Research map / Video / Video generation
Infinite Gaze Generation for Videos with Autoregressive Diffusion
JENNA KANG, Colin Groth, Tong Wu, Finley Torrens, Patsorn Sangkloy, Gordon Wetzstein, Qi Sun
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| autoregressive diffusion with saliency latents vs best baseline (DiffEye) | Infinite gaze model · DIEM | Levenshtein (mean): 3.78 -> 3.40 | Improved |
| autoregressive diffusion with saliency latents vs DiffEye | Infinite gaze model · DHF1K | DTW (mean): 4.24 -> 4.01 | Improved |
| autoregressive diffusion vs TPP-Gaze | Infinite gaze model · DIEM | Discrete Frechet (mean): 3.60 -> 4.14 | Hurt |
| autoregressive diffusion vs TPP-Gaze in user study | Infinite gaze model · 6 videos, 12 participants | closest to ground truth: 85% preference | Improved |
| replace saliency latent conditioning with MAGVIT2 tokens (ablation) | Infinite gaze model · DIEM | DTW (mean): 12.33 -> 54.92 | Hurt |
| use 45-frame prediction rather than 10 (ablation) | Infinite gaze model · DIEM | DTW (mean): 17.31 -> 12.33 | Improved |
| use 45-frame prediction rather than 10 (ablation) | Infinite gaze model · DIEM | Discrete Frechet (mean): 3.67 -> 4.14 | Hurt |
| use 45-frame prediction rather than 90 (ablation) | Infinite gaze model · DIEM | DTW (mean): 14.65 -> 12.33 | Improved |
| use ground-truth 90-frame history instead of static cold start (ablation) | Infinite gaze model · DIEM | Levenshtein / Max temporal correlation: 3.41 vs 3.40; 0.235 vs 0.235 | No clear effect |
Closest papers by meaning
- ST-DiffEye: Diffusion-based Continuous Gaze Generation via Joint Scanpath-Trajectory Modeling (NeurIPS 2026 · similarity 0.93)
- DiffEye: Diffusion-Based Continuous Eye-Tracking Data Generation Conditioned on Natural Images (NeurIPS 2025 · similarity 0.90)
- GazeFlow: From Human Gaze Behavior to Generative Egocentric Gaze Prediction (NeurIPS 2026 · similarity 0.88)
- EyEar: Learning Audio Synchronized Human Gaze Trajectory Based on Physics-Informed Dynamics (AAAI 2025 · similarity 0.88)
- StreamGaze: Gaze-Guided Temporal Reasoning and Proactive Understanding in Streaming Videos (NeurIPS 2026 · similarity 0.87)
- CaRDiff: Video Salient Object Ranking Chain of Thought Reasoning for Saliency Prediction with Diffusion (AAAI 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
