Cover Image

Rethinking the Evaluation of Alignment Methods: Insights into Diversity, Generalisation, and Safety

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Janiak, Denis, Moska, Julia, Motyka, Dawid, Seweryn, Karolina, Walkowiak, Paweł, Żuk, Bartosz, Janz, Arkadiusz
Format:	Preprint
Published:	2025
Subjects:	Machine Learning Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Unpacking Robustness in Inflectional Languages: Adversarial Evaluation and Mechanistic Insights
by: Walkowiak, Paweł, et al.
Published: (2025)

Hallucination Detection in LLMs Using Spectral Features of Attention Maps
by: Binkowski, Jakub, et al.
Published: (2025)

FactSelfCheck: Fact-Level Black-Box Hallucination Detection for LLMs
by: Sawczyn, Albert, et al.
Published: (2025)

Understanding the Effects of RLHF on LLM Generalisation and Diversity
by: Kirk, Robert, et al.
Published: (2023)

Do Generalisation Results Generalise?
by: Boglioni, Matteo, et al.
Published: (2025)

Safe, or Simply Incapable? Rethinking Safety Evaluation for Phone-Use Agents
by: Tang, Zhengyang, et al.
Published: (2026)

A Symbolic Framework for Evaluating Mathematical Reasoning and Generalisation with Transformers
by: Meadows, Jordan, et al.
Published: (2023)

The Illusion of Progress: Re-evaluating Hallucination Detection in LLMs
by: Janiak, Denis, et al.
Published: (2025)

Isotonic Quantile Regression Averaging for uncertainty quantification of electricity price forecasts
by: Lipiecki, Arkadiusz, et al.
Published: (2025)

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

StylOch at PAN: Gradient-Boosted Trees with Frequency-Based Stylometric Features
by: Ochab, Jeremi K., et al.
Published: (2025)

SafeArena: Evaluating the Safety of Autonomous Web Agents
by: Tur, Ada Defne, et al.
Published: (2025)

Sample, Don't Search: Rethinking Test-Time Alignment for Language Models
by: Faria, Gonçalo, et al.
Published: (2025)

Integrating gender inclusivity into large language models via instruction tuning
by: Wróblewska, Alina, et al.
Published: (2025)

Rethinking Regularization Methods for Knowledge Graph Completion
by: Li, Linyu, et al.
Published: (2025)

Safety Alignment as Continual Learning: Mitigating the Alignment Tax via Orthogonal Gradient Projection
by: Sun, Guanglong, et al.
Published: (2026)

Evaluating LLMs Robustness in Less Resourced Languages with Proxy Models
by: Chrabąszcz, Maciej, et al.
Published: (2025)

Advancing LLM Safe Alignment with Safety Representation Ranking
by: Du, Tianqi, et al.
Published: (2025)

Test-Time Safety Alignment
by: Saglam, Baturay, et al.
Published: (2026)

Rethinking the Role of Proxy Rewards in Language Model Alignment
by: Kim, Sungdong, et al.
Published: (2024)

Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning
by: Wang, Guoli, et al.
Published: (2026)

Towards Generalising Neural Topical Representations
by: Yang, Xiaohao, et al.
Published: (2023)

Survey of Action Recognition, Spotting and Spatio-Temporal Localization in Soccer -- Current Trends and Research Perspectives
by: Seweryn, Karolina, et al.
Published: (2023)

MULTIVERSE: Exposing Large Language Model Alignment Problems in Diverse Worlds
by: Jin, Xiaolong, et al.
Published: (2024)

Rethinking of Encoder-based Warm-start Methods in Hyperparameter Optimization
by: Płudowski, Dawid, et al.
Published: (2024)

SABER: Uncovering Vulnerabilities in Safety Alignment via Cross-Layer Residual Connection
by: Joshi, Maithili, et al.
Published: (2025)

Multilingual Safety Alignment via Self-Distillation
by: Qin, Ruiyang, et al.
Published: (2026)

PL-Guard: Benchmarking Language Model Safety for Polish
by: Krasnodębska, Aleksandra, et al.
Published: (2025)

Rethinking LLM Evaluation: Can We Evaluate LLMs with 200x Less Data?
by: Wang, Shaobo, et al.
Published: (2025)

Towards Understanding Safety Alignment: A Mechanistic Perspective from Safety Neurons
by: Chen, Jianhui, et al.
Published: (2024)

The Role of Language Imbalance in Cross-lingual Generalisation: Insights from Cloned Language Experiments
by: Schäfer, Anton, et al.
Published: (2024)

Activation Steering for Synthetic Data Generation: The Role of Diversity in Downstream Safety Detection
by: Deshpande, Vijeta, et al.
Published: (2026)

TRA: Better Length Generalisation with Threshold Relative Attention
by: Opper, Mattia, et al.
Published: (2025)

MobileSafetyBench: Evaluating Safety of Autonomous Agents in Mobile Device Control
by: Lee, Juyong, et al.
Published: (2024)

Rethinking Perplexity: Revealing the Impact of Input Length on Perplexity Evaluation in LLMs
by: Cheng, Letian, et al.
Published: (2026)

Safety Alignment Can Be Not Superficial With Explicit Safety Signals
by: Li, Jianwei, et al.
Published: (2025)

ProMIL: Probabilistic Multiple Instance Learning for Medical Imaging
by: Struski, Łukasz, et al.
Published: (2023)

Improving LLM Safety Alignment with Dual-Objective Optimization
by: Zhao, Xuandong, et al.
Published: (2025)

Course-Correction: Safety Alignment Using Synthetic Preferences
by: Xu, Rongwu, et al.
Published: (2024)

Rethinking Layer Relevance in Large Language Models Beyond Cosine Similarity
by: Hinostroza, Cristian, et al.
Published: (2026)