No article text was recorded for this version.
alignment-robustness gap
Version 1 of 148 papers54 conceptsfrozen 2026-10-04
Papers in this version
- Blog: Vulnerabilities of Large Language Models to Adversarial Attacks
- Blog: Are aligned neural networks adversarially aligned?
- Blog: Guardrails and Security for LLMs: Safe, Secure and Controllable Steering of LLM Applications
- Blog: Visual Adversarial Examples Jailbreak Aligned Large Language Models
- Blog: GuardSet-X: Massive Multi-Domain Safety Policy-Grounded Guardrail Dataset
- Blog: Jailbreaking as a Reward Misspecification Problem
- Blog: Beyond the Tip of Efficiency: Uncovering the Submerged Threats of Jailbreak Attacks in Small Language Models
- Blog: Evaluating Implicit Bias in Large Language Models by Attacking From a Psychometric Perspective
- Blog: Exploiting the Index Gradients for Optimization-Based Jailbreaking on Large Language Models
- Blog: The Hidden Risks of Large Reasoning Models: A Safety Assessment of R1
- Blog: LLM Safety Alignment is Divergence Estimation in Disguise
- Blog: Stealthy Jailbreak Attacks on Large Language Models via Benign Data Mirroring
- Blog: Finding and Fixing Undesirable Behaviors in Pretrained Language Models
- Blog: Can Rationalization Improve Robustness?
- Blog: Pressure Reveals Character: Behavioural Alignment Evaluation at Depth
- Blog: On Guardrail Models’ Robustness to Mutations and Adversarial Attacks
- Blog: CrossAnchor: Image-Anchored Text Optimization Exposes Blind Spots in Multi-line Defenses of Agentic Systems
- Blog: Who Flips? Self- and Cross-Model Counterarguments Reveal Answer Instability in LLMs
- Blog: Jailbroken: How Does LLM Safety Training Fail?
- Blog: Fake Alignment: Are LLMs Really Aligned Well?
- Blog: Exploiting Task-Level Vulnerabilities: An Automatic Jailbreak Attack and Defense Benchmarking for LLMs
- Blog: Steering Without Side Effects: Improving Post-Deployment Control of Language Models
- Blog: Playing the Fool: Jailbreaking LLMs and Multimodal LLMs with Out-of-Distribution Strategy
- Blog: Distributional Preference Alignment of LLMs via Optimal Transport
- Blog: Do Perceptually Aligned Gradients Imply Robustness?
- Blog: Hair-Trigger Alignment: Black-Box Evaluation Cannot Guarantee Post-Update Alignment
- Blog: Preference Poisoning Attacks on Reward Model Learning
- Blog: Learn Your Reference Model for Real Good Alignment
- Blog: Activation Steering for Tool-Poisoning Defense in Language-Model Agents
- Blog: Universal Jailbreak Backdoors from Poisoned Human Feedback
- Blog: Prompt Leakage effect and mitigation strategies for multi-turn LLM Applications
- Blog: SafeMERGE: Preserving Safety Alignment in Fine-Tuned Large Language Models via Selective Layer-Wise Model Merging
- Blog: Chain-of-Sanitized-Thoughts: Reducing PII Leakage in Chain-of-Thought Reasoning
- Blog: SALAD-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Models
- Blog: Divide-Then-Align: Honest Alignment based on the Knowledge Boundary of RAG
- Blog: GuardBench: A Large-Scale Benchmark for Guardrail Models
- Blog: Characterizing Large Language Model Geometry Helps Solve Toxicity Detection and Generation
- Blog: BEEAR: Embedding-based Adversarial Removal of Safety Backdoors in Instruction-tuned Language Models
- Blog: Immune: Improving Safety Against Jailbreaks in Multi-modal LLMs via Inference-Time Alignment
- Blog: Policy Teaching via Data Poisoning in Learning from Human Preferences
- Blog: PICACO: Pluralistic In-Context Value Alignment via Total Correlation Optimization
- Blog: Test-Time Immunization: A Universal Defense Framework Against Jailbreaks for (Multimodal) Large Language Models
- Blog: MLLMGuard: A Multi-dimensional Safety Evaluation Suite for Multimodal Large Language Models
- Blog: Training Socially Aligned Language Models on Simulated Social Interactions
- Blog: Towards Understanding Safety Alignment: A Mechanistic Perspective from Safety Neurons
- Blog: AdvWeb: Controllable Black-box Attacks on VLM-powered Web Agents
- Blog: Monte Carlo Tree Search Based Prompt Autogeneration for Jailbreak Attacks against LLMs
- Blog: Are Large Language Models Really Bias-Free? Jailbreak Prompts for Assessing Adversarial Robustness to Bias Elicitation
History & citation
Current · Permanent link
- This direction split from Adversarial Robustness and Knowledge Erasure in Aligned Language Models (50% overlap)
Cite this direction
Latest version (a living page):
@misc{lacuna_alignment_robustness_gap_97a6de02,
title = {alignment-robustness gap},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/alignment-robustness-gap/rdr_97a6de02b48c44459f98db23e05858e8},
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_robustness_gap_97a6de02_v1,
title = {alignment-robustness gap},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/alignment-robustness-gap/rdr_97a6de02b48c44459f98db23e05858e8/v/1},
note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}