Contexture Lab
The Contexture Lab recovers the continuity that categorical thinking erases. Reality exists as interconnected continua without natural boundaries, yet we impose categorical partitions to comprehend it—introducing bias and concealing underlying structure. We develop methods to reveal the spectra and nuance that these partitions obscure.
Lab News
-
October 16, 2026 — Talk at CCS 2026
Sadamori Kojaku presents “Uncovering simultaneous breakthroughs with a robust measure of disruptiveness” (with Munjung Kim and Yong-Yeol Ahn) at the Conference on Complex Systems (CCS 2026) in Binghamton, Friday, October 16. Paper Slides -
October 14, 2026 — Poster at CCS 2026
Chand Sahil Mansuri presents the poster “Paired Perturbation Evaluation Reveals Bounded and Non-Linear Failure Propagation in Clinical Language Pipelines” (with Sadamori Kojaku) at the Conference on Complex Systems (CCS 2026) in Binghamton, Wednesday, October 14. -
October 13, 2026 — Panel at CCS 2026
Sadamori Kojaku joined the yrCSS panel “Practical and Ethical Insights on Using AI in Complex Systems” at the Conference on Complex Systems (CCS 2026) in Binghamton, with Evangelia Panagakou, Juniper Lovato and Melanie Mitchell. -
October 09, 2026 — Talk at the yrCSS CCS Warm-Up 2026
Sadamori Kojaku spoke on reproducible research and automation, and how to regain work-life balance, at the yrCSS CCS Warm-Up 2026 at Binghamton University. Slides -
September 29, 2026 — New preprint: Doc2LoRA Provides Decodable Representations of Scientific Ideas
Chand Sahil Mansuri, Joel Zachariah, and Sadamori Kojaku represent scientific papers as LoRA adapters generated by a Doc-to-LoRA hypernetwork, so that every point in the space, including mixtures, is a language model that can be queried. arXiv:2609.38374 · Project page · Code -
September 29, 2026 — New preprint: Multi-agent discussion gains less when dissent is withheld
Chand Sahil Mansuri, Xin Wang, Mengying Li, Bryan Acton, Rory Eckardt, Dhaval Patel, and Sadamori Kojaku model when multi-agent LLM discussion improves accuracy and when it locks in a wrong consensus. Discussion can overturn an incorrect majority only when agents withhold dissent at a rate below a critical threshold. arXiv:2609.38324 -
August 07, 2026 — Wei Xiang Zheng completes BUPNUR summer research
A BUPNUR summer student researcher, Wei Xiang Zheng, finished the project on multi-agent systems for industrial applications! -
May 08, 2026 — Keynote at CompleNet 2026
Sadamori Kojaku delivered a keynote on neural embeddings and network science at CompleNet 2026 in Zaragoza, Spain (May 4–8, 2026). -
April 09, 2026 — Press coverage in Physics World
Have you published a disruptive paper? New machine-learning tool helps you check -
April 01, 2026 — Press coverage on Binghamton University News
Eureka! Researchers develop new way to detect breakthroughs in science (posted 2026-05-07) -
February 16, 2026 — Lab presents at NetSci 2026
Xuanchi Li and Sadamori Kojaku will present their work at NetSci 2026 in Boston (June 1-5, 2026). Presentations include a contributed talk on detectability thresholds in weighted modular networks and a poster on fast inference for maximum entropy configuration models. -
January 01, 2026 — Press coverage in APS Physics
APS Physics v19/63 -
October 01, 2025 — VizTrust highlighted in Communications of the ACM 🎉
Xin Wang’s VizTrust project was featured in Communications of the ACM. It’s truly an honor! -
September 23, 2025 — Invited to speak at CompleNet 2026
Sadamori Kojaku is invited for a keynote speach. Read more -
July 21, 2025 — Link prediction paper presented at ICML 2025
Our paper on implicit degree bias in link prediction was presented at ICML 2025 in Vancouver, Canada (Acceptance rate: 26.9%). -
June 01, 2025 — Lab presents 3 papers at NetSci 2025
Xin Wang and Xuanchi Li presented their work at NetSci 2025 in Maastricht, Netherlands on implicit degree bias in link prediction and near-linear time algorithms for configuration models. -
May 01, 2025 — VizTrust featured on BingUNews
Xin Wang’s VizTrust project was featured on Binghamton University News -
May 01, 2025 — VizTrust receives research grant
Xin Wang’s VizTrust project successfully landed Multidisciplinary Summer Research Grant ($3,012.58) from the Center for Cognitive Applications. -
April 01, 2025 — Lab presents at CompleNet 2025
Xuanchi Li and Sadamori Kojaku presented papers on link prediction bias and configuration model algorithms at CompleNet 2025 in Brazil. -
March 01, 2025 — Artwork selected for 2025 Art of Science competition
Sadamori Kojaku’s ‘Gravity of Ideas: Mapping Science at Binghamton University’ was chosen by Zagham Shaw, digital content creator at the Office of Research Advancement, from the Visualizing the Unseen Category. View artwork -
July 08, 2024 — Xin Wang presents at ACM CUI 2024
Xin Wang presented poster on ‘User Trust Modeling in Conversational User Interface Based on Word Embedding Bias’ at The ACM Conference on Conversational User Interfaces. -
March 20, 2024 — Xin Wang presents at NERCCS 2024
Xin Wang presented poster on analyzing patient reviews on Google Map hospital profiles through neural embedding and network modeling at NERCCS in Clarkson, NY.
The Contexture Lab is led by Sadamori Kojaku, Assistant Professor in the School of Systems Science and Industrial Engineering (SSIE) at Binghamton University.
Talks, slides and selected works: skojaku.github.io/research