Tag
VSpark 提出通过校准动作噪声 rollout、隐藏特征对齐与拒绝草稿续用三项设计来改进推测解码,在多个机器人操作任务上实现 1.28x–2.10x 的无损加速。
FlexCast 提出一种灵活的上下文-目标学习框架,将时间序列基础模型的可变上下文条件化与自回归分解解耦,通过动态采样可见上下文和掩码目标的掩码预测实现统一的掩码-预测接口,在保持有竞争力精度的同时显著降低推理延迟,并表明自回归模型的关键收益来自灵活训练而非顺序推理。
ART-MPG 提出了一种紧凑的神经自回归模型,用于流形路径引导(Manifold Path Guiding),通过拓扑条件化的 von Mises–Fisher 混合模型采样反射/透射链,在相同采样数下实现了比现有在线和离线渲染算法更低的均方误差。
The paper introduces Future Chunk Prediction (FCP), an auxiliary objective for autoregressive video diffusion models that predicts temporally separated chunks without intervening visual context, improving physical plausibility and reducing training compute. The approach requires no additional pretrained visual encoder and hints at more structured representation spaces useful for autoregressive world models.
Matrix-Game 4.0 将游戏逻辑与视觉合成分离,通过可执行引擎与自回归视频渲染器实现低延迟的实时可玩游戏世界,在单张 RTX5090 上达到 480p、60 FPS 及 112ms 按键到显示延迟,并开源代码、模型权重与低多边形游戏资产。
The paper introduces LOB-Impact, an open-source benchmark showing that neural generative limit order book models trained on next-token prediction exhibit emergent market impact when metaorders are injected, with impact exponents between 0.63 and 1.21, though relaxation decay remains unobserved likely due to short context windows.
提出多时间尺度分层世界-动作模型(MHWA),通过分层时间抽象缓解自回归世界模型开环预测中的漂移与误差累积问题,在样本高效决策上优于扁平基线。
GeoState提出了一种视角解耦的几何表示,将场景动态与相机运动分离以实现预测性3D世界建模,在Waymo上超越先前方法,并可零样本泛化到KITTI和nuScenes。
论文提出 RepForcing,通过表示空间监督(而非中间教师的轨迹回归)来初始化少步自回归视频生成器,并结合非对称 DMD,在 4 步采样下 VBench 总分从 84.04 提升至 84.66,VisionReward 提升约 59%。
NAMVIS proposes a diffusion-free framework that reformulates sparse-view multi-view image synthesis as geometry-conditioned next-scale autoregression, using Multi-scale Projective Pose Encoding to anchor generation to camera geometry. It outperforms diffusion baselines on Objaverse, GSO, and OmniObject3D in PSNR/SSIM/LPIPS while running over 3x faster.
Introduces Gumbel Straight Flow, a method for distilling autoregressive language models into one-step flow maps that enable parallel sampling.
ProAR transforms autoregressive video generation into a goal-oriented reasoning process by anchoring generation to predicted goal frames via an asymmetric attention mask and adding future representation self-alignment, achieving strong benchmark gains while training on only 25% of the steps of standard AR baselines.
The paper shows that small autoregressive transformers trained only on restricted parameter regimes of nonlinear dynamical systems (logistic/sine maps, Lorenz, Hopf) can extrapolate global dynamics far outside the training distribution, reproducing period-doubling cascades up to period 128 and the Feigenbaum constant to within 5e-4, with causal interventions revealing how control-parameter information flows through attention.
This paper introduces Rollout-Marginal Distillation (RMD) for autoregressive video diffusion, where each generated chunk is scored independently against a chunk teacher to avoid error-compounding from imperfect temporal context, followed by video-level DMD to restore coherence. RMD reportedly maintains high visual quality far beyond its training horizon and outperforms video-level DMD baselines.
This paper introduces ActionPiece, a novel action tokenization method for autoregressive vision-language-action models that uses physical rank consistency to improve the fidelity of action relationships, evaluated on benchmarks like LIBERO and SimplerEnv.
SAPO is a low-memory and compute-efficient framework for agentic reinforcement learning that shares policy and value functions in a single autoregressive backbone, outperforming PPO and GRPO in experiments on ALFWorld and WebShop.
This paper introduces HB-SJD, a batched speculative Jacobi decoding method for visual on-policy distillation that accelerates rollout generation by processing multiple tokens in parallel, reducing training time while preserving generation quality.
This paper investigates whether tabular foundation models (TFMs) like TabPFN, TabICL, TabDPT, and TabFM produce predictions consistent with any joint distribution. It demonstrates that all evaluated TFMs violate both marginalization and factorization consistency for classification and regression, questioning their Bayesian inference claims.
This paper introduces a training-free method to improve revisit consistency in autoregressive generative rendering by using temporal and spatial correspondences from the 3D engine to maintain consistent appearance when the camera revisits locations.
Introduces Surrogate Latent Policy Optimization (SLPO) to apply outcome-reward RL to autoregressive latent reasoners, enabling test-time scaling and variable-horizon policies that improve accuracy on harder instances.