Image de couverture de livre

BARRED: Synthetic Training of Custom Policy Guardrails via Asymmetric Debate

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Mazza, Arnon, Levi, Elad
Format:	Preprint
Publié:	2026
Sujets:	Computation and Language Artificial Intelligence Machine Learning
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Lightweight Safety Guardrails via Synthetic Data and RL-guided Adversarial Training
par: Ilin, Aleksei, et autres
Publié: (2025)

IntellAgent: A Multi-Agent Framework for Evaluating Conversational AI Systems
par: Levi, Elad, et autres
Publié: (2025)

Building a Foundational Guardrail for General Agentic Systems via Synthetic Data
par: Huang, Yue, et autres
Publié: (2025)

Intent-based Prompt Calibration: Enhancing prompt optimization with synthetic boundary cases
par: Levi, Elad, et autres
Publié: (2024)

MAD-OPD: Breaking the Ceiling in On-Policy Distillation via Multi-Agent Debate
par: Wang, Jianze, et autres
Publié: (2026)

SafePred: A Predictive Guardrail for Computer-Using Agents via World Models
par: Chen, Yurun, et autres
Publié: (2026)

Why Do Safety Guardrails Degrade Across Languages?
par: Zhang, Max, et autres
Publié: (2026)

Noise Injection Systemically Degrades Large Language Model Safety Guardrails
par: Shahani, Prithviraj Singh, et autres
Publié: (2025)

Disentangled Safety Adapters Enable Efficient Guardrails and Flexible Inference-Time Alignment
par: Krishna, Kundan, et autres
Publié: (2025)

Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing
par: Neill, James O', et autres
Publié: (2025)

Pluralistic Behavior Suite: Stress-Testing Multi-Turn Adherence to Custom Behavioral Policies
par: Varshney, Prasoon, et autres
Publié: (2025)

Evaluating the Performance of Large Language Models via Debates
par: Moniri, Behrad, et autres
Publié: (2024)

CONSCENDI: A Contrastive and Scenario-Guided Distillation Approach to Guardrail Models for Virtual Assistants
par: Sun, Albert Yu, et autres
Publié: (2023)

LoRA-Guard: Parameter-Efficient Guardrail Adaptation for Content Moderation of Large Language Models
par: Elesedy, Hayder, et autres
Publié: (2024)

UniGuard: Towards Universal Safety Guardrails for Jailbreak Attacks on Multimodal Large Language Models
par: Oh, Sejoon, et autres
Publié: (2024)

Does Training on Synthetic Data Make Models Less Robust?
par: Zhang, Lingze, et autres
Publié: (2025)

HAPO: Training Language Models to Reason Concisely via History-Aware Policy Optimization
par: Huang, Chengyu, et autres
Publié: (2025)

Provable Length Generalization in Sequence Prediction via Spectral Filtering
par: Marsden, Annie, et autres
Publié: (2024)

Identifying Features Associated with Bias Against 93 Stigmatized Groups in Language Models and Guardrail Model Safety Mitigation
par: Gueorguieva, Anna-Maria, et autres
Publié: (2025)

Incremental Summarization for Customer Support via Progressive Note-Taking and Agent Feedback
par: Wu, Yisha, et autres
Publié: (2025)

Large Language Model Post-Training: A Unified View of Off-Policy and On-Policy Learning
par: Zhao, Shiwan, et autres
Publié: (2026)

How Bad is Training on Synthetic Data? A Statistical Analysis of Language Model Collapse
par: Seddik, Mohamed El Amine, et autres
Publié: (2024)

XL-Suite: Cross-Lingual Synthetic Training and Evaluation Data for Open-Ended Generation
par: Iyer, Vivek, et autres
Publié: (2025)

Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task
par: Li, Kenneth, et autres
Publié: (2022)

ThinkGuard: Deliberative Slow Thinking Leads to Cautious Guardrails
par: Wen, Xiaofei, et autres
Publié: (2025)

PD$^3$: A Project Duplication Detection Framework via Adapted Multi-Agent Debate
par: Bao, Dezheng, et autres
Publié: (2025)

Stabilizing Off-Policy Training for Long-Horizon LLM Agent via Turn-Level Importance Sampling and Clipping-Triggered Normalization
par: Li, Chenliang, et autres
Publié: (2025)

Agentic Policy Optimization via Instruction-Policy Co-Evolution
par: Zhou, Han, et autres
Publié: (2025)

Towards a Theoretical Understanding of Synthetic Data in LLM Post-Training: A Reverse-Bottleneck Perspective
par: Gan, Zeyu, et autres
Publié: (2024)

Can LLMs Speak For Diverse People? Tuning LLMs via Debate to Generate Controllable Controversial Statements
par: Li, Ming, et autres
Publié: (2024)

CALICO: Conversational Agent Localization via Synthetic Data Generation
par: Rosenbaum, Andy, et autres
Publié: (2024)

A Truly Joint Neural Architecture for Segmentation and Parsing
par: Levi, Danit Yshaayahu, et autres
Publié: (2024)

Debate Helps Weak Judges Reward Stronger Models
par: Elasky, Ethan, et autres
Publié: (2026)

RigorLLM: Resilient Guardrails for Large Language Models against Undesired Content
par: Yuan, Zhuowen, et autres
Publié: (2024)

SIPDO: Closed-Loop Prompt Optimization via Synthetic Data Feedback
par: Yu, Yaoning, et autres
Publié: (2025)

EVPO: Explained Variance Policy Optimization for Adaptive Critic Utilization in LLM Post-Training
par: Pan, Chengjun, et autres
Publié: (2026)

Predicting Customer Satisfaction by Replicating the Survey Response Distribution
par: Manderscheid, Etienne, et autres
Publié: (2024)

Let Models Speak Ciphers: Multiagent Debate through Embeddings
par: Pham, Chau, et autres
Publié: (2023)

DiffLM: Controllable Synthetic Data Generation via Diffusion Language Models
par: Zhou, Ying, et autres
Publié: (2024)

Virus: Harmful Fine-tuning Attack for Large Language Models Bypassing Guardrail Moderation
par: Huang, Tiansheng, et autres
Publié: (2025)