Digest

Reset

1167 papers

2026-11-13 · arXiv:2611.16182

Same-State Pairs Make One JEPA Proof of the Platonic Representation Hypothesis Vacuous

Xin Li

The authors show that, under the hard-whitening JEPA population objective with same-state positive pairs, the induced world kernel becomes the identity, so every centered whitened state function is a global minimizer and rotation-uniqueness becomes vacuous. They further show an orbit-level failure: two minimizers can reach the maximum best-rotation residual. One surviving result is a gap-free lower bound on alignment error, but it cannot certify good-view alignment in the degenerate regime.

JEPAAI summary
2026-11-13 · arXiv:2611.16217

WorldDraft: If Code Builds the World, Can Video Models Just Render It?

Zhiyu Hou, Wen-Jie Shu, Harold Haodong Chen et al.

WorldDraft studies whether video models can act as draft-to-video generative renderers when 4D world dynamics and camera paths are provided by executable code. It introduces a benchmark (432 strictly paired cases across 48 4D world families) to jointly measure world preservation and visual realization over varying draft granularity and appearance alignment. The authors propose WorldDraft-Primer to prompt frozen video renderers for better adherence.

Video World ModelsCode availableAI summary
2026-11-13 · arXiv:2611.16221

BWM-Forge: Agentic World Programming for Embodied Intelligence

BWM Team

BWM-Forge is a code-driven framework for controllable robot video generation. It uses a coding agent to build editable Blender scene programs specifying objects, robot motion, and camera states, then renders proxy videos that provide structural and motion references for a video generation model that adds visual detail. A multi-agent harness (preprocessing, building, review, rendering) supports iterative refinement and experience reuse across tasks, robot embodiments, and viewpoints.

Video World ModelsCode availableAI summary
2026-11-13 · arXiv:2611.16375

Recursive State-Augmented Language Models: Persistent Latent-State Adaptation of Pretrained Language Models

Omer Haq

Recursive State-Augmented Language Models (Recursive SALM) maintain a persistent latent state that evolves after each observation and generates adaptations of a frozen pretrained language model via a modular read–transition–write architecture. In deterministic experiments using frozen GPT-2 small, latent-gated LoRA matches a full-context LoRA baseline on a four-turn hypothesis-inference task. Latent-state interventions show predictions depend on the state when earlier evidence is absent from the decoder context.

Latent DynamicsAI summary
2026-11-13 · arXiv:2611.16422

BlinkMimic: Training a Humanoid Motion Tracker in Minutes on a Single GPU

Fuyi Chen

BlinkMimic is a single-GPU training recipe for whole-body humanoid motion tracking, building on SuperMimic and InstantTrack. On simulated Unitree G1, the authors report one policy for all 40 LAFAN1 clips reaches a target of mean normalized return ≥0.82 with zero failed clips over three consecutive checkpoints in 3.4 minutes on an RTX 4090. The paper attributes speedups to Muon updates, a progress-triggered step-size schedule, residual actions, deployment-failure probing, and CUDA graph capture of whole iterations.

DrivingCode availableAI summary
2026-11-13 · arXiv:2611.16424

PhysCapBench: A Cross-Domain Benchmark for Physically Grounded Video Captioning

Haofeng Wang, Pengcheng Wen, Ruiqi Li et al.

PhysCapBench is a cross-domain benchmark for physically grounded video captioning across third-person scenes, egocentric manipulation, and handheld-gripper demonstrations. It provides human-verified annotations for 1,500 video sequences and evaluates captions using seven physical dimensions. The authors report that physical captions reduce reconstruction LPIPS by 8.6–9.5% (third-person) and 26.9–30.9% (egocentric) vs metadata captions, and instruction enrichment yields higher PhysCapBench scores and policy success (e.g., +7.5 percentage points for π0.5 on a RoboTwin 2.0 subtask).

DrivingAI summary
2026-11-13 · arXiv:2611.16534

Context-Predictive Low-Rank Adaptation for Face Presentation Attack Detection

Aminu Yusuf, Shakeel Muhammad Ibrahim, Sulaimon Oyeniyi Adebayo et al.

CP-JEPA adapts a pretrained CLIP (ViT-B/16) visual encoder for face presentation attack detection using low-rank adaptation (LoRA) plus context-predictive learning. A student predicts complete-image teacher features at masked spatial locations (EMA teacher targets) while retaining unmasked classification, masked classification, and prediction consistency. Inference uses only the LoRA-adapted CLIP visual encoder and binary classifier.

Latent DynamicsAI summary
2026-11-13 · arXiv:2611.16579

ActWorld-JEPA: Action-Conditioned JEPA World Models Meet Latent Planning for End-to-End Autonomous Driving

Ziying Song, Shengkai Zhang, Caiyan Jia et al.

ActWorld-JEPA is an action-conditioned JEPA world model for candidate-based end-to-end autonomous driving. It predicts candidate-specific latent futures with Plan-Conditioned JEPA (PC-JEPA) and uses Relational Residual Reranking (R3) to refine base trajectory scores via plan–future agreement/discrepancy. The authors report best PDMS on NAVSIM v1 and strong NAVSIM v2 and nuScenes open-loop planning results.

DrivingAI summary
2026-11-13 · arXiv:2611.16639

Asymmetric Multimodal Learning with Decomposed JEPAs

Yusuf Shihata

The authors show that cross-modal alignment as a sole objective can lose task-critical private and synergistic information under informational asymmetry, leading to semantic collapse on compositional VQA. They prove an “Alignment Redundancy Bound” limiting aligned latents to the redundant information atom. They introduce Decomposed JEPA (D-JEPA) with a shared core plus private fibers, averting collapse and improving downstream accuracy.

JEPAAI summary
2026-11-13 · arXiv:2611.16645

Rethinking Actions: Continuity-Aware Priors for World-Model Planning

Shaobo Yang, Zichen Ai, Ziyang Ma et al.

The authors propose LG-DKGP, a locality-gated deep-kernel Gaussian-process prior for visual world-model planning that learns action continuations from the temporal structure of expert demonstrations. Their planner uses a learned neural mean plus a temporal GP residual to form a continuity-aware prior, initializes CEM search with a training-derived exploration scale, and evaluates progress with a frozen world model.

Latent DynamicsCode availableAI summary
2026-11-13 · arXiv:2611.16647

InertiaMPDD: Multimodal Personalized Depression Detection via Emotional Inertia-Inspired Dynamics Modeling

Changzeng Fu, Songyang Wang, Chang Gao et al.

InertiaMPDD learns a healthy next-state predictor from non-depressed recordings using frozen pretrained multimodal representations. It compares predicted vs observed next latent states to form reference-derived descriptors: reference deviation, residual decay rate (AR(1) on residuals), and a Jacobian spectral descriptor. These proxies support standalone screening and a personalized correction to a multimodal population margin, achieving high fusion F1 on D-Vlog and F1 on LMVD, with limited standalone discrimination on AVEC2014.

Latent DynamicsAI summary
2026-11-13 · arXiv:2611.16728

EgoX: Egocentric Cross-Embodiment Manipulation with Embodiment Dreaming

Yaru Niu, Binghong Chen, Changwei Yao et al.

EgoX is an egocentric cross-embodiment manipulation framework that pretrains a Modularized Cross-Embodiment Transformer (MXT+) with a training-only Embodiment Dreaming objective on mixed demonstrations from multiple humans and robots, then fine-tunes on a target embodiment. The authors report higher average normalized task progress (0.63→0.80) and larger gains under out-of-distribution (0.53→0.77) across 13 task–embodiment pairs.

RoboticsAI summary
2026-11-13 · arXiv:2611.16767

Evaluating Video LLMs' Understanding of Corner Cases in Autonomous Driving

Gabriel Manalu, Minh Son Hoang, Hyeontaek Hwang et al.

The authors report DashBench, a benchmark for evaluating 25 video LLMs on 1,709 ego-view dashcam clips spanning 88 scenario-level corner-case categories, balanced across collision and non-collision hazards. Using multi-prompt, open-ended QA, the authors find a capability gap: even the top model reaches 46.52% accuracy and models often miss dangerous events and hallucinate nonexistent agents/actions. Code/data are shared at cornercase-dashbench.github.io.

DrivingAI summary
2026-11-13 · arXiv:2611.17031

Your Vision Foundation Model Is Secretly an EEG Foundation Model

Guoan Wang, Feng Liu

SeeEEG converts multichannel EEG into images that a frozen vision transformer can read, by assigning each EEG channel to a dedicated row of ViT tokens and learning only the “display” (montage/filter/contrast) parameters plus a lightweight readout head. Using a frozen DINOv3 ViT-B backbone, the authors report higher balanced accuracy than seven EEG foundation models across eight datasets and strong mean rank improvements in Cohen kappa/AUROC.

Latent DynamicsAI summary
2026-11-13 · arXiv:2611.17060

Ropedia Xperience 2: A Gold Standard for Scaling Physical AI Environments and Behaviors

Hao Li, Fangzhou Hong, Zhaoxi Chen et al.

The authors introduce Ropedia Xperience 2, a physical-AI dataset combining handheld LiDAR 3D scans (1,589 released metric-scale scenes) with 4,230 egocentric HOMIE task recordings covering 74 tasks (131 hours). The captured scenes can be converted to physically interactive simulation, human motions can be retargeted to whole-body robot trajectories, and additional robot behaviors can be collected via virtual teleoperation and long-horizon trajectory generation.

AI summary
2026-11-13 · arXiv:2611.17110

SALA: Semantically Aligned Latent Actions Enable Dual-Axis Scaling of Embodied Foundation Models

Jiachun Jin, Zetong Zhou, Haonan Wang et al.

SALA trains a latent action tokenizer with semantic alignment across human and robot embodiments using unpaired atomic-action language descriptions. Language similarities form soft neighborhoods that move human latents toward control-grounded robot latents (robot latents anchored by action reconstruction). The aligned shared latent action space enables scaling: more general human-video pre-training reduces robot action prediction error; more task-specific human-video post-training increases real-robot success rates.

Latent DynamicsAI summary
2026-11-13 · arXiv:2611.17146

HiVE-VQA: Hierarchical Interaction via Volumetric Embedding for Video Quality Assessment

Robbie Hamilton, Fan Zhang, David Bull

HiVE-VQA applies a JEPA-based frozen video encoder (V-JEPA 2.1 ViT-B) to no-reference video quality assessment. It extracts token grids from multiple depths, stacks them into a hierarchical spatio-temporal volume, and trains only a compact 3D convolutional head (5.2M parameters, 5.7% of total). Trained on LSVQ and evaluated without fine-tuning, it outperforms 17 SOTA methods on SROCC and PLCC across intra- and cross-dataset benchmarks.

JEPACode availableAI summary
2026-11-13 · arXiv:2611.17318

Spatial Foundry: A Multimodal Dataset for Spatial Intelligence

Ethan Lin, Yang Song

Spatial Foundry is a dataset of 5,280 furnished residential 3D scene configurations with aligned floor plans, perspective/panoramic RGB, metric depth, point clouds, meshes, instance-level 3D boxes, and camera trajectory–video pairs. Shared scene coordinates and editable source scenes support independent variation of camera motion vs scene edits for spatial perception and shared world modeling.

AI summary
2026-11-11 · arXiv:2611.13030

APS-VLA: Action-Guided and Future-Aware Token Pruning for Efficient Vision-Language-Action Models

Xiaoyang Xing, Chun-Mei Feng, Anthony Cohn et al.

APS-VLA is a framework for vision-language-action (VLA) token pruning that learns token selection from robot demonstrations while keeping a pretrained VLA backbone frozen. It trains a lightweight Decision Influence Estimator (DIE) using action supervision and an auxiliary action-conditioned visual transition prediction. At deployment, DIE performs one hard token selection/packing step before the frozen policy’s native action generation, enabling faster inference with small task-success changes.

DrivingAI summary
2026-11-11 · arXiv:2611.13113

Building World Models from Intrinsic Motivation

Reece Keller, Declan Kutscher, Xaq Pitkow et al.

The authors integrate 24 intrinsic objectives (curiosity, entropy, competency) into a common DreamerV3 model-based RL framework, alongside three representation-learning strategies (observation reconstruction, next-embedding prediction, current-embedding redundancy reduction). They evaluate zero-shot, few-shot, and multi-task control across locomotion, manipulation, and open-ended gaming. Curiosity with observation reconstruction gives the strongest overall family profile; episodic+lifelong novelty yields the best compositional vision transfer and adaptation.

Latent DynamicsAI summary