Pre-Deployment Safety Assessments

Follow this topic Ask about this topic

Momentum

0 papers in the last four weeks, against 1 the four weeks before. 0.0% of all new papers.

Jul 6Week of Sep 21

Latest papers 24

All topics

When Masking Helps or Hurts Robustness in Compressed CLIP: A Pre-Deployment Diagnostic
Sep 30, 2026Muhammad Zawish, Steven DavyVisual Token Pruning Mask2Former
Technological University Dublin, Ireland
Reliability Testing of Medical Model Performance under Distributed Deployment
Sep 29, 2026Yifei Wang, Xiaohan Zhang, Youtao Ding +4Model Evaluation Pre-Deployment Safety Assessments
Shanghai Jiao Tong University, China · Beihang University, China · Nanyang Technological University, Singapore +1
A Browser-Native Digital Test Range for Benchmarking 4D Ocean-Glider Planning Algorithms
Aug 13, 2026Edward Holmberg, Elias Ioup, Mahdi AbdelguerfiGeometric Bathymetric Data Pre-Deployment Safety Assessments
Gulf States Center for Environmental Informatics · LSU New Orleans, LA, USA · Ocean Sciences Division +2
Knowledge Graphs and Explainable AI as Complementary Resources for Urban Mining
Jul 10, 2026Jan Gronewald, Andreas Emrich, Nijat MehdiyevXai Explainable Artificial Intelligence
German Research Center for Artificial Intelligence (DFKI), Saarbrücken, Germany
Predicting LLM Safety Before Release by Simulating Deployment
Jul 8, 2026Marcus Williams, Hannah Sheahan, Cameron Raymond +8Pre-Deployment Safety Assessments Large Language Model Safety
OpenAI
Open Problems in AI Incident Governance
Jul 6, 2026Harleen Kaur Sidhu, Rebecca Scholefield, Nour Annan +5Artificial Intelligence Risk Pre-Deployment Safety Assessments
Independent · Sorbonne University · Rice University +4
CAGE-1: Control, Assurance, and Governance Evaluation for Enterprise Agentic AI
Jul 3, 2026Roopam W. SureAgentic Evaluations Governance
RouterVLA: Turning Smoke Tests into Supervision for Heterogeneous VLA Selection
Jun 25, 2026Xingyu Ren, Chugang Yi, Ge Ma +1Trajectory-Level Credit Destination-Only Scoring
1The Chinese University of Hong Kong · University of Maryland, College Park · 3Tsinghua University
When EER Hides Deployment Failure: Auditing Threshold Transfer and Unlabeled Score Calibration for Speech Deepfake Detectors
Jun 19, 2026Jingwen Zhou, Mingzhe WangEqual Error Rate In-Domain Accuracy
Xidian University, Xi’an, China
Frozen Foundation-Model Embeddings Discard Small-Lesion Signal in Chest Radiography: Implications for Pre-Deployment Evaluation
Jun 10, 2026Raajitha Muthyala, Zhenan Yin, Alekhya Jilla +6Chest X-Ray Frozen Vision-Language Models
Department of Biomedical Engineering and Informatics, Indiana University, 535 W Michigan St., IT 475J, Indianapolis, IN, 46202, USA · Department of Radiology and Imaging Sciences, Emory University, 1364 Clifton Rd NE, Atlanta, GA, 30322, USA
Toward Pre-Deployment Assurance for Enterprise AI Agents: Ontology-Grounded Simulation and Trust Certification
Jun 2, 2026Thanh Luong Tuan, Abhijit SanyalAgentic Deployments Pre-Deployment Safety Assessments
Golden Gate University, San Francisco, CA, USA · Data, Digital & IT, Novartis Healthcare Pvt. Ltd., Hyderabad, India
Silent Failures in Federated Personalization of Foundation Models
May 31, 2026YongKyung Oh, Alex BuiFederated Learning AI Trustworthiness
Medical & Imaging Informatics (MII) Group, University of California, Los Angeles (UCLA) Los Angeles, CA, 90024, USA
Operational AI Deployment Assurance: Governance-State Orchestration Under Threshold-Sensitive Deployment Conditions -- A Governance Framework for High-Stakes AI Systems
May 27, 2026Khalid Adnan AlsayedArtificial Intelligence Governance Governance
Ducaltus — AI Assurance & Governance Newcastle upon Tyne, United Kingdom · School of Computing, Engineering & Digital Technologies Teesside University Middlesbrough, United Kingdom
Deployment-complete benchmarking
May 25, 2026El Mustapha Mansouri, Keigo AraiPre-Deployment Safety Assessments Responses
School of Engineering, Institute of Science Tokyo, Yokohama, Kanagawa 226-8501, Japan
Healthcare LLM Benchmarks Are Only as Good as Their Explicit Assumptions
May 21, 2026Naveen Raman, Santiago Cortes-Gomez, Mateo Dulce Rubio +2Healthcare Pre-Deployment Safety Assessments
New York University
A Deployment Audit of Release-Side Risk in Conformal Triage under Prevalence Shift
May 20, 2026Chengze Li, Xiao Liu, Hanrong Zhang +7Triage Pre-Deployment Safety Assessments
University of Illinois Chicago, Chicago, IL, USA · Manteia Technologies Co., Ltd, Xiamen, China · University of Illinois Urbana-Champaign, Champaign, IL, USA +1
Training ML Models with Predictable Failures
May 14, 2026Will Schwarzer, Scott NiekumEarly Failure Prediction Pre-Deployment Safety Assessments
University of Massachusetts Amherst
RISED: A Pre-Deployment Evaluation Framework for High-Stakes AI Decision-Support Systems, with Application to Healthcare
May 13, 2026Rohith Reddy Bellibatlu, Manpreet Singh, Yash Jajoo +2Clinical Decision Support Pre-Deployment Safety Assessments
Florida International University, Miami, FL, USA. · Boston University, Boston, MA, USA. · New York University, New York, NY, USA. +2
Measuring Evaluation-Context Divergence in Open-Weight LLMs: A Paired-Prompt Protocol with Pilot Evidence of Alignment-Pipeline-Specific Heterogeneity
May 7, 2026Florian A. D. Burnat, Brittany I. DavidsonLarge Language Model Evaluation Pre-Deployment Safety Assessments
University of Bath
Code World Model Preparedness Report
May 1, 2026Daniel Song, Peter Ney, Cristina Menghini +21Pre-Deployment Safety Assessments Code Generation
MSL Preparedness Team, AI Security Team
Risk Reporting for Developers' Internal AI Model Use
Apr 27, 2026Oscar Delaney, Sambhav Maheshwari, Joe O'Brien +2Artificial Intelligence Risk Pre-Deployment Safety Assessments
† Work done while at IAPS
Bridging the Training-Deployment Gap: Gated Encoding and Multi-Scale Refinement for Efficient Quantization-Aware Image Enhancement
Apr 23, 2026Dat To-Thanh, Nghia Nguyen-Trong, Hoang Vo +2Quantization-Aware Training Mobilenetv2
University of Science, VNU-HCM, Vietnam · Vietnam National University, Ho Chi Minh City, Vietnam · University of Information Technology, VNU-HCM, Vietnam +2
Mythos and the Unverified Cage: Z3-Based Pre-Deployment Verification for Frontier-Model Sandbox Infrastructure
Apr 22, 2026Dominik BlainPre-Deployment Safety Assessments Sandbox
COBALT Formal Verification, QreativeLab Inc. Montr´eal, Qu´ebec, Canada
The Geometric Canary: Predicting Steerability and Detecting Drift via Representational Stability
Apr 20, 2026Prashant C. RajuPre-Deployment Safety Assessments

Sep 30, 2026, cs.CVJ/K move · Enter open · S save

When Masking Helps or Hurts Robustness in Compressed CLIP: A Pre-Deployment Diagnostic

Muhammad Zawish, Steven Davy

Technological University Dublin, Ireland

Visual Token Pruning Mask2Former Contrastive Language-Image Pre-Training Model Ablation Pre-Deployment Safety Assessments Inverse Design

Open paper AskPDF arXiv

This paper demonstrate that whether masking-based token pruning helps or hurts worst-group robustness can be predicted before deployment, without labels or fine-tuning. A systematic study of semantic masking across 8 spurious-correlation benchmarks shows its effect on worst-group accuracy is highly unstable: it improves accuracy by up to 82.5% relative on some datasets and degrades it by up to 100% on others. We trace this instability to spurious inversion: background patches receive higher CLIP text-similarity than the true object when the spurious attribute is background-separable, inverting the assumption every text- and attention-guided pruning method relies on. We introduce the Spurious Inversion Metric (SIM), a label-free, pre-deployment diagnostic whose sign predicts this effect with statistical significance (binomial

p=0.035

) across all 8 datasets, and remains dependable across 6 CLIP architectures with a clean foreground/background split. Naive masking is itself a major source of risk: it causes the largest average-accuracy loss of any method we evaluate, and its own per-image segmentation step is a significant runtime bottleneck. To address this, we design a batched, synchronization-free GPU segmentation routine that cuts this overhead from 3.5

\times

to 1.75

\times

baseline. Gating deployment by SIM's sign recovers masking's benefits while avoiding its worst failures, matching or exceeding a strong pruning baseline on 7 of 8 datasets.

Momentum

Related topics

Latest papers 24