Medium

Follow this topic Ask about this topic

Momentum

2 papers in the last four weeks, down 33% on the four weeks before. 0.0% of all new papers.

Jul 13Week of Sep 28

Latest papers 18

All topics

Median Temporal Ensembling: Training-Free Robust Aggregation for Action-Chunked Visuomotor Policies
Sep 22, 2026Yuhang JiangVisuomotor Policy Adversarial Corruption
University of Trento, Italy
Brain Metastases Segmentation for BraTS 2026 Task 1: A Multi-Architecture Comparison
Sep 21, 2026Mahdi Islam, Musarrat TabassumBrain Tumor Segmentation Cross Validation
Center for Computational & Data Sciences, Independent University, Bangladesh · Department of Electrical and Electronic Engineering, Independent University, Bangladesh · Independent Researcher
TempCloze: Can Video-LLMs Identify the Missing Middle?
Sep 1, 2026Wenqi Pei, Henry Hengyuan Zhao, Yilai Liu +4Egocentric Video Medium
The University of Hong Kong · National University of Singapore · Peking University
Ripple-Pivot Search: Active Parallel Decoding for Diffusion Large Language Models
Aug 12, 2026Yushi Ye, Xu Chen, Haoyun Jiang +7Diffusion Language Models Large Language Model Decoding
Cooperative Medianet Innovation Center, Shanghai Jiao Tong University · Alibaba Group · TMLR Group, Department of Computer Science, Hong Kong Baptist University +2
ReRound: Reconstructive Rounding to Resolve Midpoint Ambiguity in Calibration-Free LLM Quantization
Aug 11, 2026He-Yen Hsieh, H. T. KungPost-Training Quantization Medium
Harvard University
Probing Character-level Transformers for the Spanish L-shaped Morphome
Aug 4, 2026Akhilesh Kakolu Ramarao, Kevin Tang, Wiebke Petersen +1Morphology Transformer Architectures
Department of English Language and Linguistics, Heinrich Heine University Düsseldorf · Department of Linguistics, College of Liberal Arts and Sciences, University of Florida · Institute of Linguistics and Information Science, Heinrich Heine University Düsseldorf +1
WaterGen: Decoupling Scene and Medium in Underwater Image Generation
Jun 30, 2026Jiayi Wu, Tianfu Wang, Tianyi Xiong +6Indoor Scene Generation Scene Understanding
University of Maryland · University of South Florida · University of Florida
IHBench: Evaluating Post-Interruption Recovery in Voice Agents with Structured Workflows
Jun 17, 2026Ahmad Salimi, Wentao Ma, Yuzhi Tang +3Dialogue Benchmarks Voice Agents
Boson AI Toronto, ON, Canada · Boson AI Santa Clara, CA, USA
On Finite-sample Concentration of Median of Incomplete U-Statistics
May 30, 2026Nong Minh Hieu, Antoine LedentConcentration Bounds Kernel Method
Singapore Management University, School of Computing and Information Systems
Midpoint Generative Models
May 28, 2026Daniil Shlenskii, Nikita Gushchin, Lev Novitskiy +2Generative Models Stochastic Interpolants
1AXXX, Russia · 2Applied AI Institute, Russia · 3Kandinsky Lab, Russia
Positional Failures in Long-Context LLMs: A Blind Spot in Reasoning Benchmarks
May 22, 2026Chuyifei Zhang, Hongyu Cui, Xiaowen Huang +1Reasoning Benchmark Medium
Beijing Jiaotong University · Central South University of Forestry and Technology
Hidden-State Privacy Has an Empty Middle
May 21, 2026Alexander Okezue BellPrivacy Hidden States
Stanford University 450 Jane Stanford Way Stanford, CA 94305
Metis AI: The Overlooked Middle Zone Between AI-Native and World-Movers
May 14, 2026Xiang LiArtificial Intelligence Systems Real-World Tasks
Massachusetts General Hospital
Self-organized MT Direction Maps Emerge from Spatiotemporal Contrastive Optimization
May 12, 2026Zhaotian Gu, Molan Li, Jie Su +3Human Visual Cortex Contrastive Learning
School of System Science, Beijing Normal University, Beijing 100875, China · Qiyuan Laboratory, Beijing 100095, China · State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal2026 University, Beijing 100875, China
Phoenix-VL 1.5 Medium Technical Report
May 11, 2026Team Phoenix, :, Arka Ray +30Large Language Model Adaptation Multilingual Language Models
HTX × Mistral AI ∗
The Truth Lies Somewhere in the Middle (of the Generated Tokens)
May 11, 2026Sophie L. Wang, Phillip Isola, Brian CheungGlobal Average Pooling Semantic Representations
1MIT.
Neutron and X-ray Diffraction Reveal the Limits of Long-Range Machine Learning Potentials for Medium-Range Order in Silica Glass
Apr 23, 2026Sai Harshit Balantrapu, Atul C. Thakur, Chris Benmore +1Interatomic Potentials Powder X-Ray Diffraction
Department of Material Design and Innovation, University at Buffalo, Buffalo, NY 14260, United States of America · NVIDIA Corp., 2788 San Tomas Expressway, Santa Clara, CA 95051, United States of America · X-ray Science Division, Advanced Photon Source, Argonne National Laboratory, Lemont, IL 60439, United States of America
HeadRank: Decoding-Free Passage Reranking via Preference-Aligned Attention Heads
Apr 19, 2026Juyuan Wang, Chenxing Wang, Yuchen Fang +8Cross-Encoder Reranking Relevance
Weixin Group, Tencent, China · 2South China University of Technology, Guangzhou, China

Sep 22, 2026, cs.ROJ/K move · Enter open · S save

Median Temporal Ensembling: Training-Free Robust Aggregation for Action-Chunked Visuomotor Policies

Yuhang Jiang

University of Trento, Italy

Visuomotor Policy Adversarial Corruption Encoders Medium

Open paper AskPDF arXiv

Action-chunked visuomotor policies predict overlapping trajectories, so every executed action is covered by several predictions. Temporal ensembling smooths execution by combining these predictions with an exponentially weighted mean. One corrupted prediction can move the aggregate without bound: its breakdown point is 0. We use adversarial corruption to stress this deployed aggregator and to compare two kinds of guarantee. A metric guarantee bounds the response to a perturbation of a given size. A combinatorial guarantee instead bounds the damage when at most q of the M candidates covering a timestep are corrupted, whatever their size. Encoder adversarial fine-tuning recovers 44% of the loss under the published patch attack, but only 7.3% after the attacker's step size is increased. By contrast, the coordinate-wise median of the same candidate set keeps its recovered fraction flat as attack optimisation increases. Median temporal ensembling costs one line and requires no retraining. Across 25 (configuration, corruption-level) combinations it is never worse than the mean and is significantly better in 15. It also transfers to a second policy class, and it recovers performance under a failure with no attacker in the loop at all: camera frames that arrive blank. Its effect on clean data is configuration-dependent, from -0.04 to +0.07. We also give the boundary: corruption that shifts every covering prediction by the same amount is invisible to this whole family of statistics, and no equivariant aggregator can remove it.

Momentum

Related topics

Latest papers 18