1
LLM Sparsity Prior for Robust Feature Selection
利用大模型蕴含的稀疏性先验，为高维数据特征选择提供鲁棒策略，理论贡献显著。
arXiv:2605.23102v1 Announce Type: cross Abstract: Large language models (LLMs) offer a scalable mechanism to elicit domain-informed prior information …
2
On Robustness and Chain-of-Thought Consistency of RL-Finetuned VLMs
首份系统研究RL微调VLM的鲁棒性与思维链一致性，揭示模型脆弱性根源
arXiv:2602.12506v3 Announce Type: replace Abstract: Reinforcement learning (RL) finetuning has become a key technique for enhancing large language mod…
3
Revisiting Robustness for LLM Safety Alignment via Selective Geometry Control
通过选择性几何控制，提升大模型安全对齐的鲁棒性，为AI防御攻击提供新思路。
arXiv:2602.07340v2 Announce Type: replace Abstract: Safety alignment of large language models remains brittle under domain shift and noisy preference …
4
Reducing Political Manipulation with Consistency Training
用一致性训练算法对抗社交网络政治操纵，为AI安全提供新思路
arXiv:2605.22771v1 Announce Type: new Abstract: Large language models (LLMs) exhibit systematic political bias across a variety of sensitive contexts.…
5
Robust Reasoning Benchmark
AIME 2024数学题经13种文本扰动，测试大模型推理鲁棒性，揭示依赖格式的短板
arXiv:2604.08571v2 Announce Type: replace-cross Abstract: While Large Language Models (LLMs) achieve high performance on standard mathematical benchma…
6
Codec-Robust Attacks on Audio LLMs
针对音频大模型的攻击新范式，能在编解码预处理下依然生效，突破现有防御机制。
arXiv:2605.20519v1 Announce Type: cross Abstract: Prior attacks on Audio Large Language Models (Audio LLMs) demonstrated that carefully crafted wavefo…
7
Cross-lingual robustness of LLM-brain alignment and its computational roots
大模型与大脑在跨语言条件下的对齐鲁棒性，首次从计算根源上得到解释。
arXiv:2605.21049v1 Announce Type: new Abstract: Large language models (LLMs) reliably predict neural activity during language comprehension and transf…
8
Adaptive Probe-based Steering for Robust LLM Jailbreaking
针对LLM越狱新方法，采用自适应探针引导，克服了传统对比引导的偏差和手动调参局限，提升鲁棒性与有效性。
arXiv:2605.20286v1 Announce Type: cross Abstract: Recent work has demonstrated the potential of contrastive steering for jailbreaking Large Language M…
9
Semantic Smoothing via Novel View Synthesis for Robust SAR Image Classification
利用新视图合成实现语义平滑，显著提升SAR图像分类的鲁棒性，创新性融合3D视角与遥感识别。
arXiv:2605.16440v1 Announce Type: new Abstract: Deep neural networks are vulnerable to adversarial perturbations, limiting deployment in safety-critic…
10
Trustworthiness in Retrieval-Augmented Generation Systems: A Survey
综述RAG系统可信度挑战，涵盖事实性、鲁棒性与公平性等关键维度。
arXiv:2409.10102v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) has quickly grown into a pivotal paradigm in the develo…
11
LLM-Oriented Information Retrieval: A Denoising-First Perspective
面向大语言模型的信息检索新视角，首次提出以去噪为核心，提升检索质量与鲁棒性
arXiv:2605.00505v2 Announce Type: replace-cross Abstract: Modern information retrieval (IR) is no longer consumed primarily by humans but increasingly…
12
ARROW: Augmented Replay for RObust World models
提出ARROW增强回放框架，显著提升世界模型在分布外场景的鲁棒性。
arXiv:2603.11395v2 Announce Type: replace Abstract: Continual reinforcement learning challenges agents to acquire new skills while retaining previousl…
13
How Sensitive Are Radiomic AI Models to Acquisition Parameters?
揭示放射组学AI模型在多中心采集协议下的性能敏感性，提出量化框架以提升临床鲁棒性
arXiv:2605.14667v1 Announce Type: new Abstract: A main barrier for the deployment of AI radiomic systems in clinical routine is their drop in performa…
14
Towards Camera-Robust 3D Localization: Equation-Anchored Tool-Use for MLLMs
用方程锚定工具使用，让多模态大模型实现相机鲁棒的3D定位，突破视觉视角限制。
arXiv:2605.19528v1 Announce Type: new Abstract: 3D localization in Multimodal Large Language Models (MLLMs), including 3D object detection and 3D visu…
15
Prune, Update and Trim: Robust Structured Pruning for Large Language Models
提出新型结构化剪枝方法，实现大模型高效压缩同时保持鲁棒性，适合模型优化研究者
arXiv:2605.18331v1 Announce Type: new Abstract: Large Language Models (LLMs) have experienced significant growth and development in recent years. Howe…
16
A Causally Grounded Taxonomy for Image Degradation Robustness Evaluation
从因果视角构建图像退化鲁棒性评估分类体系，为理解视觉模型退化失效提供新框架
arXiv:2605.15906v1 Announce Type: new Abstract: Image degradations can occur during acquisition, processing, and transmission, altering visual appeara…
17
LPDS: Evaluating LLM Robustness Through Logic-Preserving Difficulty Scaling
新方法LPDS通过保留逻辑改变实体，精准测试大模型鲁棒性，避免模型因细节变化而翻车。
arXiv:2605.15393v1 Announce Type: new Abstract: As large language models (LLMs) are increasingly deployed to perform tasks with minimal human oversigh…
18
ODRPO: Ordinal Decompositions of Discrete Rewards for Robust Policy Optimization
提出序数分解离散奖励的ODRPO方法，提升LLM对齐中策略优化的鲁棒性，直面自动评分器的随机挑战。
arXiv:2605.12667v2 Announce Type: replace-cross Abstract: The alignment of Large Language Models (LLMs) utilizes Reinforcement Learning from AI Feedba…
19
Reasoners or Translators? Contamination-aware Evaluation and Neuro-Symbolic Robustness in Tax Law
最新研究：LLM在税法推理中存在数据污染风险，别被“假懂”骗了！
arXiv:2605.16052v1 Announce Type: new Abstract: Recent advances in large language models (LLMs) have significantly enhanced automated legal reasoning.…
20
When Answers Stray from Questions: Hallucination Detection via Question-Answer Orthogonal Decomposition
新方法通过正交分解问答表征，实现高效幻觉检测，兼顾准确率与分布迁移鲁棒性。
arXiv:2605.14449v1 Announce Type: cross Abstract: Hallucination detection in large language models (LLMs) requires balancing accu racy, efficiency, an…