1
国产“原子级放大镜”问世：中国科大利用拓扑探针首次实现交错磁体原子尺度自旋成像
IT之家 10 月 14 日消息，中国科学技术大学科研团队将拓扑材料制成扫描隧道显微镜针尖，发展出原子尺度自旋探测新方法，并获得钾钒硒氧表面存在二维 d 波交错磁序的直接实验证据。这一针尖在未来有望成为观察电子自旋的“原子级放大镜”。 该研究由中国科学技术大学物理学院陈仙辉院士、王震宇教授团队联合该…
2
Memorization Diagnostics for Code LLMs Should be Scale-Aware
代码大模型靠记忆还是真理解？这项研究揭示传统记忆化探针在大规模模型上严重失效，诊断方法必须随规模调整。
arXiv:2608.12771v1 Announce Type: cross Abstract: The extent to which large language models for code rely on memorization over genuine understanding r…
3
Cheap Probes Predict Expensive Training in 3D-CT Vision--Language Models
用低成本探针方法预测昂贵训练效果，快速筛选3D CT视觉语言模型的最佳编码器和压缩方案。
arXiv:2607.22771v1 Announce Type: new Abstract: Picking the frozen image encoder for a 3D~CT vision--language model (VLM), together with the token-com…
4
Before the Last Token: Diagnosing Final-Token Safety Probe Failures
揭秘大模型安全探针为何在最终token失效：早于最后一层的内部表征已暴露漏洞
arXiv:2605.12726v2 Announce Type: replace Abstract: Final-token safety probes monitor a single hidden state after prompt prefill, but jailbreak prompt…
5
Show HN: Xcover, test coverage without instrumentation, using eBPF
基于eBPF内核探针，无需插桩即可测量二进制测试覆盖率，还能灵活排除函数，堪称测试利器。
Article URL: https://github.com/maxgio92/xcover Comments URL: https://news.ycombinator.com/item?id=48831339 Points: 5 # Comments: 0
6
Probing Memorization of Tabular In-Context Learning
新研究揭示表格型上下文学习的记忆机制，为理解大模型基础行为提供新视角
arXiv:2606.31208v1 Announce Type: new Abstract: Large tabular models (LTMs), i.e., tabular foundation models leveraging in-context learning (ICL), ach…
7
From Signals to Transfer: A Factorised Study of Probe-Based Uncertainty Estimation in Large Language Models
探针法如何量化LLM的不确定性？这篇研究从信号到迁移层层拆解，为模型校准提供新的因子化视角。
arXiv:2606.27679v1 Announce Type: cross Abstract: Probe-based uncertainty estimation (UE) has emerged as a prominent approach to detect hallucinations…
8
Stop Early, Spend Less: Hidden-State Probes as a Practical Recipe for Streaming Moderation of LLM Outputs
用隐藏状态探针在流式生成中提前审核LLM输出，中断不当内容并节省算力
arXiv:2606.10487v1 Announce Type: cross Abstract: Deploying large language models in user-facing systems requires efficient output safety filtering. E…
9
X-RAY: Mapping LLM Reasoning Capability via Formalized and Calibrated Probes
一项被KDD 2026接收的新方法，通过形式化与校准探针精准映射大模型推理能力，揭示LLM深层认知图谱。
arXiv:2603.05290v2 Announce Type: replace Abstract: Large language models (LLMs) achieve promising performance, yet their ability to reason remains po…
10
Probe Before You Edit: Probing-Guided Molecular Optimization for LLM Agents in Structure-Based Drug Design
提出Probe-guided策略优化LLM代理在药物分子设计中的编辑行为，提升基于结构的分子优化效果。
arXiv:2606.00555v1 Announce Type: new Abstract: Structure-based drug design increasingly employs LLM agents to iteratively refine ligands against a ta…
11
SPM-Bench: Benchmarking Large Language Models for Scanning Probe Microscopy
原创博士级科学基准，专测LLM在扫描探针显微镜领域的专业能力。
arXiv:2602.22971v2 Announce Type: replace Abstract: As LLMs achieved breakthroughs in general reasoning, their proficiency in specialized scientific d…
12
BAT: Better Audio Transformer Guided by Convex Gated Probing
BAT用凸门控探针提升音频Transformer效率，ICML26最新研究，让音频处理更精准高效。
arXiv:2602.16305v2 Announce Type: replace-cross Abstract: Probing is widely adopted in computer vision to faithfully evaluate self-supervised learning…
13
openclaw 2026.5.31-beta.3
openclaw 是一款开源 CI/CD 工具链，提供日志限制、响应体监控、就绪探针和制品检查，助力高效 DevOps 流程。
2026.5.31 Highlights Agents and CLI-backed runtimes recover more cleanly from interrupted tool calls, stale session bindings, compaction handoffs, and…
14
Probing for Knowledge Attribution in Large Language Models
精准揭示LLM幻觉根源，区分两种违规类型，为模型知识归因与纠错提供新思路。
arXiv:2602.22787v2 Announce Type: replace-cross Abstract: Large language model (LLM) hallucinations, meaning fluent but factually incorrect generation…
15
Pressure-Testing Deception Probes in LLMs: Scaling, Robustness, and the Geometry of Deceptive Representations
探索LLM欺骗探测线性探针的扩展性、鲁棒性与欺骗表示几何特性，为AI安全提供新视角。
arXiv:2605.27958v1 Announce Type: cross Abstract: Linear probes trained on LLM activations are increasingly proposed as deception-detection metrics, y…
16
Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations
用线性探针在LLM各层中间激活中提前解码拒绝行为，揭示安全机制的可预测性
arXiv:2605.28553v1 Announce Type: new Abstract: In this paper, we investigate whether refusal behavior can be predicted from LLM intermediate activati…
17
Causal Tongue-Tie: LLMs Can Encode Causal Direction, But Their Yes/No Outputs Fail to Express
大模型能编码因果方向，但输出却返回反常识答案，揭示隐藏状态与表达的不匹配。
arXiv:2605.25891v1 Announce Type: cross Abstract: We find a mismatch between what large language models encode about a causal question and what they a…
18
A Multi-Probe Audit of Clinical-Interview Depression Detection Benchmarks
四种互补探针审计临床访谈抑郁检测基准，揭示DAIC/E-DAIC和CMD等评估的潜在缺陷，推动更可靠的AI心理健康诊断。
arXiv:2605.23977v1 Announce Type: new Abstract: This paper audits benchmark evaluation in clinical-interview depression detection through four complem…
19
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…
20
Position: Let's Develop Data Probes to Fundamentally Understand How Data Affects LLM Performance
提出“数据探针”新范式，从因果层面揭示数据如何影响大模型性能，AI研究必读。
arXiv:2605.18801v1 Announce Type: cross Abstract: Data is fundamental to large language models (LLMs). However, understanding of what makes certain da…