Safe imitation learning with a safety layer for flexible training.
problem Flexible yet safe imitation learning for complex tasks.
method Theory and modular method with a safety layer for continuous policy, adversarial training, and worst-case safety guarantees.
result Robustness advantage of safety layer during training compared to test time.
The paper tackles safe exploration in RL by a conservative safety critic.
problem Safe exploration in reinforcement learning (RL) when partially trained policies are deployed.
method Learning a conservative safety estimate through a critic, provably bounding catastrophic failures.
result The approach provably converges to competitive task performance with significantly lower catastrophic failure rates.
RLVR maintains safety while improving reasoning capabilities in LLMs.
problem Safety-capability tradeoff in fine-tuning LLMs.
method Reinforcement Learning with Verifiable Rewards (RLVR) and theoretical analysis.
result RLVR can enhance reasoning while maintaining safety guardrails.
CARL safely adapts RL agents for safety-critical tasks.
problem Safety hazards in RL for safety-critical tasks.
method CARL combines model-based RL and cautious adaptation.
result CARL achieves higher rewards with fewer failures in safety-critical tasks.
New method provides scalable safety guarantees for RL agents.
problem Safe reinforcement learning in real-life scenarios.
method State-augmentation and shield design for probabilistic avoidance.
result Strict formal safety guarantees for RL agents, scalable and practical.
Safe learning of stochastic dynamics with safety constraints.
problem Learning controlled stochastic dynamics with safety constraints.
method Iterative expansion of a safe control set using kernel-based confidence bounds.
result The method ensures safe exploration and efficient estimation of system dynamics.
Paper explores anchoring for vision models, improving generalization and safety.
problem Anchoring can lead to undesirable shortcuts, limiting generalization.
method Introduced a new anchored training protocol with a regularizer to mitigate undesirable shortcuts.
result Significant performance gains in generalization and safety metrics.
SECRM-2D improves RL-based autonomous driving with safety guarantees.
problem Safety and efficiency trade-offs in RL-based autonomous driving.
method RL-based controller with safety constraints for efficient and comfortable driving.
result SECRM-2D avoids crashes and improves efficiency and comfort compared to baselines.
This paper formalizes AI safety using hypothesis testing in GenAI.
problem Ensuring safety of generative AI tools that create realistic content.
method Formalization of computational safety through hypothesis testing and signal processing.
result Demonstrates how AI safety can be assessed quantitatively using mathematical frameworks.
Safe Active Learning (SAL) for GPODE models improves model performance without compromising safety.
problem Efficient and safe data collection for GPODE models.
method Sequential data suggestion constrained by safety, maximizing information gain.
result SAL GPODE outperforms standard data collection methods.
Machine learning (ML) techniques are increasingly applied to decision-making and control problems in Cyber-Physical Systems among which many are safety-critical, e.g., chemical plants, robotics, autonomous vehicles. Despite the significant benefits brought by ML techniques, they also raise additional safety issues beca…
Enhances safety of 3D object detection neural networks.
problem Ensuring robustness and safety of 3D object detection systems.
method Symbolic error propagation, specialized loss function, safety-aware non-max-inclusion algorithm.
result Improved safety and robustness of 3D object detection neural networks.
Artificial Neural Networks (ANNs) have demonstrated remarkable utility in various challenging machine learning applications. While formally verified properties of their behaviors are highly desired, they have proven notoriously difficult to derive and enforce. Existing approaches typically formulate this problem as a p…
Safe reinforcement learning framework using optimal transport for robustness.
problem Robustness and safety in deep reinforcement learning with limited data assumptions.
method Optimal transport perturbations to construct worst-case virtual state transitions.
result Significantly improved safety at deployment time compared to standard methods.
We propose Trusted Neural Network (TNN) models, which are deep neural network models that satisfy safety constraints critical to the application domain. We investigate different mechanisms for incorporating rule-based knowledge in the form of first-order logic constraints into a TNN model, where rules that encode safet…
Urban traffic systems worldwide are suffering from severe traffic safety problems. Traffic safety is affected by many complex factors, and heavily related to all drivers' behaviors involved in traffic system. Drivers with aggressive driving behaviors increase the risk of traffic accidents. In order to manage the safety…
Study probabilistic safety of BNNs under adversarial attacks.
problem Evaluate vulnerability of BNNs to adversarial attacks.
method Relaxation techniques from non-convex optimization to compute probabilistic safety bounds.
result Certify probabilistic safety of BNNs with millions of parameters.
A novel approach for safe offline RL using latent safety constraints.
problem Balancing safety constraints and reward maximization in offline RL.
method Conditional Variational Autoencoders for latent safety modeling, Constrained Reward-Return Maximization.
result Our approach maintains safety compliance while optimizing rewards, outperforming existing methods.
Reinforcement learning is a promising approach to synthesizing policies for challenging robotics tasks. A key problem is how to ensure safety of the learned policy---e.g., that a walking robot does not fall over or that an autonomous car does not run into an obstacle. We focus on the setting where the dynamics are know…
A new method reduces the computational burden of safety alignment for large language models.
problem Safety concerns in large language models and the need to align them with human preferences.
method Optimal dualization approach to reduce constrained alignment to an unconstrained problem.
result Our algorithms MoCAN and PeCAN significantly reduce computational burden and improve training stability.
Autonomous agents trained via reinforcement learning present numerous safety concerns: reward hacking, negative side effects, and unsafe exploration, among others. In the context of near-future autonomous agents, operating in environments where humans understand the existing dangers, human involvement in the learning p…
Machine learning algorithms are increasingly influencing our decisions and interacting with us in all parts of our daily lives. Therefore, just like for power plants, highways, and myriad other engineered sociotechnical systems, we must consider the safety of systems involving machine learning. In this paper, we first …
Simple online monitor detects unsafe LLM outputs.
problem LLMs generate unsafe outputs despite training.
method Thresholding external verifier signal to decide alarms.
result Simple design competitive with advanced methods.
The paper shows why AI safety doesn't generalize across tasks.
problem AI safety fails to generalize across unseen tasks.
method Theoretical analysis of linear-quadratic control with H∞-robustness, empirical demonstrations in simulated quadcopter navigation and CRM. result The mapping from task specification to an optimal controller has a higher Lipschitz constant with safety requirements than without, indicating inherent complexity of safety.
Study evaluates conformal prediction methods for safety in vision models under shifts and long-tailed data.
problem Safety guarantees of conformal prediction methods under distribution shifts and long-tailed data.
method Empirical evaluation of post-hoc and training-based conformal prediction methods on large-scale datasets and models.
result Performance of conformal prediction methods degrades significantly under distribution shifts and long-tailed data.
In many real-world reinforcement learning (RL) problems, besides optimizing the main objective function, an agent must concurrently avoid violating a number of constraints. In particular, besides optimizing performance it is crucial to guarantee the safety of an agent during training as well as deployment (e.g. a robot…
Paper reviews robustness in machine learning models and discusses training and certification methods.
problem Ensuring reliability of machine learning models in safety-critical systems.
method Reviews formalisms and discusses training and certification techniques.
result Identifies future research directions in robust machine learning.
In order to satisfy safety conditions, an agent may be constrained from acting freely. A safe controller can be designed a priori if an environment is well understood, but not when learning is employed. In particular, reinforcement learned (RL) controllers require exploration, which can be hazardous in safety critical …
A new RL model ensures safe learning in uncertain environments.
problem Safe reinforcement learning in uncertain, partially observable environments.
method Lyapunov-based uncertainty quantification and Transformers for memory.
result Significant improvement in safety and optimality in grid-world tasks.
Neural networks are increasingly deployed in real-world safety-critical domains such as autonomous driving, aircraft collision avoidance, and malware detection. However, these networks have been shown to often mispredict on inputs with minor adversarial or even accidental perturbations. Consequences of such errors can …
Neural Networks are being integrated into safety critical systems, e.g., perception systems for autonomous vehicles, which require trained networks to perform safely in novel scenarios. It is challenging to verify neural networks because their decisions are not explainable, they cannot be exhaustively tested, and finit…
ELS framework improves safety alignment by dynamically steering LLMs towards helpful responses.
problem Over-Refusal in Aligned Large Language Models
method Fine-tuning free framework using an Energy-Based Model (EBM) to dynamically steer LLMs during inference.
result Extensive experiments show a significant reduction in false refusals (from 57.3% to 82.6%) while maintaining safety performance.
Safe-EF improves federated learning for non-smooth, constrained optimization.
problem Federated learning's communication bottlenecks with high-dimensional model updates.
method Error feedback (EF) for non-smooth convex optimization with safety constraints.
result Safe-EF matches lower complexity bounds and ensures safety constraints.
Study investigates how machine learning models degrade over time, leading to patient safety issues.
problem Overtime degradation of machine learning models in clinical settings.
method Used MIMIC-IV dataset to train models replicating commercial approaches, observing and analyzing degradation over a decade.
result An RNN model built on Epic features degrades from 0.729 AUC to 0.525 AUC over a decade, highlighting technical and clinical drift as root causes.
A new framework ensures model safety by retaining old model capabilities while improving new tasks.
problem Catastrophic forgetting in continual learning systems.
method Retention-centric framework with data-dependent constraints and an efficient constrained optimization algorithm.
result The method ensures that new models retain important capabilities of old models while improving new tasks.
Safe RL with binary feedback using SABRE algorithm.
problem Safe reinforcement learning with binary safety feedback.
method SABRE algorithm, combining active learning and reinforcement learning.
result Provable safe policy with high probability, no unsafe actions during training.
Classifier-based AI safety gates fail in self-improvement, even with advanced verification methods.
problem Maintaining reliable oversight of AI systems as they improve over iterations.
method Comprehensive empirical testing on neural controllers and MuJoCo benchmarks, using various classifiers and verification methods.
result Classifier-based safety gates fail in maintaining reliable oversight, even with advanced verification methods.
Survey of alignment techniques for large language models.
problem Ensuring large language models align with human values.
method Analysis of diverse alignment methods and training paradigms.
result Preference-based methods offer more flexibility for nuanced alignment.
This paper explores formal verification for autonomous systems, identifying limitations and proposing improvements.
problem Ensuring safety of autonomous systems like self-driving cars and drones.
method Formal verification techniques based on formal methods, analyzing three assumptions and their limitations.
result Preliminary work to improve the strength of evidence provided by formal verification.
Deep learning has become an increasingly common technique for various control problems, such as robotic arm manipulation, robot navigation, and autonomous vehicles. However, the downside of using deep neural networks to learn control policies is their opaque nature and the difficulties of validating their safety. As th…
Autonomous vehicles rely on machine learning to solve challenging tasks in perception and motion planning. However, automotive software safety standards have not fully evolved to address the challenges of machine learning safety such as interpretability, verification, and performance limitations. In this paper, we revi…
Fine-tuning LLMs improves capability but harms safety, study finds.
problem Balancing capability and safety in LLM fine-tuning.
method Theoretical framework and numerical experiments for two safety-aware fine-tuning strategies.
result Characterization of fundamental limits of safety-capability trade-off in LLM fine-tuning.
This paper significantly improves on, and finishes to validate, an approach proposed in previous research in which safety outcomes were predicted from attributes with machine learning. Like in the original study, we use Natural Language Processing (NLP) to extract fundamental attributes from raw incident reports and ma…
RAGuard improves safety in LLMs for offshore wind maintenance.
problem Conventional LLMs fail with specialised or unexpected scenarios in offshore wind maintenance.
method Integrates safety-critical documents alongside technical manuals in RAG framework.
result RAGuard increases safety recall from almost 0% to over 50% while maintaining technical recall above 60%.
Survey of algorithms for testing AI-driven CPS safety.
problem Testing AI-driven CPS for safety in complex environments.
method Survey of applied algorithms for safety validation.
result Survey of existing tools and techniques for safety validation.
Deep learning methods are widely regarded as indispensable when it comes to designing perception pipelines for autonomous agents such as robots, drones or automated vehicles. The main reasons, however, for deep learning not being used for autonomous agents at large scale already are safety concerns. Deep learning appro…
Before deploying autonomous agents in the real world, we need to be confident they will perform safely in novel situations. Ideally, we would expose agents to a very wide range of situations during training, allowing them to learn about every possible danger, but this is often impractical. This paper investigates safet…
Bayesian approach scores influential training examples for model predictions.
problem Enhance interpretability and safety of machine learning models.
method Formulate TDA as a Bayesian information-theoretic problem, scoring subsets by information loss.
result Method aligns with classical influence scores while promoting diversity for subsets.