#01Nov 19, 2026
cs.CV
RL Post-Training in Diffusion Transformers Is Activation Steering
Tuna Han Salih Meral, Kaan Oktay, Alec Helbling and 3 more
Reinforcement learning (RL) post-training substantially improves the visual quality, text rendering, and compositional accuracy of diffusion transformers. Because capabilities such as rendering requested text or composing multiple objects are inherently prompt-dependent, a natural hypothesis is that improving them requires learning complex, input-dependent transformations. To isolate how much of these gains an input-independent shift explains, we probe each RL-tuned model with a deliberately simple parameterization: a fixed set of steering vectors added to the block outputs of the frozen base model. On SD3.5-M, these vectors hold 65x fewer parameters than the RL-tuned LoRA (0.29M against 18.8M). We fit them to match the velocity predictions of public RL-tuned checkpoints, with no reward in the objective. Although the vectors are identical for every prompt, token, and sampling seed, they recover 90.0-96.3% of the reward gains of six RL-tuned image models, and 89.8% and 111.9% on two RL-tuned video models. Initialized at zero and trained with online RL, the vectors alone exceed the reward gains of weight-based RL on text rendering (OCR) and human preference (PickScore) on SD3.5-M. A constant shift thus suffices for most of the gain of policy-gradient RL post-training: without new input-dependent transformations, it steers the pretrained generator toward what it can already compute.
#02Nov 19, 2026
cs.CV
One Prior, Two Dynamics: Repurposing Video Diffusion Transformers for Self-Supervised Latent Action Learning
Yuxin Jiang, Wenqiang Sun, Jiaqi Li and 4 more
Videos reveal how the world changes, but rarely the actions that cause those changes. Latent action learning recovers such supervision from visual transitions, yet often learns the underlying dynamics largely from scratch. We ask whether a pretrained video generator can instead provide the prior for both inverse and forward dynamics. We introduce O$^{2}$LaF, built on a simple principle: one prior, two dynamics. We initialize both models from the same pretrained video DiT and adapt them to their distinct roles: a feed-forward transition encoder with learnable action queries for inverse dynamics, and a causal, action-conditioned video generator for forward dynamics. By controlling information access and optimizing both dynamics end-to-end through a compact bottleneck, O$^{2}$LaF learns transferable actions from unlabeled videos. Across navigation and robotic manipulation, O$^{2}$LaF sets a new state of the art in few-shot action probing and enables stronger zero-shot action transfer across visual contexts and embodiments. Beyond generative video priors, we further extend our framework to predictive representation priors such as V-JEPA2, achieving competitive performance. Our findings suggest a broader paradigm for latent-action learning: repurposing pretrained video models to discover transferable actions and predict their consequences.
#03Nov 19, 2026
cs.CV
SGAR: Feed-Forward 3D Reconstruction via Sparse Geometry-Guided Attention
Aoxiang Fan, Maria Pilligua, Corentin Dumery and 2 more
Feed-forward models have been very successful in 3D reconstruction with ad hoc attention designs that relate all view patches densely. However, the cost of this dense global attention grows quadratically with the number of views, preventing the model from applying to larger-scale reconstruction settings. In this paper, we argue that most dense attention computations are unnecessary from a classical structure-from-motion perspective: 3D structures stem from geometrically matching regions in 2D. We bring this principle to learning-based feed-forward reconstruction with our Sparse Geometric Attention Reconstruction (SGAR) method. In SGAR, each patch token attends to only a small set of patch tokens from other views, chosen by a lightweight learned key selection module trained to predict cross-view 3D proximity of tokens. Thus, dense global attention is replaced by much more efficient geometric-similarity-guided sparse attention. SGAR is a general method that can be applied to different feed-forward models by fine-tuning. In our experiments, we show that SGAR is the fastest among both state-of-the-art dense and sparse methods and the most accurate sparse one. It runs up to 12-times faster than dense attention at 1,000 views with stable pose accuracy where existing sparse methods fail, and shows the best performance in reconstructing more complete geometries on the same time budget.
#04Nov 19, 2026
cs.CL
MemFolder: Structure-Guided Memory Folding for Efficient Agent Memory Management
Xuyang Liu, Chenwen Lu, Junxi Wang and 6 more
Long-term agents accumulate large external memory stores as interactions unfold, increasing the costs of storage, retrieval, and response generation. Efficiency is constrained by two coupled factors: a growing record count increases retrieval overhead, while longer retrieved contexts increase prefill computation and decoding-time attention costs. Existing methods prune or consolidate records, but pruning can discard useful evidence, whereas consolidation may produce fewer yet longer retrieval units. We introduce MemFolder, a structure-guided memory folding framework that sequentially reduces record count and context length. The framework comprises two core components that share memory groups, using cross-record structure to guide both evidence consolidation and compact representation. RecordFolder grounds entity mentions, progressively selects representative yet complementary anchors, and combines semantic and entity affinity with temporal chaining to consolidate compatible records into fewer retrieval units. ContextFolder encodes the original records within these groups into compact text and uses verification feedback to revise encodings when key facts cannot be recovered. Experiments on the three text-based memory benchmarks LoCoMo, PersonaMem, and LongMemEval and the two multimodal benchmarks M3-Bench-web and M3-Bench-robot demonstrate the strong performance of our methods. Notably, on Mem0, RecordFolder retains 99.0% of the original average performance across the three text-based benchmarks with 50% fewer records, exceeding the strongest baseline by 1.9 percentage points in performance retention. Combined with ContextFolder, existing record compression methods further shorten their output records while retaining most of their performance.
#05Nov 19, 2026
cs.CV
Plan4D: Generative Plannable 4D Worlds
Hangwei Zhang, Rujing Dang, Ruimin Lin and 3 more
Most world models for future prediction prioritize aestetic quality of generated videos, but offer only weak physical grounding and limited geometric consistency over time, due to the lack of explicit 4D modeling. In this paper, we introduce plan4d, a possible next milestone towards building a generative plannable 4D world model that predicts future RGB videos together with dense 4D geometry, including depth, cameras, point clouds, and scene flow from an observed scene context and text-specified actions. Since training data with paired actions and dense 4D annotations is inherently scarce, plan4d leverages the complementary priors of foundational video generators and 3D reconstructors. Specifically, we propose a model-stitching architecture that couples Depth Anything 3 with LongCat-Video and augments it with an auxiliary DPT head for scene-flow prediction. The resulting model allows us to plan and generate complex, long-horizon, and fine-grained actions in a dynamic 3D scenes, while substantially improving both visual fidelity and geometric accuracy of the generated future videos. Extensive experiments on both synthetic and real-world datasets demonstrate that plan4d achieves state-of-the-art performance in generative 4D modeling of environmental reactions. Project Page: https://fogradio.github.io/plan4d-project