1
联通反诈大模型揪出涉诈 VOIP 黑盒，如有可疑人员不请自来“维修网络”要警惕
IT之家 9 月 6 日消息，中国联通官方昨日在公众号平台披露，中国联通抚顺市分公司近日依托自主研发的反诈大模型系统，联合当地公安机关，精准识别并捣毁被用于电信诈骗的非法 VOIP 通信设备，守护群众财产安全。 8 月 24 日上午 9 时，中国联通反诈大模型发出紧急预警，监测到某辖区存在疑似涉诈线…
2
Hierarchy-Aware Supervised Uncertainty Estimation for Black-box LLM Taxonomic Reasoning
黑盒LLM分类推理的不确定性估计，巧用层级感知监督提升准确性。
arXiv:2608.22839v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used for scientific decision support, yet reliable con…
3
Improved Confidence Estimates for Black-Box Large Language Models
黑盒大模型置信度不准？这项研究提出新方法，让答案可信度更可靠。
arXiv:2608.19323v1 Announce Type: new Abstract: Uncertainty quantification (UQ) is essential for the safe deployment of large language models (LLMs). …
4
Fair ASR: Re-Evaluating Black-Box Jailbreaks under Shared Target-Call Budgets
挑战黑盒越狱评估的公平性，引入共享调用预算新框架，重新定义攻击成功率，值得AI安全研究者细读。
arXiv:2608.17360v1 Announce Type: cross Abstract: Reliable jailbreak evaluation is essential for assessing LLM safety, but most existing studies rely …
5
Targeted Counterfactual Fingerprinting for Black-Box LLM Ownership Verification
黑盒大模型也能验明正身？反事实指纹技术为LLM所有权验证提供新思路，防窃取。
arXiv:2608.08195v1 Announce Type: cross Abstract: Large language models (LLMs) are high-value assets that can be derived through redeployment, fine-tu…
6
Decomposed Entailment for Factuality Checking and Hallucination Detection
一句话看懂LLM幻觉检测新框架，轻量黑盒免参考，不止摘要还跨任务实测。
arXiv:2608.05823v1 Announce Type: new Abstract: The reliability of Large Language Models (LLMs) is often compromised by factual inconsistencies, inclu…
7
BBOWP-Bench: Evaluating LLMs on Black-Box Optimization Word Problems
专测大模型解决黑盒优化文字题的能力，基准设计值得关注
arXiv:2608.02612v1 Announce Type: cross Abstract: Formulating an optimization problem strongly affects the quality of the final solution, yet good for…
8
Stemma: Induced Decision Regions Reveal LLM Provenance
提出Stemma方法，利用诱导决策区域揭示大模型来源，提升黑盒测试的鲁棒性。
arXiv:2607.25880v1 Announce Type: cross Abstract: LLM provenance testing asks whether a suspect LLM belongs to the same lineage as a source. Existing …
9
Guardrails as Scapegoats: Auditing Unfaithful Safety Refusals in Tool-Augmented LLM Agents
重磅论文揭露：主流LLM Agent评估框架忽略静默故障，提出轻量黑盒审计方案，精准检测恶意拒绝漏洞
arXiv:2607.19449v1 Announce Type: new Abstract: Evaluation frameworks for tool-augmented LLM agents focus overwhelmingly on capability metrics or expl…
10
Interventional Grounding Audits: Black-Box Premise-Dependency Tests for LLM Chain-of-Thought via Predicate Substitution
黑盒审计方法通过前提替换干预，检测LLM思维链是否真正依赖前提，而非表面合理。
arXiv:2607.13069v1 Announce Type: new Abstract: Large language models produce chain-of-thought (CoT) reasoning that appears logically sound yet may no…
11
Auto-DSM Under the Lens: A Black-Box Evaluation Framework for LLM-Based DSM Generation
首个面向LLM生成DSM的黑盒评估框架，系统检验大模型在数字系统建模任务上的真实水准，值得关注。
arXiv:2607.05985v1 Announce Type: new Abstract: This paper presents a black-box evaluation framework to systematically assess the ability of Large Lan…
12
Trust-Region Noise Search for Black-Box Alignment of Diffusion and Flow Models
扩散与流模型的黑盒对齐新方法，信任区域噪声搜索带来高效微调
arXiv:2603.14504v2 Announce Type: replace-cross Abstract: Optimizing the noise samples of diffusion and flow models is an increasingly popular approac…
13
Black-Box Inference of LLM Architectural Properties with Restrictive API Access
通过限制性API访问，黑盒推断LLM架构属性的新方法，揭示模型内部结构。
arXiv:2607.01313v1 Announce Type: cross Abstract: In practice, most commercial LLM providers do not publicly release details of underlying LLM archite…
14
AlienLM: Alienization of Language for API-Boundary Privacy in Black-Box LLMs
用语言「陌生化」技术为大模型API边界隐私保驾护航，黑盒场景下实现隐私保护新思路
arXiv:2601.22710v2 Announce Type: replace-cross Abstract: Modern LLMs are increasingly accessed via black-box APIs, requiring users to transmit sensit…
15
AI, Trust, and Teaming: The Humans-as-Handlers Approach for Autonomous and Opaque AI Systems
当AI系统越来越黑盒，人类如何与它协作？这篇论文提出"人类即操控者"的全新思路，直击信任与自主性的核心矛盾。
arXiv:2607.00523v1 Announce Type: cross Abstract: Artificial intelligence (AI) is becoming ubiquitous, and across domains, increasingly autonomous sys…
16
Verify when Uncertain: Beyond Self-Consistency in Black Box Hallucination Detection
黑盒环境下用验证代替自一致性，大幅提升大模型幻觉检测准确率，值得关注。
arXiv:2502.15845v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) often hallucinate, limiting their reliability in sensitive appl…
17
CBD: API-Only LLM Black-Box Unlearning through Controlled Behavioral Divergence
仅通过API接口实现LLM黑盒遗忘，用行为分歧控制精准擦除特定知识，模型编辑与隐私保护新方向。
arXiv:2606.27683v1 Announce Type: cross Abstract: Edge devices increasingly invoke large language models (LLMs) through API services for context aware…
18
Stabilizing black-box algorithms through task-oriented randomization
通过任务导向的随机化方法，有效提升黑盒算法的稳定性，为机器学习优化提供新思路。
arXiv:2606.25269v1 Announce Type: cross Abstract: As black-box models become foundational to modern research, ensuring their stability is paramount fo…
19
LLM Program Optimization via Retrieval Augmented Search
用检索增强搜索做黑盒程序优化，beam search逐步迭代，LLM优化新思路。
arXiv:2501.18916v2 Announce Type: replace Abstract: Recent work has demonstrated the potential of large language models (LLMs) for program optimizatio…
20
When Compression Becomes an Attack Surface: Black-Box Attacks on Prompt-Compressed LLM Agents
提示压缩已成攻击面：黑盒手段可操纵压缩后文本，让LLM智能体偏离原意，安全研究必读。
arXiv:2510.22963v4 Announce Type: replace-cross Abstract: Prompt compression is increasingly deployed in LLM agents to reduce latency and cost, but it…