizers and Autoregressive Models for Dynamic Resolution Generations
VibeToken enables autoregressive image generation at arbitrary resolutions using 64 tokens for 1024x1024 images with 3.94 gFID, constant 179G FLOPs, and better efficiency than diffusion or fixed AR baselines.
-
Generative Refinement Networks for Visual Synthesis
GRN uses hierarchical binary quantization and entropy-guided refinement to set new ImageNet records of 0.56 rFID for reconstruction and 1.81 gFID for class-conditional generation while releasing code and models.
-
Emu3.5: Native Multimodal Models are World Learners
Emu3.5 is a native multimodal world model pre-trained on over 10 trillion vision-language tokens with next-token prediction, post-trained via reinforcement learning, and accelerated by Discrete Diffusion Adaptation for efficient interleaved generation and world exploration.
-
MMaDA: Multimodal Large Diffusion Language Models
MMaDA is a unified multimodal diffusion model using mixed chain-of-thought fine-tuning and a new UniGRPO reinforcement learning algorithm that outperforms specialized models in reasoning, understanding, and text-to-image tasks.
-
Emu3: Next-Token Prediction is All You Need
Emu3 shows that next-token prediction on a unified discrete token space for text, images, and video lets a single transformer outperform task-specific models such as SDXL and LLaVA-1.6 in multimodal generation and perception.