Search: “pretraining”

TacDyn-WAM: Learning Implicit Tactile Dynamics in a Heterogeneous Visuo-Tactile World Action Model

arXiv:2610.00638v1 Announce Type: new Abstract: World action models improve robotic manipulation by conditioning actions on predicted futures, yet existing tactile variants largely inherit video-gener

researcharxiv-cs-ro

2 Oct 2026

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

Towards Fast and Disentangled Counterfactuals for Visual Foundation Models

DGX agent

arXiv:2610.00895v1 Announce Type: cross Abstract: Foundation models remain vulnerable to spurious correlations and ``Clever Hans'' strategies. Explainable machine learning can find and remove such str

model-releasesarxiv-cs-cv

2 Oct 2026

Agentic Relative Camera Pose Estimation via Learned Ranking and Verification

arXiv:2609.38755v1 Announce Type: new Abstract: A wide range of approaches have been developed for camera pose estimation, including correspondence-based methods, end-to-end pose regression, and recen

safetyarxiv-cs-cv

Amortized Data Borrowing with Exchangeability-Aware Neural Posterior Estimation

arXiv:2609.38902v1 Announce Type: new Abstract: Augmenting small concurrent studies with external or historical cohorts is attractive in drug development, where enrollment is slow, follow-up is expens

Beyond Rephrasing: Book-Level Organization Improves Synthetic Textbook Data for Mid-Training

arXiv:2607.28109v2 Announce Type: replace Abstract: Synthetic textbook data has improved language model pre-training, but prior work largely treats the benefit as a property of generated content or lo

researcharxiv-cs-ai

ChronoGraph: Functional 4D Scene Graphs with Vision-Language Models for Interaction Understanding and Grounded Planning

arXiv:2609.39665v1 Announce Type: new Abstract: Embodied agents must determine where to act, anticipate the resulting scene changes, and interpret observed outcomes to guide subsequent actions. This r

tutorialsarxiv-cs-ai

Cue the Flow: Steering Flow-Matching Policies for Open-World Delivery Manipulation

arXiv:2609.38989v1 Announce Type: new Abstract: Open-world goods delivery requires mobile manipulators to follow free-form user instructions and manipulate potentially novel objects. Existing dual-sys

safetyarxiv-cs-ro

Decoupling Multi-Contrast Super-Resolution: Self-Supervised Implicit Re-Representation for Unpaired Cross-Modal Synthesis

arXiv:2505.05855v4 Announce Type: replace Abstract: Multi-contrast super-resolution (MCSR) is crucial for enhancing MRI but current deep learning methods are limited. They typically require large, pai

Efficient Multi-Modal Planning with Reward-Guided Preference Optimization for Autonomous Driving

arXiv:2609.38862v1 Announce Type: cross Abstract: Safe and efficient trajectory planning is essential in autonomous driving. However, existing end-to-end approaches often fall short in both computatio

EVOKE: Eliciting World Knowledge in Agents for Transferable Decision-Making

arXiv:2609.38334v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed as agents for multi-step decision-making, yet transfer poorly to unseen environments. World-model

safetyarxiv-cs-cl

FAST: Flow Any Scene Transformer

arXiv:2609.39748v1 Announce Type: new Abstract: Scaling has become a primary driver of progress in language and vision foundation models, yet its role in precise correspondence matching remains undere

Hardware

Gestalt: a meta-foundation model for astronomy

arXiv:2609.38312v1 Announce Type: cross Abstract: The Platonic Representation Hypothesis predicts that sufficiently scaled foundation models converge on a shared representation of the world. As each n

hardwarearxiv-cs-lg

Applications

Gromov-Wasserstein Distillation for Inductive Multi-View Embedding

arXiv:2609.40047v1 Announce Type: new Abstract: Gromov-Wasserstein multidimensional scaling (GW-MDS) learns low-dimensional representations from relational data but remains transductive, providing no

applicationsarxiv-cs-lg

Grounding Time-Series Foundation Models in Digital Twin Topology for Predictive Maintenance

arXiv:2609.40071v1 Announce Type: cross Abstract: Digital twins increasingly support downstream analytical tasks that depend on time-series data, motivating interest in time-series foundation models (

How to Reduce Localization Ambiguity? Geometry-Semantic Constrained BEV Representation Learning for Satellite-Ground Localization

arXiv:2609.39127v1 Announce Type: new Abstract: Satellite-ground localization estimates the planar position and yaw orientation of a ground camera within a geo-referenced satellite image. Most recent

PLRS-IC: A Dual-Calibration Framework for Chest X-Ray Vision-Language Alignment

arXiv:2609.39266v1 Announce Type: new Abstract: Fine-grained vision-language alignment in chest radiography enables zero-shot classification, grounding, and segmentation without task-specific annotati

StreamRig: Exploiting Intra-Rig Geometry for Streaming Multi-Camera Odometry

arXiv:2609.40244v1 Announce Type: new Abstract: Mobile robots and vehicles carry synchronized multi-camera rigs, yet many streaming 3D foundation models are designed for monocular input, leaving effic

The Row Normalization Puzzle in Muon

arXiv:2609.39114v1 Announce Type: new Abstract: This paper examines how row-wise renormalization affects Muon, focusing on the gap between NorMuon's worst-case guarantees and its practical performance

TSGL: Teacher-Student Graph Learning for 3DGS Compression

arXiv:2609.38635v1 Announce Type: cross Abstract: 3D Gaussian Splatting (3DGS) is a popular representation for novel view synthesis. However, 3DGS contains millions of Gaussian primitives, each with r

Agents

Cooperative Multi-Agent Vision-Language-Action Models via Reinforced Fine Tuning

arXiv:2609.36588v1 Announce Type: cross Abstract: We study reinforcement learning (RL) methods for cooperative multi-agent Vision-Language-Action (VLA) models. This problem is challenging because VLAs

agentsarxiv-cs-ai

Local Ai

Emergent Specialization in Populations of Self-Supervised Collaborative Vision Experts Without a Shared Gate or Cross-Agent Gradients

arXiv:2609.36770v1 Announce Type: new Abstract: Can a population of neural networks develop a useful division of labor without a shared gate or gradients between agents? We study a setting where each

local-aiarxiv-cs-ai

Fine-Tuning on Self-Generated and Reward-Weighted Data: Learning Dynamics, Convergence Rates, and Benefits of Off-Policyness

arXiv:2609.36945v1 Announce Type: new Abstract: We study the learning dynamics of fine-tuning a policy model on self-generated and reward-weighted data, with particular focus on a generalized version

GLASS: Global Latent Aggregation with Slot-based Set Decoding for Scalable All-Atom Crystal Generation

arXiv:2609.37158v1 Announce Type: new Abstract: Generative models for crystals enable the discovery of novel structures, but scaling all-atom generation to larger systems such as metal--organic framew

Making Analog Training Scale: Co-Designing Mapping, Optimizer, and Converters

model-releasesarxiv-cs-lg

arXiv:2609.36584v1 Announce Type: new Abstract: Analog in-memory computing (AIMC) offers an alternative for model training by executing matrix operations directly where weights are stored. However, sc

MatToolBench: Benchmarking Multimodal Agents in Real-World Materials Science Workflows

arXiv:2609.37053v1 Announce Type: new Abstract: Multimodal GUI agents have achieved impressive results on general software benchmarks, yet their ability to operate professional scientific software rem

Message Passing Does More with Less for In-Context Learning on Graphs

arXiv:2609.37057v1 Announce Type: new Abstract: Achieving strong performance with graph neural networks (GNNs) typically requires training and hyperparameter tuning for each dataset, incurring repeate

PHASE: A Physiology-Guided Hierarchical Foundation Model for Intracranial EEG

arXiv:2609.36087v1 Announce Type: cross Abstract: Clinicians and neuroscientists have long analyzed intracranial electroencephalography (iEEG) through directly measurable physiological characteristics

Hardware

Probe-Space Preconditioning for Fast and Stable Zero-Order Training

arXiv:2609.38095v1 Announce Type: new Abstract: Backpropagation (BP) dominates deep learning but imposes a massive memory tax. For example, training OPT-30B with Adam requires approx 600GB of GPU memo

hardwarearxiv-cs-lg

Representational and Functional Robustness to Electrode Montages in EEG Foundation Models

arXiv:2609.36288v1 Announce Type: new Abstract: EEG foundation models (EEG-FMs) are intended to generalize across different datasets by learning representations that, ideally, are invariant to dataset

Scheduling Recursive Reasoning in Looped Transformers

arXiv:2609.36653v1 Announce Type: new Abstract: Recurrent reasoning models have attracted growing attention for scaling test-time computation, typically by iteratively refining latent states with shar

Unlocking the Critic: Reward-Free Policy Optimization for LLM Post-Training

arXiv:2609.37119v1 Announce Type: cross Abstract: Recent approaches to reinforcement learning (RL) post-training for large language models increasingly remove the critic to reduce training instability

safetyarxiv-cs-ai

Adjoint Guidance Flow: Amortized Critic Guidance for VLA Policies

arXiv:2609.34944v1 Announce Type: cross Abstract: Flow-based Vision-Language-Action (VLA) policies are typically trained by behavior cloning and thus do not explicitly optimize long-term task return.

safetyarxiv-cs-cv

Copper-Policy: Focus on the Representation for Robust Robot Manipulation

model-releasesarxiv-cs-lg

arXiv:2609.32779v1 Announce Type: cross Abstract: World Action Models (WAMs) acquire behavioral priors by modeling future scene evolution, but predicting detailed futures in pixel or latent space incu

FlexLoop: Depth-Elastic Looped Policies for Adaptive Test-Time Computation in Deep RL

arXiv:2609.34488v1 Announce Type: new Abstract: Looped architectures scale computation by reusing the same parameters across recurrent steps, and recent work shows that they substantially improve deep

From Pixel to Poses: Object-centric Tool Manipulation Learning from Human Demonstrations

arXiv:2609.35375v1 Announce Type: new Abstract: Scaling up robotic manipulation is primarily bottlenecked by the scarcity of real-world robot data. While recent approaches leverage human video demonst

safetyarxiv-cs-ro

Generative Uncertainty as a Self-supervised Signal for Semantic Similarity Learning

arXiv:2609.35341v1 Announce Type: new Abstract: Evaluating semantic similarity between videos is a fundamental challenge in computer vision, essential for tasks ranging from out-of-distribution (OOD)

How Does 'English (US)' Become the Default? Triangulating Structural Bias Towards American English Across the LLM Pipeline

arXiv:2604.04204v2 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly embedded in educational, professional, and public infrastructure, yet widely used platforms expo

Improving Video Sparse Attention with Fine-grained Router and Sparse Rebasing

arXiv:2609.32882v1 Announce Type: new Abstract: We present VSA2, a frontier trainable sparse attention for video DiTs. VSA2 includes a variety of new architectural features and training procedures tha

researcharxiv-cs-cv

Learning Geometrically-Grounded Amodal 3D Representations for View-Generalizable Robotic Manipulation

arXiv:2601.22988v2 Announce Type: replace Abstract: Real-world robotic manipulation requires visuomotor policies capable of robust 3D scene reasoning under varying camera viewpoints. While recent 3D-a

safetyarxiv-cs-ro

LEGATO: Large-scale End-to-end Generalizable Approach to Typeset OMR

arXiv:2506.19065v3 Announce Type: replace Abstract: We propose Legato, a new end-to-end model for optical music recognition (OMR), a task of converting music score images to machine-readable documents

researcharxiv-cs-cv

Livin' on a Prior: Likelihood Score Approximation for Inverse Problems

model-releasesarxiv-cs-lg

arXiv:2609.34446v1 Announce Type: new Abstract: Generative models have found great success as data-driven methods of solving inverse problems. Two popular approaches work either by combining a pretrai

OneCanvas: 3D Scene Understanding via Panoramic Reprojection

arXiv:2606.19253v2 Announce Type: replace Abstract: Existing approaches to 3D scene understanding in Vision-Language Models (VLMs) either rely on complex, model-specific geometry encoders or large tra

OneSign: Unifying Sign Language Understanding Tasks with One Model

arXiv:2609.33090v1 Announce Type: cross Abstract: SLU encompasses a diverse set of tasks, including ISLR, CSLR, and SLT. Although these tasks share basic semantic and linguistic foundations, they are

researcharxiv-cs-cv

PanoFuse: Panorama-Enhanced Vision-Language-Action Learning with Decoupled Semantic-Geometric Routing

arXiv:2609.31717v1 Announce Type: new Abstract: Vision-Language-Action (VLA) policies have shown promising performance in language-conditioned robotic manipulation. However, most existing VLA systems

QiYao-M: Multimodal Time Series Foundation Model with Role-Aware Modeling of Endogenous and Exogenous Modalities

arXiv:2609.34842v1 Announce Type: new Abstract: Existing multimodal time series foundation models (TSFMs) typically model heterogeneous modalities through largely shared mechanisms, overlooking the di

tutorialsarxiv-cs-lg

Query Expansion and Key Specialization in Transformer Attention Geometry

arXiv:2609.34273v1 Announce Type: cross Abstract: The projection of queries and keys are central to the attention mechanism in Transformer architectures. While they are mathematically symmetric, they

REALIS: A Curated Dataset for Studying the Challenges of AI Image Detection

arXiv:2609.32734v1 Announce Type: new Abstract: AI-generated image detectors are often evaluated on benchmarks where real and synthetic images differ in content, quality, or generation artifacts, allo

Local Ai

SpatialSpeak: QA-Native Reconstruction with Local and Global Context for Spatial Chain-of-Thought Reasoning

arXiv:2609.33616v1 Announce Type: new Abstract: Vision-language models (VLMs) can benefit from geometric priors for multi-view spatial reasoning, yet answer-only training does not directly supervise t

local-aiarxiv-cs-cv