Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
The KDD conference has seen remarkable growth since its origins as an IJCAI workshop in Detroit in 1989, evolving into a full-fledged research conference in 1995, underscoring the important role data mining as a field has played in extracting knowledge and actionable insights from vast troves of data that is being generated in the digital world around us. This year we received a record 755 submissions to the research program, from which 133 papers were accepted, for an aggregate acceptance rate of 17.6% (quite similar to recent years). Among the academic conferences, the KDD conference has typically more of an emphasis on research motivated by real-world applications. It is important to keep in mind that it is this synergy of research in areas like algorithms, computational geometry, database, graph theory, machine learning, natural language processing, statistics, visualization and many others when applied to problems arising in diverse fields such as web, medicine, climatology, marketing that drives our field forward, makes it vibrant and fun - who would know that ideas in computational geometry can be adapted to construct fast algorithms to improve online advertising and movie recommendations? The breadth of topics covered in this year's research program is truly comprehensive, including social networks, privacy, text mining, predictive modeling, time-series forecasting, spatial data analysis, geometry, and more. We are very fortunate to have 4 world-class keynote speakers this year spanning industry and academia, providing inspirational talks on cutting-edge techniques and issues in web mining, information networks, statistical inference for big data, and social computing. The process of whittling down the initial 734 submissions to the final set of 133 accepted papers required the coordination and time of a large number of willing volunteers. The program committee (PC) consisted of over 350 reviewers (PC members) and 50 senior PC members. In the first phase each submitted paper was automatically assigned to 3 reviewers (after a bidding process). Once the reviews from each of the 3 reviewers were completed, the program chairs rejected papers that did not receive much support from any of the reviewers. We rejected 259 papers at this stage. Special care was taken to minimize the error of rejecting a potentially good paper at this stage. The papers that survived the first phase were assigned to the senior PC members based on their bids, they had the option of initiating a discussion for any of their papers, e.g., if there was significant divergence in scores among reviewers, or if a paper was on the borderline of being accepted. Following the discussion phase, the senior PC members provided a recommendation score and a detailed meta-review for each paper. In the final phase, we (the program chairs) analyzed all of this information, starting with the obvious accept and reject decisions, and then gradually focusing in more detail on the papers near the borderline, seeking additional reviews and input from the PC and senior PC members where appropriate. We also initiated a shepherding phase with 15 papers having the opportunity of fixing mild issues we thought would be possible to address before they can be accepted. 13 of them were accepted after thorough revisions. Finally, it is quite likely that in hindsight some worthy papers may have been rejected as part of this process - these errors are an unfortunate reality of modern computer science conferences, and hard to avoid when a very large number of decisions have to be made over a short time span based on a subjective reviewing process. Nevertheless, we, the PC chairs, are responsible for those unfortunate errors and welcome suggestions on the matter.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it