No article text was recorded for this version.
alignment tax (safety-utility trade-off)
Version 1 of 1852 papers1,060 conceptsfrozen 2026-10-04
Papers in this version
- Blog: Automatic Pseudo-Harmful Prompt Generation for Evaluating False Refusals in Large Language Models
- Blog: The Alignment Problem in Context
- Blog: A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Models
- Blog: Alignment with Preference Optimization Is All You Need for LLM Safety
- Blog: Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems
- Blog: SafeCFG: Redirecting Harmful Classifier-Free Guidance for Safe Generation
- Blog: Safety Subspaces are Not Distinct: A Fine-Tuning Case Study
- Blog: FairMT-Bench: Benchmarking Fairness for Multi-turn Dialogue in Conversational LLMs
- Blog: Inference-Time Policy Steering through Human Interactions
- Blog: PEFT-as-an-Attack! Jailbreaking Language Models during Federated Parameter-Efficient Fine-Tuning
- Blog: Safety in Large Reasoning Models: A Survey
- Blog: VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration
- Blog: Why So Toxic?: Measuring and Triggering Toxic Behavior in Open-Domain Chatbots
- Blog: Feature-Guided SAE Steering for Refusal-Rate Control using Contrasting Prompts
- Blog: Why So Toxic?: Measuring and Triggering Toxic Behavior in Open-Domain Chatbots
- Blog: Tradeoffs Between Alignment and Helpfulness in Language Models
- Blog: AdvisorQA: Towards Helpful and Harmless Advice-seeking Question Answering with Collective Intelligence
- Blog: Jailbroken: How Does LLM Safety Training Fail?
- Blog: On Evaluating the Durability of Safeguards for Open-Weight LLMs
- Blog: Holistic Evaluation of Text-to-Image Models
- Blog: Safety Tax: Safety Alignment Makes Your Large Reasoning Models Less Reasonable
- Blog: Multilingual Jailbreak Challenges in Large Language Models
- Blog: AI Alignment Strategies from a Risk Perspective: Independent Safety Mechanisms or Shared Failures?
- Blog: Jailbroken: How Does LLM Safety Training Fail?
- Blog: Interpreting and Steering LLMs with Mutual Information-based Explanations on Sparse Autoencoders
- Blog: A StrongREJECT for Empty Jailbreaks
- Blog: On the Essence and Prospect: An Investigation of Alignment Approaches for Big Models
- Blog: Towards Tool Use Alignment of Large Language Models
- Blog: Flexible Realignment of Language Models
- Blog: You Can't Eat Your Cake and Have It Too: The Performance Degradation of LLMs with Jailbreak Defense
- Blog: Enhancing Model Defense Against Jailbreaks with Proactive Safety Reasoning
- Blog: TriPlay-RL: Tri-Role Self-Play Reinforcement Learning for LLM Safety Alignment
- Blog: "To Survive, I Must Defect": Jailbreaking LLMs via the Game-Theory Scenarios
- Blog: VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration
- Blog: On the Essence and Prospect: An Investigation of Alignment Approaches for Big Models
- Blog: A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy
- Blog: A survey of safety and trustworthiness of large language models through the lens of verification and validation
- Blog: Capabilities of GPT-4 on Medical Challenge Problems
- Blog: Can LLMs Follow Simple Rules?
- Blog: You Can't Eat Your Cake and Have It Too: The Performance Degradation of LLMs with Jailbreak Defense
- Blog: Large Language Model Instruction Following: A Survey of Progresses and Challenges
- Blog: Improving Consistency Identification in Task-oriented Dialogue Through Multi-Agent Collaboration
- Blog: Safe RLHF: Safe Reinforcement Learning from Human Feedback
- Blog: ToolSword: Unveiling Safety Issues of Large Language Models in Tool Learning Across Three Stages
- Blog: Llama 2: Open Foundation and Fine-Tuned Chat Models
- Blog: PRISM: Robust VLM Alignment with Principled Reasoning for Integrated Safety in Multimodality
- Blog: DEAN: Deactivating the Coupled Neurons to Mitigate Fairness-Privacy Conflicts in Large Language Models
- Blog: LLMs for Generating and Evaluating Counterfactuals: A Comprehensive Study
- Blog: OffTopicEval: When Large Language Models Enter the Wrong Chat, Almost Always!
- Blog: BadRobot: Jailbreaking LLM-based Embodied AI in the Physical World
- Blog: Disentangled Safety Adapters Enable Efficient Guardrails and Flexible Inference-Time Alignment
- Blog: PoseGuard: Pose-Guided Generation with Safety Guardrails
- Blog: Stealthy and Persistent Unalignment on Large Language Models via Backdoor Injections
- Blog: Towards Comprehensive and Efficient Post Safety Alignment of Large Language Models via Safety Patching
- Blog: MidPO: Dual Preference Optimization for Safety and Helpfulness in Large Language Models via a Mixture of Experts Framework
- Blog: Evaluating Defences against Unsafe Feedback in RLHF
- Blog: Capability-Based Scaling Trends for LLM-Based Red-Teaming
- Blog: SPA: Achieving Consensus in LLM Alignment via Self-Priority Optimization
- Blog: Self-Guard: Defending Large Reasoning Models via enhanced self-reflection
- Blog: DREAM: Disentangling Risks to Enhance Safety Alignment in Multimodal Large Language Models
Showing 60 of 852 papers.
History & citation
Current · Permanent link
- Mitigating the Alignment Tax in Large Language Models merged into this direction (71% overlap)
- Parameter Space Surgery for LLM Safety Preservation merged into this direction (50% overlap)
Cite this direction
Latest version (a living page):
@misc{lacuna_alignment_tax_safety_utility_trade_off_dde39f15,
title = {alignment tax (safety-utility trade-off)},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/alignment-tax-safety-utility-trade-off/rdr_dde39f1517c94a16a4a06d1857f3d487},
note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}This exact version (version 1, frozen text):
@misc{lacuna_alignment_tax_safety_utility_trade_off_dde39f15_v1,
title = {alignment tax (safety-utility trade-off)},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/alignment-tax-safety-utility-trade-off/rdr_dde39f1517c94a16a4a06d1857f3d487/v/1},
note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}