Search: “safety”

From Pairs to Sequences: Track-Aware Policy Gradients for Keypoint Detection

arXiv:2602.20630v4 Announce Type: replace Abstract: Keypoint-based matching is a fundamental component of modern 3D vision systems, such as Structure-from-Motion (SfM) and SLAM. Most existing learning

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

From Weight Perturbation to Feature Attribution for Explaining Fully Connected Neural Networks

DGX agent

arXiv:2605.15328v1 Announce Type: new Abstract: Fully Connected Neural Networks (FCNNs) are often regarded as simple and intuitive architectures, yet they serve as the foundation for more complex mode

safetyarxiv-cs-lg

18 May 2026

GAP: Geometric Anchor Pre-training for Data-Efficient Visuomotor Learning of Manipulation Tasks

arXiv:2605.15836v1 Announce Type: cross Abstract: Learning visuomotor policies from scarce expert demonstrations remains a core challenge in robotic manipulation. A primary hurdle lies in distilling h

Gaussian Relational Graph Transformer

arXiv:2605.15575v1 Announce Type: new Abstract: Relational graph learning models relational databases as graphs and has demonstrated superior performance on a wide range of relational predictive tasks

Generalized Policy Gradient with History-Aware Decision Transformer for Reliable Routing over Graph Signals

arXiv:2508.17218v3 Announce Type: replace-cross Abstract: Reliable path planning in stochastic transportation networks requires decisions that account for uncertain and correlated travel times on irre

Ghosted Layers: Unconstrained Activation Alignment for Recovering Layer-Pruned LLMs

arXiv:2605.15491v1 Announce Type: cross Abstract: Layer pruning removes entire Transformer decoder blocks from large language models, but introduces a mismatch between the hidden state received by the

GOMA: Toward Structure-Driven Multimodal Alignment from a Graph Signal Smoothing Perspective

arXiv:2605.15723v1 Announce Type: cross Abstract: Multimodal alignment is commonly learned from isolated image-text pairs via CLIP-style dual encoders, leaving the relational context among entities la

HoloMotion-1 Technical Report

arXiv:2605.15336v1 Announce Type: cross Abstract: In this report, we present HoloMotion-1, a humanoid motion foundation model for zero-shot whole-body motion tracking. A key innovation of HoloMotion-1

HoMMI: Learning Whole-Body Mobile Manipulation from Human Demonstrations

arXiv:2603.03243v2 Announce Type: replace Abstract: We present Whole-Body Mobile Manipulation Interface (HoMMI), a data collection and policy learning framework that learns whole-body mobile manipulat

safetyarxiv-cs-ro

HyperDiT: Hyper-Connected Transformers for High-Fidelity Pixel-Space Diffusion

arXiv:2605.15741v1 Announce Type: new Abstract: Pixel-space diffusion models bypass the reconstruction bottleneck of Variational Autoencoders (VAEs) but face a fundamental 'granularity dilemma': captu

Imperfect World Models are Exploitable

arXiv:2605.15960v1 Announce Type: new Abstract: We propose a novel definition of model exploitation in reinforcement learning. Informally, a world model is exploitable if it implies that one policy sh

Import AI 457: AI stuxnet; cursed Muon optimizer; and positive alignment

This newsletter explores three AI topics: potential weaponization risks of AI systems (framed as 'AI stuxnet'), issues with the Muon optimizer in machine learning, and progress or approaches toward po

safetyimport-ai

Improve Large Language Model Systems with User Logs

arXiv:2602.06470v2 Announce Type: replace-cross Abstract: Scaling training data and model parameters has long driven progress in large language models (LLMs), but this paradigm is increasingly constra

it’s funny how people here just make stuff up.

it’s funny how people here just make stuff up. @GaryMarcus Take away LLM & every single AI application today goes back to the stone age, driverless cars will immediately break down, all the apps will

safetygary-marcus--x

just as i feared

just as i feared AI can be a powerful tool for opening people up to new perspectives, yet we find that people actually prefer to use “sycophantic” AI systems that reinforce their pre-existing beliefs.

safetygary-marcus--x

Learn Where Outcomes Diverge: Efficient VLA RL via Probabilistic Chunk Masking

arXiv:2605.16154v1 Announce Type: new Abstract: Reinforcement learning (RL) allows vision-language-action (VLA) policies to generalize beyond their training distribution by optimizing directly for tas

Learning Bilevel Policies over Symbolic World Models for Long-Horizon Planning

arXiv:2605.15975v1 Announce Type: new Abstract: We tackle the challenge of building embodied AI agents that can reliably solve long-horizon planning problems. Imitation learning from demonstrations ha

Learning Sim-Grounded Policies for Bimanual Rope Manipulation from Human Teleoperation Data

arXiv:2605.16043v1 Announce Type: cross Abstract: Deformable Linear Objects (DLOs) such as ropes and cables are widely encountered in both household and industrial applications, yet remain challenging

Linked Multi-Model Data on Russian Domestic and Foreign Policy Speeches

arXiv:2605.15886v1 Announce Type: new Abstract: This paper introduces a dataset of interlinked multimodal political communications from the Russian government, addressing persistent deficiencies in th

safetyarxiv-cs-cl

LRCP: Low-Rank Compressibility Guided Visual Token Pruning for Efficient LVLMs

arXiv:2605.15621v1 Announce Type: new Abstract: Large vision-language models (LVLMs) achieve strong multimodal understanding, but their inference cost grows rapidly with the number of visual tokens, e

LUIVITON: Learned Universal Interoperable VIrtual Try-ON

arXiv:2509.05030v2 Announce Type: replace Abstract: To enable large-scale reuse of real-world 3D assets, where garments and characters rarely share skeletons, templates, or dense correspondences, we p

MAgSeg: Segmentation of Agricultural Landscapes in High-Resolution Satellite Imagery using Multimodal Large Language Models

arXiv:2605.16179v1 Announce Type: new Abstract: Agricultural landscape segmentation in the Global South is challenging as it is characterized by fragmented plots, high intra-class variance, and a scar

MaTe: Images Are All You Need for Material Transfer via Diffusion Transformer

arXiv:2605.15660v1 Announce Type: new Abstract: Recent diffusion-based methods for material transfer rely on image fine-tuning or complex architectures with assistive networks, but face challenges inc

Mind Dreamer: Untethering Imagination via Active Latent Intervention on Latent Manifolds

arXiv:2605.16030v1 Announce Type: new Abstract: Model-Based Reinforcement Learning (MBRL) leverages latent imagination for sample efficiency, yet remains constrained by Historical Tethering: imaginati

Missing-Data-Induced Phase Transitions in Spectral PLS for Multimodal Learning

arXiv:2601.21294v2 Announce Type: replace Abstract: Partial Least Squares (PLS) learns shared structure from paired data via the top singular vectors of the empirical cross-covariance (PLS-SVD), but m

Monotone and Separable Set Functions: Characterizations and Neural Models

arXiv:2510.23634v3 Announce Type: replace-cross Abstract: Motivated by applications for set containment problems, we consider the following fundamental problem: can we design set-to-vector functions s

NavRL++: A System-Level Framework for Improving Sim-to-Real Transfer in Reinforcement Learning-Based Robot Navigation

arXiv:2605.15559v1 Announce Type: new Abstract: Recent years have witnessed significant progress in autonomous navigation using reinforcement learning. However, existing approaches largely emphasize r

safetyarxiv-cs-ro

Neutral-Reference Prompting for Vision-Language Models

arXiv:2605.15615v1 Announce Type: new Abstract: Efficient transfer learning of vision-language models (VLMs) commonly suffers from a Base-New Trade-off (BNT): improving performance on unseen (new) cla

Note that as of last summer Waymo said that LLMs/VLMs were experimental, and that their foundation model then could “process only a small am…

Note that as of last summer Waymo said that LLMs/VLMs were experimental, and that their foundation model then could “process only a small amount of image frames, does not incorporate accurate 3D sensi

safetygary-marcus--x

Nudging Beyond the Comfort Zone: Efficient Strategy-Guided Exploration for RLVR

arXiv:2605.15726v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) has emerged as a scalable paradigm for improving the reasoning capabilities of large language mode

Offline Reinforcement Learning with Universal Horizon Models

arXiv:2605.15603v1 Announce Type: cross Abstract: Model-based reinforcement learning (RL) offers a compelling approach to offline RL by enabling value learning on imagined on-policy trajectories. Howe

OmniVL-Guard: Towards Unified Vision-Language Forgery Detection and Grounding via Balanced RL

arXiv:2602.10687v3 Announce Type: replace-cross Abstract: Existing forgery detection methods are often limited to uni-modal or bi-modal settings, failing to handle the interleaved text, images, and vi

On Kernel Eigen-alignments of KRR: Reconstruction and Generalization

arXiv:2605.15240v1 Announce Type: cross Abstract: This paper investigates the critical role of eigenalignments between the kernel matrix and learning targets in achieving robust generalization in lear

OpenFrontier: General Navigation with Visual-Language Grounded Frontiers

arXiv:2603.05377v2 Announce Type: replace-cross Abstract: Open-world navigation requires robots to make decisions in complex everyday environments while adapting to flexible task requirements. Convent

Opponent State Inference Under Partial Observability: An HMM-POMDP Framework for 2026 Formula 1 Energy Strategy

arXiv:2603.01290v3 Announce Type: replace Abstract: The 2026 Formula 1 technical regulations introduce a fundamental change to energy strategy: under a 50/50 internal combustion engine / battery power

Pessimistic Risk-Aware Policy Learning in Contextual Bandits

arXiv:2605.15620v1 Announce Type: cross Abstract: We study risk-aware offline policy learning, aiming to learn a decision rule from logged data that is optimal under general risk criteria. This proble

phi-Balancing for Mixture-of-Experts Training

arXiv:2605.15403v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) models rely on balanced expert utilization to fully realize their scalability. However, existing load-balancing methods are lar

Polynomial Neural Sheaf Diffusion: A Spectral Filtering Approach on Cellular Sheaves

arXiv:2512.00242v3 Announce Type: replace-cross Abstract: Sheaf Neural Networks equip graph structures with a cellular sheaf: a geometric structure which assigns local vector spaces (stalks) and a lin

Preconditioned Regularized Wasserstein Proximal Sampling

arXiv:2509.01685v2 Announce Type: replace-cross Abstract: We consider sampling from a Gibbs distribution by evolving finitely many particles. We propose a preconditioned version of a recently proposed

PSD: Pushing the Pareto Frontier of Diffusion LLMs via Parallel Speculative Decoding

arXiv:2605.15609v1 Announce Type: new Abstract: Diffusion large language models (dLLMs) generate text by iteratively denoising masked token sequences. Although dLLMs can predict all masked positions i

safetyarxiv-cs-cl

RanSOM: Second-Order Momentum with Randomized Scaling for Constrained and Unconstrained Optimization

arXiv:2602.06824v2 Announce Type: replace-cross Abstract: Momentum methods, such as Polyak's Heavy Ball, are the standard for training deep networks but suffer from curvature-induced bias in stochasti

RE-SAC: Disentangling aleatoric and epistemic risks in bus fleet control: A stable and robust ensemble DRL approach

arXiv:2603.18396v3 Announce Type: replace Abstract: Bus holding control is challenging due to stochastic traffic and passenger demand. While deep reinforcement learning (DRL) shows promise, standard a

ReactiveGWM: Steering NPC in Reactive Game World Models

arXiv:2605.15256v1 Announce Type: new Abstract: Current game world models simulate environments from a subjective, player-centric perspective. However, by treating the Non-Player Character (NPC) merel

ReAlign: Generalizable Image Forgery Detection via Reasoning-Aligned Representation

arXiv:2605.16080v1 Announce Type: new Abstract: The rise of AI-generated images (AIGIs) poses growing challenges for digital authenticity, prompting the need for efficient, generalizable image forgery

Reference-Free Reinforcement Learning Fine-Tuning for MT: A Seq2Seq Perspective

arXiv:2605.15976v1 Announce Type: cross Abstract: Production machine translation relies overwhelmingly on encoder-decoder Seq2Seq models, yet reinforcement learning approaches to MT fine-tuning have l

Reference Games as a Testbed for the Alignment of Model Uncertainty and Clarification Requests

arXiv:2601.07820v2 Announce Type: replace Abstract: In human conversation, both interlocutors play an active role in maintaining mutual understanding. When listeners are uncertain about what speakers

safetyarxiv-cs-cl

Res^2CLIP: Few-Shot Generalist Anomaly Detection with Residual-to-Residual Alignment

arXiv:2605.16171v1 Announce Type: new Abstract: Few-shot Generalist Anomaly Detection requires models to generalize to novel categories without retraining, posing significant challenges in real-world

Residual Reinforcement Learning for Robot Teleoperation under Stochastic Delays

arXiv:2605.15480v1 Announce Type: cross Abstract: Stochastic communication delays in teleoperation introduce signal discontinuities that undermine control stability and degrade control performance. Co