Cover Image

What Makes an Evaluation Useful? Common Pitfalls and Best Practices

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Gekker, Gil, Segal, Meirav, Lahav, Dan, Nevo, Omer
Format:	Preprint
Published:	2025
Subjects:	Machine Learning Artificial Intelligence Computation and Language Cryptography and Security Computers and Society
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

A Content-Based Framework for Cybersecurity Refusal Decisions in Large Language Models
by: Linder, Noa, et al.
Published: (2026)

Leveraging LLM Inconsistency to Boost Pass@k Performance
by: Dalal, Uri, et al.
Published: (2025)

Cybench: A Framework for Evaluating Cybersecurity Capabilities and Risks of Language Models
by: Zhang, Andy K., et al.
Published: (2024)

LLM Platform Security: Applying a Systematic Evaluation Framework to OpenAI's ChatGPT Plugins
by: Iqbal, Umar, et al.
Published: (2023)

Making AI-Assisted Grant Evaluation Auditable without Exposing the Model
by: Bicakci, Kemal
Published: (2026)

DOMBA: Double Model Balancing for Access-Controlled Language Models via Minimum-Bounded Aggregation
by: Segal, Tom, et al.
Published: (2024)

The Pitfalls of "Security by Obscurity" And What They Mean for Transparent AI
by: Hall, Peter, et al.
Published: (2025)

Misaligned Roles, Misplaced Images: Structural Input Perturbations Expose Multimodal Alignment Blind Spots
by: Shayegani, Erfan, et al.
Published: (2025)

Just Do It!? Computer-Use Agents Exhibit Blind Goal-Directedness
by: Shayegani, Erfan, et al.
Published: (2025)

Urania: Differentially Private Insights into AI Use
by: Liu, Daogao, et al.
Published: (2025)

Clio: Privacy-Preserving Insights into Real-World AI Use
by: Tamkin, Alex, et al.
Published: (2024)

Can a large language model be a gaslighter?
by: Li, Wei, et al.
Published: (2024)

An In-Depth Investigation of Data Collection in LLM App Ecosystems
by: Wu, Yuhao, et al.
Published: (2024)

Superficial Safety Alignment Hypothesis
by: Li, Jianwei, et al.
Published: (2024)

IsolateGPT: An Execution Isolation Architecture for LLM-Based Agentic Systems
by: Wu, Yuhao, et al.
Published: (2024)

Prompt Attacks Reveal Superficial Knowledge Removal in Unlearning Methods
by: Jang, Yeonwoo, et al.
Published: (2025)

Generative AI Security: Challenges and Countermeasures
by: Zhu, Banghua, et al.
Published: (2024)

Best-of-Venom: Attacking RLHF by Injecting Poisoned Preference Data
by: Baumgärtner, Tim, et al.
Published: (2024)

Watermarking Makes Language Models Radioactive
by: Sander, Tom, et al.
Published: (2024)

Intent Laundering: AI Safety Datasets Are Not What They Seem
by: Golchin, Shahriar, et al.
Published: (2026)

NYU CTF Bench: A Scalable Open-Source Benchmark Dataset for Evaluating LLMs in Offensive Security
by: Shao, Minghao, et al.
Published: (2024)

What is in Your Safe Data? Identifying Benign Data that Breaks Safety
by: He, Luxi, et al.
Published: (2024)

OBLIVIATE: Robust and Practical Machine Unlearning for Large Language Models
by: Xu, Xiaoyu, et al.
Published: (2025)

GaussMark: A Practical Approach for Structural Watermarking of Language Models
by: Block, Adam, et al.
Published: (2025)

HeavyWater and SimplexWater: Distortion-Free LLM Watermarks for Low-Entropy Next-Token Predictions
by: Tsur, Dor, et al.
Published: (2025)

Defense Against the Dark Prompts: Mitigating Best-of-N Jailbreaking with Prompt Evaluation
by: Armstrong, Stuart, et al.
Published: (2025)

TurboFuzzLLM: Turbocharging Mutation-based Fuzzing for Effectively Jailbreaking Large Language Models in Practice
by: Goel, Aman, et al.
Published: (2025)

RLShield: Practical Multi-Agent RL for Financial Cyber Defense with Attack-Surface MDPs and Real-Time Response Orchestration
by: Nayak, Srikumar
Published: (2026)

PIArena: A Platform for Prompt Injection Evaluation
by: Geng, Runpeng, et al.
Published: (2026)

Rethinking How to Evaluate Language Model Jailbreak
by: Cai, Hongyu, et al.
Published: (2024)

Cross-Session Threats in AI Agents: Benchmark, Evaluation, and Algorithms
by: Azarafrooz, Ari
Published: (2026)

LLM Cyber Evaluations Don't Capture Real-World Risk
by: Lukošiūtė, Kamilė, et al.
Published: (2025)

On Evaluating The Performance of Watermarked Machine-Generated Texts Under Adversarial Attacks
by: Liu, Zesen, et al.
Published: (2024)

There Are No Silly Questions: Evaluation of Offline LLM Capabilities from a Turkish Perspective
by: Yilmaz, Edibe, et al.
Published: (2026)

Expected Harm: Rethinking Safety Evaluation of (Mis)Aligned LLMs
by: Chen, Yen-Shan, et al.
Published: (2026)

Automatic Pseudo-Harmful Prompt Generation for Evaluating False Refusals in Large Language Models
by: An, Bang, et al.
Published: (2024)

Confidential Guardian: Cryptographically Prohibiting the Abuse of Model Abstention
by: Rabanser, Stephan, et al.
Published: (2025)

Adversarial Augmentation and Active Sampling for Robust Cyber Anomaly Detection
by: Benabderrahmane, Sidahmed, et al.
Published: (2025)

Robust Safety Monitoring of Language Models via Activation Watermarking
by: Aremu, Toluwani, et al.
Published: (2026)

The Wolf Within: Covert Injection of Malice into MLLM Societies via an MLLM Operative
by: Tan, Zhen, et al.
Published: (2024)