Ronald Luc

It was wonderful to receive the award last week and present our team’s thoughts now, reflecting on the 10 years after the initial publication. Trace data never tells the whole story of the software development process and we run the risk of misinterpretation when we don’t know where the perils are in our dataset. Investigating dataset perils is a type of work that the mining community should standardize. Everytime there is a dataset that becomes central to the community’s analyses and papers, there should be a corresponding perils paper. The more popular the dataset, the more critical the role of the perils paper! I hope to see this trend continue, especially in the era of developers using AI tools, which I suspect is going to result in a popular dataset (if not more) soon. Until then, many thanks to the community for the recognition, and congratulations to all the authors!

63

2 komentáře

An interesting project "Paper without code": LLM reads an article, writes a detailed extract on it, draws pictures and graphs, and writes code to test ideas in the article (from scratch or based on the code in the article repository). Link to the example from the blog: https://lnkd.in/eNSjECx7 Project code: https://lnkd.in/ebGdpTtM

3

What about #Java & #AI? Lots of opinions about where it is today but where is it going? That's an awesome question to ask. Nicolai Parlog gave an impressive briefing in his YouTube Newscast earlier this month ... If you're curious, check out the Newscast. #itnews #technews #softwareengineering #ml #machinelearning #softwaredevelopment #itmanagers #devops #datascience https://lnkd.in/gfSrxre6

1

If you work in the AI field, you have probably already visited the pages of the main conferences to see the accepted papers. However, you may have had difficulty searching for papers in your area of interest using natural language. To address this issue for the upcoming CVPR 2024, I built a Gradio application on Hugging Face Spaces where I indexed all the abstracts of the accepted papers using sentence transformers and FAISS, enabling semantic search over the papers. Link to check the app: https://lnkd.in/d_9pd2_B

39

1 komentář

Finally managed to sort my thoughts about everything I saw at the Conference of Complex Systems 2024! Overall it was a great event, some great keynotes, and some great talks and lightnings. Key highlights for me: - Thomaso Aste gave a great presentation of his work on using network representations for data-driven modelling - some amazing applications in areas as diverse as neuro-biology and bond pricing! - Cheick Tidiane Ba showed a fascinating use of temporal network analysis using Raphtory to investigate the collapse of the Terra/Luna stable coin - Andrea Bamberg Migliano showed how network theory can be applied to the anthropological analysis of hunter gatherer tribes In terms of takeaways, I will definitely try to keep up with the literature on data-driven modelling and the use of graphs and networks theory in that space! I see some great applications in making high-dimensional data tractable and queryable. Some useful links: https://lnkd.in/eS25TVMz https://lnkd.in/e3RGdiqT https://lnkd.in/ewFjA_kd #CCS2024, #graphsandnetworks, #complexsystems

23

1 komentář

The paper is out too! https://lnkd.in/dTaY2NUi We go into significant detail about reproducing benchmark numbers for open and closed source models. We release a repository to reproduce these evals with standardized prompts and metrics: https://lnkd.in/d3zXiPYz We also release a multi-modal version of MT-Bench which uses LLM-as-a-judge to better evaluate models on their utility as a multimodal assistant: https://lnkd.in/dQ6v_y-b

21

4 komentáře