The paper introduces SGF+, a method for autoregressive video generation that decouples the roles of denoising and context writing by assigning separate parameters to each, improving visual quality and temporal consistency without additional training data or longer training periods.
The chart will appear after repeat observations. The current metric comes from the source.
15 source votesReal observations only. History before source connection is not reconstructed.
This approach enhances video generation quality and consistency by separating parameter roles, making it suitable for long-duration video creation without extra training.
GRACE is a two-stage framework for efficient video generation that compresses a pretrained video autoencoder while maintaining compatibility with a pretrained DiT model. It uses a frozen base latent and learns a residual latent to preserve information, aligning the compressed latent with the DiT's feature space for optimized generation.
yesterdayDaily PapersCheckerBench is a benchmark for evaluating coding agents' ability to synthesize static-analysis checkers from defect specifications, including tasks derived from 300 CVEs across multiple repositories and language ecosystems. CheckerLab, an evaluation framework, measures diagnostic contrast, patch localization, and tool use, with results showing that current agents struggle to develop reliable checkers, achieving a maximum Pass@1 of 45.33%.
2 days agoDaily PapersThe paper investigates on-policy distillation (OPD) in cross-tokenizer settings, focusing on alignment coverage and supervision reliability. It finds that strict 1:1 token alignment covers most student-generated tokens despite vocabulary mismatches, and that restricting reverse KL to a top-16 subset of shared vocabulary achieves comparable accuracy to full shared-vocabulary OPD, while adding span supervision reduces accuracy.
2 days agoDaily PapersThe paper introduces WING, a framework that transfers interaction knowledge from human egocentric videos to robot policies by separating observer-induced motion from hand-object interactions and using spectral analysis to identify shared temporal structures between human and robot behaviors. It achieves high success rates on multiple benchmarks and demonstrates strong performance in real-world tasks.
6 days agoDaily PapersThe paper introduces NAVA-WAM, a method for pretraining action policies directly from observation-only videos without relying on action-annotated robot data. It uses two stages: first, pretraining on videos with visual transition supervision, and second, post-training with action-labeled demonstrations to improve robot control.
6 days agoDaily PapersPointWAM is a 3D world action model that decomposes the environment and robot hands into separate components and predicts their 3D point trajectories within a shared space-time coordinate frame. It uses large-scale human demonstration videos for pre-training and improves performance on dexterous manipulation tasks compared to previous methods.
6 days ago