Cover Image

Aegis2.0: A Diverse AI Safety Dataset and Risks Taxonomy for Alignment of LLM Guardrails

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Ghosh, Shaona, Varshney, Prasoon, Sreedhar, Makesh Narsimhan, Padmakumar, Aishwarya, Rebedea, Traian, Varghese, Jibin Rajan, Parisien, Christopher
Format:	Preprint
Published:	2025
Subjects:	Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Safety Through Reasoning: An Empirical Study of Reasoning Guardrail Models
by: Sreedhar, Makesh Narsimhan, et al.
Published: (2025)

Pluralistic Behavior Suite: Stress-Testing Multi-Turn Adherence to Custom Behavioral Policies
by: Varshney, Prasoon, et al.
Published: (2025)

CantTalkAboutThis: Aligning Language Models to Stay on Topic in Dialogues
by: Sreedhar, Makesh Narsimhan, et al.
Published: (2024)

Unsupervised Extraction of Dialogue Policies from Conversations
by: Sreedhar, Makesh Narsimhan, et al.
Published: (2024)

Towards Inference-time Category-wise Safety Steering for Large Language Models
by: Bhattacharjee, Amrita, et al.
Published: (2024)

Training a General Purpose Automated Red Teaming Model
by: Padmakumar, Aishwarya, et al.
Published: (2026)

AEGIS: Online Adaptive AI Content Safety Moderation with Ensemble of LLM Experts
by: Ghosh, Shaona, et al.
Published: (2024)

Defenses & Enablers For Skill Injection Attacks on Terminal Based Agents
by: Fujinuma, Yoshinari, et al.
Published: (2026)

Improving Romanian LLM Pretraining Data using Diversity and Quality Filtering
by: Negoita, Vlad, et al.
Published: (2025)

SafeSteer: Interpretable Safety Steering with Refusal-Evasion in LLMs
by: Ghosh, Shaona, et al.
Published: (2025)

Semi-Supervised Learning for Large Language Models Safety and Content Moderation
by: Dinuta, Eduard Stefan, et al.
Published: (2025)

SmellBench: Evaluating LLM Agents on Architectural Code Smell Repair
by: Dinu, Ion George, et al.
Published: (2026)

Improving Legal Judgement Prediction in Romanian with Long Text Encoders
by: Masala, Mihai, et al.
Published: (2024)

Meta-learning how to Share Credit among Macro-Actions
by: Hosu, Ionel-Alexandru, et al.
Published: (2025)

Aegis: Taxonomy and Optimizations for Overcoming Agent-Environment Failures in LLM Agents
by: Song, Kevin, et al.
Published: (2025)

Complexity-based code embeddings
by: Folea, Rares, et al.
Published: (2026)

"Înţelegi Româneşte?'' A Recipe for Romanian Vision-Language Models
by: Masala, Mihai, et al.
Published: (2026)

Guardrails in Logit Space: Safety Token Regularization for LLM Alignment
by: Bach, Thong, et al.
Published: (2026)

GIT-CXR: End-to-End Transformer for Chest X-Ray Report Generation
by: Sîrbu, Iustin, et al.
Published: (2025)

HelpSteer2: Open-source dataset for training top-performing reward models
by: Wang, Zhilin, et al.
Published: (2024)

Addressing LLM Diversity by Infusing Random Concepts
by: Agrawal, Pulin, et al.
Published: (2026)

Safety Guardrails for LLM-Enabled Robots
by: Ravichandran, Zachary, et al.
Published: (2025)

Aegis:An Advanced LLM-Based Multi-Agent for Intelligent Functional Safety Engineering
by: Shi, Lu, et al.
Published: (2024)

Surfacing Semantic Orthogonality Across Model Safety Benchmarks: A Multi-Dimensional Analysis
by: Bennion, Jonathan, et al.
Published: (2025)

Why LLM Safety Guardrails Collapse After Fine-tuning: A Similarity Analysis Between Alignment and Fine-tuning Datasets
by: Hsiung, Lei, et al.
Published: (2025)

Does Writing with Language Models Reduce Content Diversity?
by: Padmakumar, Vishakh, et al.
Published: (2023)

A self-supervised neural-analytic method to predict the evolution of COVID-19 in Romania
by: Stochiţoiu, Radu D., et al.
Published: (2020)

VLN-Video: Utilizing Driving Videos for Outdoor Vision-and-Language Navigation
by: Li, Jialu, et al.
Published: (2024)

Current state of LLM Risks and AI Guardrails
by: Ayyamperumal, Suriya Ganesh, et al.
Published: (2024)

MultiMatch: Multihead Consistency Regularization Matching for Semi-Supervised Text Classification
by: Sirbu, Iustin, et al.
Published: (2025)

Containment Verification: AI Safety Guarantees Independent of Alignment
by: Moon, Royce, et al.
Published: (2026)

Beyond Single-Agent Safety: A Taxonomy of Risks in LLM-to-LLM Interactions
by: Bisconti, Piercosma, et al.
Published: (2025)

Evaluating the Diversity and Quality of LLM Generated Content
by: Shypula, Alexander, et al.
Published: (2025)

CR4T: Rewrite-Based Guardrails for Adolescent LLM Safety
by: An, Heajun, et al.
Published: (2026)

An Empirical Study of Aegis
by: Saragih, Daniel, et al.
Published: (2024)

Disentangled Safety Adapters Enable Efficient Guardrails and Flexible Inference-Time Alignment
by: Krishna, Kundan, et al.
Published: (2025)

AegisLLM: Scaling Agentic Systems for Self-Reflective Defense in LLM Security
by: Cai, Zikui, et al.
Published: (2025)

PSG-Agent: Personality-Aware Safety Guardrail for LLM-based Agents
by: Wu, Yaozu, et al.
Published: (2025)

SafeHarbor: Hierarchical Memory-Augmented Guardrail for LLM Agent Safety
by: Liu, Zhe, et al.
Published: (2026)

Bridging the Safety Gap: A Guardrail Pipeline for Trustworthy LLM Inferences
by: Han, Shanshan, et al.
Published: (2025)