Tag
KITE 提出一种知识注入的时间序列编码器，将工程师的领域先验（语义先验）通过可切换的协变量与注意力偏置机制注入自监督事件预测器，在稀有事件预测上显著优于无先验基线，并支持跨域迁移。
AI tools amplify users' existing knowledge, benefiting experts significantly while providing novices with fluent output they cannot evaluate, potentially widening the expertise gap.
The article argues that domain expertise is the most important skill in using LLMs effectively, using Terence Tao's ChatGPT conversation as an example of expert-level prompting.
A thought-provoking thread arguing that automating engineering tasks, especially with AI agents like Claude, is more important than ever. It advocates encoding domain knowledge as infrastructure to accelerate development and enable contributions from non-engineers.
This paper introduces DKCD, a framework that enhances causal discovery from unstructured data in high-expertise domains by integrating domain knowledge graphs with LLM-based reasoning to identify latent causal factors and improve annotation accuracy.
The author shares a perspective that AI rewards those with deep domain knowledge, enabling them to apply expertise across multiple fields, and argues against the notion that AI will replace jobs.
This paper proposes a domain knowledge-based temporal-spatial graph convolution network for ECG recognition that uses PRQST landmarks and double-stream directed graphs to model intra- and inter-cycle dependencies, achieving state-of-the-art F1 scores on the First Chinese ECG Intelligent Competition dataset.
The article discusses that building LLM-driven AI tools still requires capturing domain knowledge, though it's easier than previous AI generations because knowledge doesn't need to be strictly structured.
The author replies to comments on their viral post about LLMs eroding their software engineering career, discussing how AI automation is reducing the need for deep domain knowledge in fintech and the challenges of maintaining diligence in a vibecoding culture.
A tweet comparing AI adoption to replacing employees with geniuses lacking domain knowledge, resulting in chaos.
BODHI is a domain knowledge prompting method that improves LLM-based generation of formal OS kernel specifications by augmenting few-shot prompts with a structured C-to-Python translation guide, achieving up to 96.73% Pass@1 on the OSV-Bench benchmark.
YC General Partner @t_blom presented a talk on building self-improving, AI-native companies, emphasizing recursive AI loops and reducing headcount through AI automation.
This paper introduces FINESSE-Bench, a suite of eight specialized benchmarks with 3,993 questions for hierarchical evaluation of financial competencies in large language models, covering professional certification topics and applied trading tasks.