AI safety and alignment

Alignment, interpretability, red-teaming, bias and privacy: the research on what these systems do when they misbehave.

17,884 stories · RSS feed

arXiv Computer Vision

Sep 2

Adversarial Attack and Disturbance Detection by Hadamard-Coded Output Representations for Object Detection and Semantic Segmentation

arXiv:2606.09536v2 Announce Type: replace Abstract: Conventional one-hot encodings often yield poorly calibrated models, being overconfident under attack, and letting entropy-based detection algorith...

By Lucas G\"ornhardt, Timo Bartels, Niklas Schwarz, Tim Fingscheidt

computer-vision benchmarks safety

arXiv Computer Vision

Sep 2

Efficient and Robust Absolute Pose Estimation via Gravity-Prior-Driven Transformation Decoupling and Pose Refinement

arXiv:2609.00713v1 Announce Type: new Abstract: Estimation of the absolute pose of an object is an essential task for various robotic applications. Recently, incorporating gravity direction as prior...

By Hu Cao, Qianyi Yang, Xinyi Li, Jiong Liu, Yinlong Liu, Alois Knoll

computer-vision robotics benchmarks safety

arXiv Computer Vision

Sep 2

SurgiATM: A Physics-Guided Plug-and-Play Model for Deep Learning-Based Smoke Removal in Laparoscopic Surgery

The paper introduces SurgiATM, a lightweight physics-guided module for removing surgical smoke from laparoscopic endoscopic frames. It integrates a physics-based atmospheric model with a data-driven deep learning approach via a Mixture-of-Experts output stage, using a Laplacian-like error distribution to model smoke. SurgiATM adds only two hyperparameters and no extra trainable weights, enabling easy integration into existing desmoking architectures and improving accuracy and stability across multiple datasets and procedures.

By Mingyu Sheng, Jianan Fan, Dongnan Liu, Guoyan Zheng, Ron Kikinis, Weidong Cai

safety

arXiv Machine Learning

Sep 2

Explicit Interaction Architectures for Dynamical Learning: A Controlled Study of Structural Inductive Bias

arXiv:2606.19101v2 Announce Type: replace-cross Abstract: We investigate a structure-first approach to dynamical learning in which the organization of stateful interactions is prescribed explicitly r...

By Augusto Sarti

safety

arXiv Computation and Language

Sep 2

Slow to See, Slow to Suppress: Understanding the Effects of Modality in Context-Memory Conflicts

arXiv:2609.00293v1 Announce Type: new Abstract: We investigate how vision-language models (VLMs) handle context-memory conflicts; that is, situations in which the model is given information in contex...

By Athulith Paraselli, Etha Tianze Hua, Ellie Pavlick

llms rag multimodal safety

arXiv Computation and Language

Sep 2

Location-Aware Language Models via Secondary Embeddings

arXiv:2609.00454v1 Announce Type: new Abstract: Pretrained transformer-based language models achieve strong performance across a wide range of NLP tasks but remain limited in encoding geo-locational...

By Gokul Srinivasagan, Munir Georges

llms rag nlp benchmarks safety

arXiv Computation and Language

Sep 2

Enoki: Efficient Multi-Level Hallucination Detection

arXiv:2609.00581v1 Announce Type: new Abstract: Ensuring factuality remains a critical challenge for deploying LLMs in high-stakes settings. Existing hallucination detectors usually operate at a sing...

By Elisei Rykov, Timur Ionov, Nikolay Ivanov, Maksim Savkin, Maksim Makarenko, Alexander Panchenko, Vasily Konovalov, Julia Belikova

llms safety

arXiv Computation and Language

Sep 2

Investigating Assistant Bias in LLM User Simulators Using a Role Vector

arXiv:2609.00608v1 Announce Type: new Abstract: LLM-based user simulators are increasingly used to evaluate autonomous agents at scale, in place of costly human evaluations. Despite this promise, the...

By Daeheon Jeong, Yoonjoo Lee, Eugene Choi, Sinie van der Ben, Juho Kim

llms agents safety

arXiv Computer Vision

Sep 2

Conditional Flow Matching for Cross-Field MRI Harmonisation

arXiv:2609.00960v1 Announce Type: new Abstract: Magnetic resonance images of the same subject look markedly different across field strengths, which complicates the comparison and pooling of data acro...

By Baris Imre, Aram Salehi, Levente Baljer, Andrew Webb, Marius Staring, Efe Ilicak

diffusion fine-tuning safety

arXiv Computation and Language

Sep 2

Trust Your Guide Only When Certain: Uncertainty-Aware Sparse Alignment at Inference Time

arXiv:2609.00624v1 Announce Type: new Abstract: A prominent paradigm in inference-time alignment employs lightweight supervisors to steer Large Language Models (LLMs). Through empirical analysis, we...

By Zeen Zhu, Zhuo Li, Weiyang Guo, Liye Zhao, Haibing Di, Yequan Wang, Jing Li

llms benchmarks safety

arXiv Machine Learning

Sep 2

NeuroPriv: Adversarial Representation Learning for Privacy in Wearable EEG Systems

arXiv:2609.00390v1 Announce Type: cross Abstract: Wearable EEG systems may expose sensitive information beyond their intended health function, creating substantial risks to neuroprivacy. In this work...

By Sarmistha Sarna Gomasta, Bhawana Chhaglani, Prashant Shenoy

safety

arXiv Machine Learning

Sep 2

Real-Time Neuromorphic Spectrum Intelligence Simulator

arXiv:2609.00585v1 Announce Type: cross Abstract: We present the Real-Time Neuromorphic Spectrum Intelligence Simulator (RT-NuSIS), a modular framework to study spiking neural network (SNN) and memri...

By Navaneetha Krishnan Kamalakannan

agents benchmarks safety

arXiv Machine Learning

Sep 2

Exploring Sparse Autoencoders in Text-Based Causal Confounding Adjustment

arXiv:2609.01322v1 Announce Type: cross Abstract: In many settings, studying causal questions based on text data requires adjusting for confounding information within texts. Yet there is a tradeoff i...

By Mian Zhong, Katherine A. Keith, Anjalie Field

safety

arXiv Machine Learning

Sep 2

Variable Selection for Feature-Based Newsvendor

arXiv:2609.01544v1 Announce Type: cross Abstract: Feature-based newsvendor models use observable covariates to tailor inventory decisions, aiming to balance holding and shortage costs under demand un...

By Zhaoliang Yuan, Jie Wang

safety

arXiv Machine Learning

Sep 2

Stress Testing Unlearning Algorithms

arXiv:2608.22527v2 Announce Type: replace Abstract: Recently, machine unlearning, the removal of specific training data influence from a model, has gained increasing attention. In large language mode...

By Noam Diamant, Neta Glazer, Ethan Fetaya

llms benchmarks safety

arXiv Machine Learning

Sep 2

Fair Minimum Labeling: Efficient Temporal Network Activations for Reachability and Equity

arXiv:2510.03899v3 Announce Type: replace-cross Abstract: Balancing resource efficiency and fairness is critical in networked systems that support modern learning applications. We introduce the \emph...

By Lutz Oettershagen, Othon Michail

safety

arXiv Computation and Language

Sep 2

Subword Segmental BabyLMs: Learning to Tokenise for Sample-Efficient Pretraining

arXiv:2609.01151v1 Announce Type: new Abstract: In the standard LM training pipeline, subword tokenisation is applied as a preprocessing step. Subword segmental language modelling is an alternative p...

By Francois Meyer

computer-vision safety

arXiv Machine Learning

Sep 2

Silence is Golden: Mitigating Hallucinations in Large Audio-Language Models via Layer-Weighted Vector Steering

arXiv:2510.12851v2 Announce Type: replace-cross Abstract: Large Audio-Language Models (LALMs) excel in Audio QA but often suffer from hallucinations ungrounded in the audio. To our knowledge, we are...

By Tsung-En Lin, Kuan-Yi Lee, Hung-Yi Lee

llms benchmarks safety

arXiv Computation and Language

Sep 2

LLMPEDIA: Browsing, Verifying, and Comparing the Parametric Encyclopedic Knowledge of LLMs

arXiv:2609.01182v1 Announce Type: new Abstract: Flagship language models appear saturated on benchmarks like MMLU (Hendrycks et al., 2021), scoring above 90% - yet benchmarks test only what the exper...

By Muhammed Saeed, Simon Razniewski

llms benchmarks safety

arXiv Computer Vision

Sep 2

Beyond Language Priors: Diagnosing and Fixing Visual-Origin Hallucinations in Multimodal LLM

arXiv:2609.00231v1 Announce Type: new Abstract: Existing research on object hallucination in multimodal large language models (MLLMs) predominantly attributes the problem to language priors such as o...

By Peiyang Xu, Xiaopei Zhu, Jun Zhu, Xiaolin Hu

llms rag fine-tuning multimodal benchmarks safety

Get the daily briefing