“Co-Evolving Structured Knowledge and Reasoning in Language Models” has been accepted as a conference paper at the Conference on Language Modeling (COLM) 2026. AI-MI authors: Kilian Q. Weinberger, AI-MI Associate Director; Yoav Artzi and Jennifer J. Sun, AI-MI senior...
Publications
AI-MI–Supported Research Cuts Diffusion Sampling Cost 10× by Tuning the Timestep Schedule
An AI-MI–supported study shows how to cut the cost of generating an image with a diffusion model by roughly ten times — not by changing the model, but by choosing better points along its sampling schedule. Sampling from a diffusion model means…
AI-MI–Supported Study Confirms Superconductivity Across a Family Built Around an AI-Predicted Compound
An artificial-intelligence method predicted PtPb3Bi would superconduct. A new AI-MI–supported study in Chemistry of Materials confirms it, and finds superconductivity across the wider MPb4-xBix family (M = Au, Pd, Rh).
AI-MI Research Opens the AI-Driven Scientific Discovery Session at KDD 2026
AI-MI–supported research is on the program at KDD 2026, the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, running August 9–13 at ICC Jeju in Korea.
AI-MI–Supported Study Turns Moiré Flat-Band Engineering Into Design Rules
Twisted two-dimensional materials can host flat electronic bands — the setting for superconductivity, correlated insulators, and topological order — but which of those states appears has largely been worked out one material at a time.
AI-MI–Supported Study Rebuilds the Theory of MgB₂, the Record-Tᴄ Conventional Superconductor
A new AI-MI–supported study reconstructs the theory of MgB₂ — the phonon-mediated superconductor with the highest known critical temperature, roughly 39 K — from minimal ab initio input. In “Quantum geometry and critical temperature enhancement in…
AI-MI–Supported Research Introduces Continuous-Query Language Models That Externalize Factual Knowledge
An AI-MI–supported study proposes a new architecture for language models that externalizes factual knowledge to a knowledge base rather than storing it in model weights. “Co-LMLM: Continuous-Query Limited Memory Language Models” by Yair Feldman, Linxi Zhao, Nathan...
Separating Prediction and Memory Improves Transformers
An AI-MI–supported study proposes that standard Transformers overload a single computation stream to both predict the next token and store state for future tokens—and that disentangling these roles yields more efficient, higher-performing models. Giovanni Monea,...
AI-MI Research Identifies Anisotropy Key to Magnon-Mediated Superconductivity in 2D Ferromagnets
An AI-MI–supported theoretical study shows that a small easy-plane magnetic anisotropy is the key ingredient enabling magnon-mediated superconductivity in 2D itinerant ferromagnets. Vladimir Calvera, Heqiu Li, Yijie Wang, B. Andrei Bernevig (Princeton), and Andrey V....