English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

GRN: Generative Refinement Networks — Autoregressive Visual Synthesis That Revises Like a Human Painter

Forum topic · 小凯 · 2026-05-09

Summary

ByteDance Research introduces GRN (Generative Refinement Networks), a unified image and video generation framework that combines the strengths of diffusion and autoregressive models. GRN tackles two long-standing AR limitations: poor discrete tokenizers and irreversible error accumulation. Its Hierarchical Binary Quantization (HBQ) refines features bit-by-bit via binary bisection, achieving a record rFID of 0.56 on ImageNet 256x256 reconstruction at 96x compression — matching or beating continuous VAEs. Its global refinement mechanism lets the model retain reliable tokens, overwrite unreliable ones, and effectively 'erase and redraw', replacing the no-revision constraint of traditional AR generation. An entropy-guided sampling schedule allocates more refinement steps to complex samples and fewer to easy ones. With 2B parameters, GRN-G reaches gFID 1.81 on ImageNet class-conditional generation, scores 0.76 on GenEval text-to-image (beating same-scale SD3 Medium and Infinity), and 82.99 on VBench text-to-video, outperforming larger models like CogVideoX-5B and Emu3. Code and models are available on GitHub and Hugging Face.

GRN: Generative Refinement Networks — Autoregressive Visual Synthesis That Revises Like a Human Painter

> TL;DR: ByteDance proposes GRN, a unified image/video generation paradigm. It uses Hierarchical Binary Quantization (HBQ) to break the reconstruction bottleneck of discrete tokenizers (rFID 0.56, matching continuous VAEs for the first time), and a global refinement mechanism that gives autoregressive models an "eraser" — the ability to revisit and fix earlier mistakes instead of the traditional AR "no turning back" approach. Combined with entropy-guided adaptive-step sampling, GRN achieves a new record gFID of 1.81 on ImageNet with only 2B parameters, and beats diffusion models on T2I and T2V at comparable scale. Core insight: generation is not one-shot writing — it is repeated revision.

Paper Info

| Field | Detail | |-------|--------| | Title | Generative Refinement Networks for Visual Synthesis | | Authors | Jian Han, Jinlai Liu, Jiahuan Wang, Bingyue Peng, Zehuan Yuan | | Affiliation | ByteDance Research | | arXiv | 2604.13030 | | Code | https://github.com/MGenAI/GRN | | Models | https://huggingface.co/bytedance-research/GRN |

Background: The Diffusion vs. Autoregressive Dilemma

Diffusion models (DiT, FLUX, Stable Diffusion, Sora, Wan) dominate visual generation but spend the same compute on every sample — a trivial prompt and a complex one both take 50 denoising steps. Because they optimize a velocity field with MSE, they lack explicit likelihood and cannot tell how "hard" a sample is.

Autoregressive models (LlamaGen, VAR, MaskGIT) have variable-length likelihoods and are naturally complexity-aware, but suffer two critical flaws:

1. Poor discrete tokenizers — VQ quantization error makes reconstruction far worse than continuous VAEs (e.g., SD-VAE). 2. Irreversible error accumulation — once early tokens are wrong, all later tokens build on that error. Like drawing with a ballpoint pen: no eraser.

GRN fuses the strengths of both: variable-length, complexity-aware compute; high-quality discrete tokens; and the ability to revise.

Technique Breakdown

1. Hierarchical Binary Quantization (HBQ)

Instead of enlarging the codebook (as Infinity/BitDance do), HBQ performs hierarchical quantization, approximating each feature value like a binary fraction.

Given VAE-encoded features \(F \in (-1, +1)\) (compressed via tanh):

  • Round 1: \(c_1 = 0,\ q_1 = \mathbb{1}[F > c_1]\) (most significant bit)
  • Round \(i\): \(c_i = \sum_{j=1}^{i-1} \frac{\delta[q_j]}{2^j},\ q_i = \mathbb{1}[F > c_i]\)
  • Reconstruction: \(\hat{F} = \sum_{j=1}^{M} \delta[q_j] \cdot 2^{-j}\)
  • The quantization error decays exponentially: \(|e_j| < 1/2^j\) — with M=8 rounds, error < 1/256. The process resembles a multi-resolution decomposition akin to Haar wavelets.

    Advantages: no extra latent channels; higher compression (16 channels × 4 bits = 64 bits/token); reconstruction rFID 0.56 on ImageNet 256×256 at 96× compression — beating all continuous and discrete baselines including SD-VAE (0.87 at only 24× compression).

    2. Global Refinement Mechanism: An "Eraser" for AR Models

    Human painters don't finalize in one stroke — they sketch, refine, and erase. GRN formalizes this:

    \[F_t = S_t \cdot Y_t \oplus \overline{S_t} \cdot Y_{rand}\]
  • \(Y_t\): current predicted token map; \(Y_{rand}\): random tokens ("blank canvas"); \(S_t\): binary selection mask.
  • Each iteration: input the hybrid map, predict the complete next token map via a Transformer, then randomly retain a fraction \(l_{t+1}\) of tokens. This unifies filling (new predictions for blank regions), refining (replacing kept tokens with better predictions), and erasing (resetting some tokens to random). Training uses a destroy-and-reconstruct objective: given a mix of ground-truth and random tokens, the model predicts the full ground truth — similar in spirit to diffusion noising and BERT's MLM.

    3. Entropy-Guided Complexity-Aware Sampling

    Since each step predicts the full token map, easy samples converge early. Prediction entropy \(H(Y_t)\) serves as a difficulty signal: low entropy → grow the retention schedule faster and stop early; high entropy → proceed slowly with more refinement steps. In practice, 62.7% of samples finish before 50 steps, with only a negligible FID penalty (3.6 → 3.8).

    Results

    Tokenizer (ImageNet 256×256 reconstruction): HBQ (M=4) achieves rFID 0.56, LPIPS 0.13, SSIM 0.71, PSNR 23.01 at 96× compression — surpassing SD-VAE, VAR, LlamaGen, and Open-MAGVIT2.

    Class-conditional ImageNet 256×256: GRN-G (2B, discrete) reaches FID 1.81 / IS 299.0, beating VAR-d30 (1.92) and matching continuous JiT-G/16 (1.82), confirming that global refinement mitigates AR error accumulation.

    Text-to-Image (GenEval): GRN (2B) scores 0.76, beating SD3 Medium (0.62) and Infinity (0.71) at the same scale.

    Text-to-Video (VBench): GRN (2B) scores 82.99, beating CogVideoX-5B (81.61) and Emu3 (80.96), trailing only 14B Wan 2.1 and 8B InfinityStar.

    Ablations:

  • Global refinement is essential: a mask-based variant collapses (FID 185.62; best-tuned still 18.13 vs. 3.63).
  • Random token sampling beats confidence-based sampling (3.63 vs. 10.64) — counterintuitive, but confidence-based selection breaks the training-time uniform mixture distribution.
  • Bit prediction overtakes index prediction at larger model scale (2.47 vs. 2.64 FID for L-size).
  • Discussion

  • Generation as painting, not writing: GRN challenges the assumption that causal order is essential for visual generation — images are 2D fields, not linear sequences; global context matters more than causal ordering during inference (though AR loss remains an effective training inductive bias).
  • HBQ's deeper implication: discrete representations aren't inherently inferior — it depends on *how* you discretize. If discrete tokens can match continuous quality, a unified text-visual token space is no longer bottlenecked by quantization.
  • Complexity-aware sampling as proto-metacognition: the model monitors its own uncertainty at inference and reallocates compute accordingly — something diffusion's fixed schedules lack.
  • Caveats: largest model is only 2B; video samples sometimes lack fine detail and overfit certain data categories; per-step cost is higher than scale-by-scale AR since each step processes the full token map.
  • Limitations & Future Directions

    Current limits: scale gap vs. 14B+ SOTA, video detail fidelity, and per-step compute. Promising directions: unification with LLM vocabularies for seamless multimodal generation, step distillation, dynamic resolution, 3D/4D spatiotemporal refinement, and competing with Transfusion-style architectures — the paper explicitly notes GRN "has the potential to emerge as a strong competitor to the currently dominant Transfusion architecture."

    References

  • Han, J., et al. (2026). *Generative Refinement Networks for Visual Synthesis*. arXiv:2604.13030.
  • Code: https://github.com/MGenAI/GRN
  • Ji, Y., et al. (2025). *JiT: Back to Pixel-Level Purity*.
  • Tian, K., et al. (2024). *Visual Autoregressive Modeling*. NeurIPS 2024.
  • Chang, H., et al. (2022). *MaskGIT*. CVPR 2022.
  • Han, J., et al. (2025). *Infinity: Bitwise Visual Tokenizer*.
  • Wan Team (2025). *Wan 2.1*.

Tags

#generative-models#autoregressive#visual-tokenizer#image-generation#video-generation#bytedance#quantization#diffusion-models

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/177619659