Skip to main content
Search

# Autoregressive Rollout

5 related articles

Subscribe to updates
  1. arXiv

    Granular flow simulation moves most message passing onto coarse grids: HCG-GNN trains 6.47 times faster and rolls out 10,000 grains in 31 seconds

    Particle-resolved granular flow simulation has been limited by repeated information passing over the full particle graph. HCG-GNN moves 5 of its 7 message-passing steps onto two levels of coarse graphs built from occupied cells, matching the multi-step accuracy of a conventional particle-level GNN in inclined-plane granular flow benchmarks while training 6.47 times faster with 46.2% less GPU memory; a model trained on up to 5000 grains predicted a 2 s release of 10000 grains in 31 s, versus 2000 s for DEM under the same benchmark conditions.

  2. arXiv

    SUMNO unrolls a multigrid V-cycle into a feed-forward network: 9.4M parameters, best in five of seven flow settings

    Fluid surrogate models have long reused U-shaped multiscale designs from vision, but their resampling, skip fusion and normalization were not built for flow fields. SUMNO spatially unrolls a multigrid V-cycle into a feed-forward U-shaped network, using parameter-free downsampling and bilinear upsampling for restriction and prolongation, and shifted-window attention at the coarsest level, with window masks set per-axis periodicity, for global coupling. Across seven settings on four flow benchmarks it is best in five and second in two, cutting 10-step rollout MSE by about 30% on Pinball and RBC, with 9.4M parameters.

  3. arXiv

    Diffusion language models can learn from autoregressive teachers: JustOPD lifts Dream-7B's math reasoning while keeping any-order decoding

    Autoregressive models generate strictly left to right, while diffusion language models denoise in arbitrary order, so the former's token-level supervision does not transfer directly. JustOPD forces the diffusion student to roll out left to right during training, letting a Qwen3-14B teacher supervise every intermediate state, while inference keeps any-order decoding. With Dream-7B as student and Qwen3-14B as teacher on GSM8K and MATH500, scores rise over the base model across several divergences.

  4. arXiv

    EVO-WAM lets world action models self-train on their own generated video, lifting RoboTwin unseen-task success from 26.9% to 68.0%

    The work proposes EVO-WAM, a framework that enables complete autoregressive rollouts without external execution feedback via state prediction and anchored multi-frame context, then selects task-completing prefixes with a vision-language model and verifies video-action consistency with an inverse dynamics model for iterative self-training; on seven unseen RoboTwin 2.0 tasks it raises average success from 26.9% to 68.0% for Cosmos3 and from 28.5% to 46.4% for DreamZero, and on three real-world long-horizon composite tasks it raises Cosmos3 from 20.0% to 76.7%.

  5. ArchesClimate-SSP, trained with an energy-score loss, reaches 0.9739 K land-temperature RMSE on unseen SSP5-3.4, beating MESMER-M's 1.0963 K

    The study introduces ArchesClimate-SSP (AC-SSP), built on the ArchesWeather architecture and trained with an energy-score loss plus a variogram loss while conditioning on forcing concentrations (CO2, CH4, N2O, six aerosol species, ozone), and trained on IPSL-CM6A-LR monthly data for 2015-2100 to autoregressively generate SSP scenarios unseen in training; on the held-out overshoot scenario SSP5-3.4 its 2090-2100 land surface temperature RMSE is 0.9739 K, lower than the statistical emulator MESMER-M's 1.0963 K, and its interannual variability of 0.6050 is closer to the IPSL target's 0.5166 than MESMER-M's 0.6375.

You’re all caught up