safety

DiPS: Dialogue Policy Selection for High-Stakes Persuasion Agents

arXiv:2607.01557v1 Announce Type: cross Abstract: Large Language Models (LLMs) often struggle with persuasion in high-stakes scenarios. People's individual personalities and concerns require tailored

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Don't Let Gains FADE: Breaking Down Policy Gradient Weights in RL

DGX agent

arXiv:2607.01490v1 Announce Type: cross Abstract: Reinforcement learning post-training dramatically improves LLM reasoning, but suffers from training instability and diversity collapse. Advantage func

safetyarxiv-cs-ai

3 Jul 2026

DriveVLM-RL: Neuroscience-Inspired Reinforcement Learning with Vision-Language Models for Safe and Deployable Autonomous Driving

arXiv:2603.18315v2 Announce Type: replace-cross Abstract: Traditional reinforcement learning (RL) methods rely on manually engineered rewards or sparse collision signals, which fail to capture the ric

DRL-CLBA: A Clean Label Backdoor Attack for Speech Classification via DDPG Reinforcement Learning

arXiv:2607.01729v1 Announce Type: new Abstract: Deep learning models for speech classification are vulnerable to backdoor attacks, where malicious triggers cause misclassification at inference time. W

Efficient Waste Sorting for Circular Economy: A Confidence-guided comparison between One-Vs-All and One-Vs-Rest Classification Strategies with Human-in-the-Loop for Automated Waste Sorting

arXiv:2607.02230v1 Announce Type: cross Abstract: The complexity of waste disposal regulations across European countries poses significant challenges for the residents and hinders the transition to a

ElephantAgent: Contextual State Continuity in Agentic Systems

arXiv:2607.01919v1 Announce Type: new Abstract: Agentic systems enhance their capabilities by invoking external tools and maintaining persistent memory. However, these external dependencies introduce

Episodic-to-Semantic Consolidation Without Identity Drift

arXiv:2607.01988v1 Announce Type: new Abstract: Long-running adaptive intelligent agents face a structural tension between knowledge consolidation and information integrity. Memory consolidation is co

Epistemic Goggles: A Pretrained Module that Induces an Epistemic Frame via Gradient Editing

arXiv:2607.01690v1 Announce Type: new Abstract: Finetuning a language model on documents that are explicitly annotated as fictional results in a model that still actually believes the documents' core

ESC: Emotional Self-Correction for Reliable Vision-Language Models

arXiv:2607.02089v1 Announce Type: cross Abstract: Vision-language models (VLMs) have achieved strong performance across diverse multimodal tasks, yet they remain vulnerable to unreliable reasoning. Ex

Exploring Large Language Models for Access Control Policy Synthesis and Summarization

arXiv:2510.20692v2 Announce Type: replace-cross Abstract: Cloud computing is ubiquitous, with a growing number of services being hosted on the cloud every day. Typical cloud compute systems allow admi

Fast and Accurate Anomaly Detection in Time Series

arXiv:2607.02046v1 Announce Type: new Abstract: Anomaly detection is a critical and evolving field in Machine Learning, with applications targeting different domains such as cybersecurity, finance, he

From Approximation to Emergence: A Theory of Deep Learning

arXiv:2607.01311v1 Announce Type: new Abstract: Deep learning has outgrown any single mathematical explanation. From Approximation to Emergence develops a unified, proof-oriented account of modern dee

Full Bayesian Reinforcement Learning via LF-IBIS

arXiv:2607.01741v1 Announce Type: cross Abstract: Reinforcement Learning (RL) is a sequential decision-making framework in which an agent learns optimal policies through interaction with an environmen

Generalization in offline RL: The structure is more important than the amount of pessimism

arXiv:2607.02288v1 Announce Type: cross Abstract: While pessimism counteracts overestimation bias in offline reinforcement learning (RL), being overly conservative has been associated with hindering c

GF-DiT: Scheduling Parallelism for Diffusion Transformer Serving

arXiv:2606.13501v2 Announce Type: replace-cross Abstract: Diffusion Transformers (DiTs) have become the dominant architecture for image and video generation, creating growing demand for efficient DiT

Guided Action Flow: Q-Guided Inference for Flow-Matching Vision-Language-Action Policies

arXiv:2607.02092v1 Announce Type: cross Abstract: Flow-matching vision-language-action policies generate robot action chunks through an iterative transport process, creating an opportunity for test-ti

HAL: Inducing Human-likeness in LLMs with Alignment

arXiv:2601.02813v3 Announce Type: replace Abstract: Aligning language models to qualitative behavioral traits, such as human-likeness, remains difficult because they are hard to define, measure, and o

Hardware-Enforced Semantic Coordination for Safety-Critical Real-Time Autonomous Systems

arXiv:2607.02376v1 Announce Type: new Abstract: Recent advances in agentic AI are producing increasingly complex autonomous systems that integrate large language models, world models, optimization eng

Kara: Efficient Reasoning LLM Serving via Sliding-Window KV Cache Compression

arXiv:2607.01237v1 Announce Type: cross Abstract: Reasoning language models often generate long chain-of-thought (CoT), which accumulates a massive KV cache during the decoding phase and incurs high d

kNNGuard: Turning LLM Hidden Activations into a Training-Free Configurable Guardrail

arXiv:2607.02072v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly deployed in domains requiring guardrails to detect unsafe, off-topic, or adversarial prompts. Existing g

Latent Policy Barrier: Learning Robust Visuomotor Policies by Staying In-Distribution

arXiv:2508.05941v2 Announce Type: replace Abstract: Visuomotor policies trained via behavior cloning are vulnerable to covariate shift, where small deviations from expert trajectories can compound int

Learning Agile Intruder Interception using Differentiable Quadrotor Dynamics

arXiv:2607.02472v1 Announce Type: new Abstract: This paper presents a methodology for learning a control policy to intercept an intruder using the 3D direction unit vector to the intruder and the inte

Learning-based Multi-agent Race Strategies in Formula 1

arXiv:2602.23056v2 Announce Type: replace Abstract: In Formula 1, race strategies are adapted according to evolving race conditions and competitors' actions. This paper proposes a reinforcement learni

Learning Semantic Atomic Skills for Multi-Task Robotic Manipulation

arXiv:2512.18368v2 Announce Type: replace Abstract: Scaling imitation learning to diverse multi-task robot manipulation remains challenging due to suboptimal demonstrations, behavioral multi-modality,

Lightweight Safe Reinforcement Learning for End-to-End UAV Navigation

arXiv:2607.01794v1 Announce Type: cross Abstract: With the rapid development of autonomous aerial systems, Unmanned Aerial Vehicles (UAVs) are increasingly deployed in applications such as inspection,

MAGIK: Mapping to Analogous Goals via Imagination-enabled Knowledge Transfer

arXiv:2506.01623v4 Announce Type: replace Abstract: Humans excel at analogical reasoning - applying knowledge from one task to a related one with minimal relearning. In contrast, reinforcement learnin

Mean Field Reinforcement Learning

arXiv:2607.01525v1 Announce Type: cross Abstract: This monograph provides an introduction to mean field reinforcement learning through the lens of Markov decision processes arising from large-populati

Memora: A Harmonic Memory Representation Balancing Abstraction and Specificity

arXiv:2602.03315v2 Announce Type: replace Abstract: Agent memory systems must accommodate continuously growing information while supporting efficient, context-aware retrieval for downstream tasks. Abs

MetaTune: Adjoint-based Meta-tuning via Robotic Differentiable Dynamics

arXiv:2603.27313v2 Announce Type: replace Abstract: Disturbance observer-based control has shown promise in robustifying robotic systems against uncertainties. However, tuning such systems remains cha

Mirror Illusion Art

arXiv:2607.02015v1 Announce Type: cross Abstract: Mirror Illusion Art is a novel reflection-conditioned 3D illusion where one object yields two target appearances (front and mirror). The task is formu

MolSight: A Graph-Aware Vision-Language Model for Unified Chemical Image Understanding

arXiv:2607.01982v1 Announce Type: cross Abstract: Using molecular large language models (LLMs) as a unified framework for understanding molecular structures and functions is emerging as a new trend in

Morphology-Aware Sample Assignment: Overcoming IoU Insensitivity for Surface Defect Detection

arXiv:2606.13723v2 Announce Type: replace-cross Abstract: Intersection-over-Union (IoU), as a pivotal metric for evaluating the spatial alignment between candidate proposals and ground-truth annotatio

Multi-modal Rail Crossing Safety Analysis

arXiv:2607.01365v1 Announce Type: cross Abstract: Given one or more images of a railway crossing, can we leverage visual cues that allow us to robustly estimate how safe it is? Can we improve our abil

Multi-Objective Exploration and Preference Optimization via Mutual Information

arXiv:2607.01392v1 Announce Type: new Abstract: Aligning large language models with diverse and heterogeneous human values requires multi-objective alignment methods to effectively trade off conflicti

safetyarxiv-cs-cl

Multilayer Q-Matrix-Embedded Neural Network for Cognitive Diagnosis (M-QCDNet): Structure-Aware Deep Learning Architecture for Psychometric Interpretability

arXiv:2607.01278v1 Announce Type: new Abstract: The research proposes a multilayer Q-matrix-embedded neural network for cognitive diagnosis (M-QCDNet), which integrates the structural interpretability

Navigating the Alignment-Calibration Trade-off: A Pareto-Superior Frontier via Model Merging

arXiv:2510.17426v3 Announce Type: replace-cross Abstract: The 'alignment tax' of post-training is typically framed as a drop in task accuracy. We show it also involves a severe loss of calibration, ma

NeoMap: Training-free Novel-View Synthesis from Single Images and Videos

arXiv:2607.01962v1 Announce Type: cross Abstract: We study the challenging problem of novel view video synthesis from single images or monocular videos. Existing methods, which operate under the assum

Neuron-Aware Data Selection for Annotation-Free LLM Self-Distillation

arXiv:2607.02460v1 Announce Type: cross Abstract: Post-training large language models (LLMs) without real-world interaction feedback or human-labeled supervision remains challenging, particularly in s

Object Aligner: A Configurable JSON Schema Similarity Score for Graphs, Applied to LLM Prompt Optimization

arXiv:2607.01972v1 Announce Type: cross Abstract: Large language models (LLMs) are often asked to produce JSON conforming to a fixed schema, powering information extraction, tool calling, agentic plan

On the Role of Computation in Reinforcement Learning

arXiv:2602.05999v4 Announce Type: replace Abstract: How does the amount of compute available to a reinforcement learning (RL) policy affect its learning? Can policies using a fixed amount of parameter

On the Sample Efficiency of Inverse Dynamics Models for Semi-Supervised Imitation Learning

arXiv:2602.02762v2 Announce Type: replace Abstract: Semi-supervised imitation learning (SSIL) consists in learning a policy from a small dataset of action-labeled trajectories and a much larger datase

One Demonstration Is Enough for Real-World Robotic Reinforcement Learning

arXiv:2607.01651v1 Announce Type: new Abstract: Learning effective robot control policies on physical hardware is challenging due to costly data collection and the difficulty of reward specification.

Online Resource Allocation with Continuous Random Consumption: Regret under Degeneracy

arXiv:2607.02196v1 Announce Type: new Abstract: We study online resource allocation when both rewards and consumption sizes may be continuously distributed. Requests arrive sequentially and must be ac

Online Safety Monitoring for LLMs

arXiv:2607.02510v1 Announce Type: new Abstract: Despite alignment training, LLMs remain prone to generating unsafe outputs at deployment time. Monitoring outputs online and raising an alarm when safet

OpenAI offers feds a stake, Anthropic gets out of AI model jail and Meta wants to be a neocloud

OpenAI reportedly has floated giving the U.S. government a 5% stake in the company, perhaps the start of a series of such stakes in other AI companies as well. This no doubt has traditional anti-indus

safetysiliconangle

Optimizing Visual Generative Models via Distribution-wise Rewards

arXiv:2607.02291v1 Announce Type: new Abstract: Conventional reinforcement learning strategies for visual generation typically employ sample-wise reward functions, yet this practice frequently results

Overthink-Triggered Slowdown Attacks on LVLM-Based Robotic Systems

arXiv:2607.01518v1 Announce Type: cross Abstract: Large Vision-Language Models (LVLMs) have been increasingly integrated into robotic systems. However, these models may exhibit overthinking behaviors,

Playing 20 Question Game with Policy-Based Reinforcement Learning

arXiv:1808.07645v5 Announce Type: replace-cross Abstract: The 20 Questions (Q20) game is a well known game which encourages deductive reasoning and creativity. In the game, the answerer first thinks o