1
2
马斯克 SpaceXAI 发布旗下最强编码和知识处理模型 Grok 4.7，百万词元输入 2 美元起
IT之家 9 月 22 日消息，SpaceXAI 今日宣布推出新一代 Grok 4.7，主打编程和知识工作场景。 官方称其采用全新更大规模基础模型，并通过更长时间的强化学习训练提升复杂任务处理能力，是该公司目前最强的编码和知识处理模型 —— 速度是同类模型的两倍，价格却只有一半。 如前所述，Grok…
3
Palantir CEO 卡普：美国前沿 AI 实验室或在谋求“国有化”，以避开知识产权诉讼
IT之家 9 月 18 日消息，Palantir CEO 亚历克斯 · 卡普认为，Anthropic 呼吁政府加强 AI 监管，背后并不只是安全考量。 当地时间 17 日，卡普接受 CNBC 采访时指出，Anthropic 等前沿 AI 实验室 真正谋求的可能是国有化，从而获得政府保护 ，应对未来因…
4
国内首套拥有完全自主知识产权的 EMA 工业化装置成功建成
IT之家 8 月 26 日消息，中化国际昨日宣布，8 月 24 日，中化国际 EMA（乙烯-马来酸酐共聚物）项目装置全线贯通，首批合格产品顺利下线。这是 国内首套拥有完全自主知识产权的 EMA 工业化装置 ，打破该产品核心生产技术长期由海外企业主导的局面，成功填补国内空白。 IT之家注：EMA 是一…
5
ConvergeWriter: Data-Driven Bottom-Up Article Construction
告别大纲失控，用数据自底向上构建长文，让事实与逻辑同步生长
arXiv:2509.12811v2 Announce Type: replace Abstract: Large Language Models (LLMs) have shown remarkable prowess in text generation, yet producing long-…
6
Credal Large Language Models for Semantic Commitment under Uncertainty
集合LoRA适配器构建置信区间，让大模型区分“不知道”与“不确定”，告别过度自信的胡说八道。
arXiv:2608.23244v1 Announce Type: cross Abstract: Large language models (LLMs) often produce fluent but incorrect answers with unwarranted confidence.…
7
SAFE-G: Structure-aware Faithful Evidence-guided Generation for Knowledge-based Visual Question Answering
结构感知+证据引导，让知识型视觉问答的推理过程更可信、可验证。
arXiv:2608.21796v1 Announce Type: cross Abstract: Knowledge-based Visual Question Answering (KB-VQA) aims to answer queries that necessitate reasoning…
8
I built 270 AI agents in an Obsidian vault – zero API costs, all open source
在Obsidian中一键构建270个AI代理，零API成本全开源，让知识库秒变智能自动化工作流。
Article URL: https://gist.github.com/NenadConnor/b9bb71d56c9e70a9072c55e262885b8b Comments URL: https://news.ycombinator.com/item?id=49424678 Points: …
9
From Mastery Profile to Simulated Response: Stochastic Student Knowledge Graphs (SSKG) for Faithful LLM Student Simulation
用随机学生知识图谱让LLM忠实还原不同掌握水平的学生表现，摆脱内置能力干扰，教育模拟更可靠。
arXiv:2608.21668v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used to simulate students at different mastery levels. T…
10
What AstroPT knows about galaxies, and what that can teach us about LLMs
从星系到语言模型，AstroPT揭示了预训练模型如何习得结构知识，为LLM研究提供跨学科洞见。
arXiv:2608.22614v1 Announce Type: new Abstract: Interpretability research increasingly asks when concepts emerge during training and whether linear pr…
11
MACD: Multi-Agent Clinical Diagnosis with Self-Learned Knowledge for LLM
多智能体协作+自学习知识，让大模型在临床诊断中更可靠。
arXiv:2509.20067v5 Announce Type: replace Abstract: Large language models (LLMs) have shown promise in supporting medical diagnosis, with prompting-ba…
12
Toward Effective and Reliable LLM Agents via Dynamic Ontology
用动态本体给智能体装上知识骨架，可靠性与效率双提升！
arXiv:2608.22974v1 Announce Type: new Abstract: Large language model (LLM) agents rely heavily on knowledge encoded in model parameters or presented a…
13
Knowing but Not Saying: Preventing Factual Access Failures in LLM SFT via Recall-Anchored Distillation
大模型微调后“知道却说不出口”的难题，用召回锚定蒸馏可精准破解。
arXiv:2608.20794v1 Announce Type: new Abstract: Supervised fine-tuning (SFT) can degrade factual behavior outside the target domain. This degradation …
14
Enhancing LLMs in Predictive Political QA with Semi-Structured Data
半结构化数据让大模型预测政治走向更靠谱？看这项研究如何提升问答准确性。
arXiv:2608.21218v1 Announce Type: new Abstract: Predictive political question answering (QA), such as predicting how a political actor will vote, goes…
15
Xbox 技术总监：我们支持保护知识产权，正与 Take-Two 和 R 星密切合作
IT之家 8 月 23 日消息，《侠盗猎车手 6》开发商 Rockstar 和发行商 Take-Two Interactive 最近都遇上了麻烦：本周早些时候开始，大批游戏实机内容突遭泄露，不仅曝光完整的莱昂尼达地图，还出现了疑似已经接近最终版本的实机画面。 网名 Cyberleek 的泄密者又放出…
16
Show HN: Visual way to understand things in parallel
用画布把 AI 对话变成可分支的知识树，随时展开术语和疑问，所有相关线程一目了然，不用切换标签页。
Hey HN! Really excited to be showing what I think is a better way to understand complex topics with AI. You start a chat with your question, and whene…
17
Building an Offline AI Security Scanner That Doesn't Hallucinate Findings
离线AI安全扫描器如何用规则引擎和知识图谱约束本地大模型，杜绝幻觉误报
How AiSec Studio pipes every scan through parser → rule engine → knowledge graph before a local LLM ever sees it — and why that order matters. 60-seco…
18
Show HN: Learn Leap, an AI tutor that teaches from your own material
AI tutor 直接从你的资料出题，读书科研的好帮手，告别泛泛而谈的通用问答
I built Learn Leap because I found myself constantly asking ChatGPT questions while reading research papers. I wanted something that already understoo…
19
Show HN: Encrypt messages locally into self-contained links
本地加密生成自毁链接，信息只存于链接中，无需服务器留存，阅后即焚保护隐私。
All encrypting and signing uses post-quantum primitives. This is a static page; nothing is ever sent to the server. Comments URL: https://news.ycombin…
20
Show HN: okfctl
用 Go 写成的 okfctl 帮你检查 Markdown 知识库的健康度和链接图，一个静态二进制即可运行。
I built a CLI tool to manage Open Knowledge Format (OKF) bundles. Knowledge bases, second brains, and LLM-Wikis now have a toolchain to complement the…