1
GPT-Red: Unlocking Self-Improvement for Robustness
OpenAI用自我对弈实现自动化红队测试，大幅提升AI安全与提示注入防御能力。
Explore GPT-Red, OpenAI’s automated red teaming system that uses self-play to improve AI safety, alignment, and prompt injection robustness.
2
LLM-PDESR: Robust PDE Discovery via Subdomain Weighted Residuals and LLM-Guided Symbolic Hypothesis Generation
AI引导符号假设生成，子域加权残差法让噪声数据中偏微分方程发现更鲁棒精准
arXiv:2607.10546v1 Announce Type: new Abstract: Discovering governing partial differential equations (PDEs) from noisy observational data is a fundame…
3
YUKTI: From Natural-Language Situations to Robust, Verifiable Decisions An Uncertainty-Typed Proposition IR, Assumption-Robust Pareto Frontiers, and a Regret Certificate
语言模型做决策不可靠？YUKTI提出从自然语言到鲁棒可验证决策的新框架，挑战传统单目标优化的置信度陷阱。
arXiv:2607.09706v1 Announce Type: new Abstract: Language models turn a worded situation into a numeric plan, and the dominant pipelines (NL4Opt, OptiM…
4
Can We Trust LLM's Logic? Quantifying Uncertainty, Coherence, and Robustness via a Graph-Based Framework
用图论框架量化大模型逻辑推理的可信度，为评估LLM可靠性提供新方法。
arXiv:2607.08017v1 Announce Type: new Abstract: Large-Language Models (LLMs) can be prone to flawed and unfaithful reasoning that decoding strategies …
5
PeTeR: Post-Training Robustification of Probabilistic Circuits
概率电路后训练鲁棒化新方法PeTeR，增强模型可靠性与鲁棒性，性能更优。
arXiv:2607.07671v1 Announce Type: new Abstract: Probabilistic circuits (PCs) can model complex joint distributions while supporting exact and efficien…
6
Programming over Thinking: Efficient and Robust Multi-Constraint Planning
提出一种"在思考之上编程"的新范式，高效稳健解决多约束规划难题，AI规划领域重要突破。
arXiv:2601.09097v3 Announce Type: replace Abstract: Multi-constraint planning involves identifying, evaluating, and refining candidate plans while sat…
7
Truth or Sophistry? LoFa: A Benchmark for LLM Robustness Against Logical Fallacies
首个评估大模型逻辑谬误鲁棒性的基准，揭示LLM在诡辩面前的漏洞，被ACL 2026收录。
arXiv:2606.31039v1 Announce Type: new Abstract: Large Language Models (LLMs) exhibit strong semantic capabilities, yet their resilience to manipulativ…
8
RoPoLL: Robust Panel of LLM Judges
新方法RoPoLL让LLM评判者更鲁棒，18倍参数优势却更精准，有效对抗偏见污染。
arXiv:2606.30931v1 Announce Type: new Abstract: The LLM Jury, a Panel of LLM Evaluators (PoLL) reporting consensus scores, has become a practical alte…
9
AURORA: Asymmetry and Update-Induced Rotation for Robust Hallucination Detection in Large Language Models
基于不对称性与更新诱导旋转的创新方法，有效提升大语言模型幻觉检测的鲁棒性。
arXiv:2606.29545v1 Announce Type: new Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities across a wide range of natural …
10
Low-Agreeableness Persona Conditioning for Safe LLM Fine-Tuning
提出用低宜人性人格条件化提升LLM微调安全性，平衡对齐与攻击鲁棒性。
arXiv:2606.27709v1 Announce Type: cross Abstract: Recent work has shown that fine-tuning large language models (LLMs) for social warmth degrades factu…
11
Towards Reliable Recommender Systems for Rating Data
探索如何提升评分数据推荐系统的可靠性，提出基于数据特性的新方法
arXiv:2412.20802v3 Announce Type: replace-cross Abstract: Recommender systems are widely used in the digital landscape to match users with content fit…
12
Equivariance and Augmentation for Bayesian Neural Networks
贝叶斯神经网络与等变性结合，数据增强理论新突破，提升模型鲁棒性
arXiv:2606.26273v1 Announce Type: new Abstract: Symmetries are important for many deep learning tasks, ranging from applications in the sciences to me…
13
GEOALIGN: Geometric Rollout Curation for Robust LLM Reinforcement Learning
ICML 2026录用论文，针对LLM强化学习中的rollout采样提出几何优化方法，提升模型鲁棒性。
arXiv:2606.26917v1 Announce Type: cross Abstract: Online reinforcement learning is widely used to align large language models (LLMs) with reward signa…
14
Cliff Tokens: Identifying Single-Token Failure Triggers in LLM Mathematical Reasoning
本文发现LLM推理中的“悬崖词”——单个token即可导致数学运算失败，揭示模型脆弱性根源。
arXiv:2606.25524v1 Announce Type: new Abstract: Large language models (LLMs) reach high accuracy in mathematical reasoning, but individual traces on t…
15
Show HN: MAVS-GC – An Open-Source Governance Architecture for AI Systems
开源AI治理架构MAVS-GC，通过分层评估专家系统，增强算法在不利场景下的稳定性与适应性。
Hey HN, For some period of the time, I have been working on an open source project called MAVS-GC (Multi Adaptive Vetting Systems-Governance Core). Th…
16
CORE-BREW: LLR-Based Soft Decoding for Robust Multi-Bit LLM Watermarking
突破性LLM水印方案CORE-BREW采用软解码，提升多比特水印的鲁棒性与误报控制
arXiv:2606.24163v1 Announce Type: cross Abstract: Reliable provenance for LLM outputs requires multi-bit watermarks that remain robust under editing w…
17
Tuning without Peeking: Provable Generalization Bounds and Robust LLM Post-Training
提出“无窥视调优”方法，为大模型后训练提供可证明的泛化界限与鲁棒性保障。
arXiv:2507.01752v4 Announce Type: replace-cross Abstract: Gradient-based optimization is the workhorse of deep learning, offering efficient and scalab…
18
Paraphrasing Attack Resilience of Various AI-Generated Text Detection Methods
探究多种AI生成文本检测方法在面对改写攻击时的鲁棒性，为内容安全提供新视角。
arXiv:2605.14240v1 Announce Type: cross Abstract: The recent large-scale emergence of LLMs has left an open space for dealing with their consequences,…
19
NIM4-ASR: Towards Efficient, Robust, and Customizable Real-Time LLM-Based ASR
提出NIM4-ASR架构，实现高效、鲁棒且可定制的基于LLM的实时语音识别。
arXiv:2604.18105v2 Announce Type: replace-cross Abstract: Integrating large language models (LLMs) into automatic speech recognition (ASR) has become …
20
How Fragile Are Training-Free AI-Generated Image Detectors? A Controlled Audit of Score Direction, Preprocessing, and Compression
揭秘无训练AI图像检测器的脆弱性，系统评估分数方向、预处理与压缩的三重影响。
arXiv:2606.20488v1 Announce Type: new Abstract: Training-free detectors of AI-generated images promise generator-agnostic deployment without classifie…