1
中信证券：港股扰动趋缓卖空料将收敛，短期推荐创新药、航空、机器人及工业金属
36氪获悉，中信证券最新研报指出，目前港股未平仓卖空占市值比例虽已从6月中旬的高点小幅回落至2.43%，但仍接近历史均值三倍标准差以上的水平，内外部扰动因素逐步缓和下，预计未来将有较大的回落空间。但近期港股也有较为明显的pair trading迹象，短期我们仍建议关注基本面确定性较高和有事件催化的方…
2
Do LLMs Truly Generalize in the Molecular Domain? A Perturbation-Based Analysis
扰动分析揭示：大模型在分子领域是否真正具备泛化能力？最新研究带来关键验证。
arXiv:2607.01800v1 Announce Type: new Abstract: Large Language Models (LLMs) have recently shown promise in molecular discovery, yet a gap remains bet…
3
中信证券：宏观扰动加剧，商品延续分化
36氪获悉，中信证券研报指出，2026年6月美伊开始和谈并签约，原油价格大幅回落，但冲突对通胀形成更长期的影响，宏观扰动加剧。美联储是否启动加息对黄金等贵金属产生显著影响。展望三季度，大宗品价格将延续分化。整体我们看好需求端支撑明显的铜、碳酸锂、电解铝、煤炭等。
4
Knowledge Graphs and Reasoning LLMs for Finding Simple Yet Effective Transcriptomic Perturbation Predictors
利用知识图谱增强LLM推理，预测基因扰动对转录组的影响，为虚拟细胞建模提供新思路。
arXiv:2606.08816v1 Announce Type: new Abstract: Predicting the effect of an unseen gene knockout perturbation on transcriptomic gene expression remain…
5
STRIDE: Training Data Attribution via Sparse Recovery from Subset Perturbations
用稀疏恢复技术从子集扰动中精确归因训练数据，破解大模型因果干预的计算难题
arXiv:2606.05165v1 Announce Type: cross Abstract: Training Data Attribution (TDA) seeks to trace a model's predictions back to its training data. The …
6
A Local Perturbation Theory for Cross-Domain Interference and Recovery in Multi-Domain RL
多领域强化学习中的跨域干扰与恢复难题，这篇论文提出全新局部扰动理论，为RL领域交叉应用提供新思路。
arXiv:2606.02398v1 Announce Type: new Abstract: Reinforcement learning (RL) post-training improves large language models (LLMs) on individual domains …
7
Plausibility Is Not Prediction: Contrastive Evidence for LLM-Based Cellular Perturbation Reasoning
直击LLM作为虚拟细胞模拟器的软肋：合理性不等于预测能力，新证据揭示模型推理的潜在缺陷。
arXiv:2606.01042v1 Announce Type: new Abstract: Perturbation experiments are central to understanding cellular mechanisms, but remain costly and spars…
8
LLM Anonymization Against Agentic Re-Identificatio
针对智能体LLM的文本匿名化新威胁模型与防御方法：弱上下文线索可被交叉引用导致重新识别，现有方法难以兼顾隐私与文本价值。
arXiv:2605.30848v1 Announce Type: cross Abstract: Agentic LLMs with web search change the threat model for text anonymization: weak contextual cues ca…
9
Consolidating Rewarded Perturbations for LLM Post-Training
把后训练循环从样本空间移到权重空间，用高斯扰动采样+奖励排序集成专家，开辟LLM微调新范式。
arXiv:2605.31494v1 Announce Type: cross Abstract: Post-training of language models is commonly framed as a sample-score-update loop implemented by gra…
10
Toxic HallucinAItions: Perturbing Prompts and Tracing LLM Circuits
揭示大语言模型产生毒性幻觉的内部机制，通过扰动提示词并追踪神经网络电路路径，为AI安全提供新思路。
arXiv:2605.30913v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly deployed in conversational settings where user tone ra…
11
Linear Ensembles Wash Away Watermarks: On the Fragility of Distributional Perturbations in LLMs
揭示LLM水印在多重模型访问下的致命缺陷：独立扰动轻松被线性集成抹除，对AI安全与版权保护提出新挑战。
arXiv:2605.30501v1 Announce Type: new Abstract: Watermarking embeds statistical signatures in AI-generated text for detection and attribution. We reve…
12
A Perturbation Approach to Unconstrained Linear Bandits
提出一种扰动方法，解决无约束线性赌博机的遗憾下界与最优算法设计问题，ICML 2026收录。
arXiv:2603.28201v2 Announce Type: replace Abstract: We revisit the standard perturbation-based approach of Abernethy et al. (2008) in the context of u…
13
Learning to Perturb Hidden Representations for Generalizable Deep Learning
一种通过扰动隐藏表示来增强深度学习模型泛化性的新方法，值得关注。
arXiv:2605.29525v1 Announce Type: new Abstract: Deep neural networks process data through a cascade of representations: input features, hidden activat…
14
QUIVER: A Formal Framework for Quantifying Perturbation Propagation and Bifurcation in Compound AI Systems
首次为复合AI系统提供量化扰动传播与分叉的正式数学框架，堪称AI可靠性研究的里程碑式工具。
arXiv:2605.23956v1 Announce Type: new Abstract: Compound AI systems that chain multiple LLM calls into directed computation graphs are now the dominan…
15
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…
16
Stop Drawing Scientific Claims from LLM Social Simulations Without Robustness Audits
微调提示词就能颠覆LLM社会模拟结论，稳健性审计刻不容缓。
arXiv:2605.18890v1 Announce Type: cross Abstract: The scientific claims drawn from LLM social simulations should be no stronger than the robustness au…
17
Adversarial Fragility and Language Vulnerability in Clinical AI: A Systematic Audit of Diagnostic Collapse Under Imperceptible Perturbations and Cross-Lingual Drift in Low-Resource Healthcare Settings
临床AI系统在细微扰动和多语言场景下存在诊断崩溃风险，这篇系统性审计揭开了安全漏洞。
arXiv:2605.16993v1 Announce Type: cross Abstract: Current clinical artificial intelligence (AI) systems are evaluated almost exclusively on clean, sta…
18
Adaptive Layerwise Perturbation: Unifying Off-Policy Corrections for LLM RL
统一离策略修正的自适应逐层扰动方法，为LLM强化学习提供更高效的训练策略。
arXiv:2603.19470v3 Announce Type: replace Abstract: Off-policy problems such as policy staleness and training--inference mismatch have become a major …