Novelty Search

Follow this topic Ask about this topic

Momentum

2 papers in the last four weeks, with none the four weeks before. 0.0% of all new papers.

Jul 13Week of Sep 28

Latest papers 10

All topics

Let CSP Be Your ANCHOR: Adaptive Crystal Search over Frozen Structure Priors
Sep 27, 2026Emma Lei Hovmand, Jonas Elsborg, Melih Kandemir +1Novelty Search Crystal Structure Prediction
Department of Energy Conversion and Storage, Technical University of Denmark · Department of Mathematics and Computer Science, University of Southern Denmark
RINI: Seeing the Prior Is Not Enough
Sep 27, 2026Hongyi Du, Tianyi Zhang, Heng Wang +7Novelty Search AI-Assisted Scientific Research
University of Illinois Urbana-Champaign · Harvey Mudd College · National University of Singapore +2
Intelligence from Learnable Novelty
Jul 20, 2026Yanbo Zhang, Michael LevinNovelty Search Intrinsic Motivation
Allen Discovery Center at Tufts University, Medford, MA, 02155, USA · Wyss Institute for Biologically Inspired Engineering at Harvard University, Boston, MA, 02115, USA
Novelty Search with Cross-Task Collaborative Discovery
Jul 1, 2026Jiao Liu, Yanchi Li, Hua Yu +2Novelty Search Evolutionary Optimization
College of Computing & Data Science, Nanyang Technological University, Singapore · School of Computer Science, China University of Geosciences · School of Intelligent Rail Transportation, Dalian Jiaotong University +1
Heuresis: Search Strategies for Autonomous AI Research Agents Across Quality, Diversity and Novelty
Jun 23, 2026Antonis Antoniades, Deepak Nathani, Ritam Saha +6Novelty Search AI Agent Evaluation
University of California, Santa Barbara · Hexo AI
Novelty-based Tree-of-Thought Search for LLM Reasoning and Planning
May 7, 2026Leon Hamm, Zlatan AjanovicNovelty Search Tree of Thoughts
1RWTH Aachen
Novelty-Based Generation of Continuous Landscapes with Diverse Local Optima Networks
Apr 23, 2026Kippei Mizuta, Shoichiro Tanaka, Shuhei Tanaka +1Novelty Search Evolutionary Optimization
The University of Fukuchiyama, Fukuchiyama, Kyoto, Japan
Evolving Many Worlds: Towards Open-Ended Discovery in Petri Dish NCA via Population-Based Training
Apr 13, 2026Uljad Berdica, Jakob Foerster, Frank Hutter +1Neural Cellular Automata Novelty Search
1FLAIR, University of Oxford · 2ELLIS Institute Tübingen · University of Freiburg +2
Illuminating the Three Dogmas of Reinforcement Learning under Evolutionary Light
Jul 15, 2025Mani Hamidi, Terrence W. DeaconReinforcement Learning Novelty Search
Department of Computer Science, University of Tübingen, Germany · Department of Anthropology, University of California, Berkeley, USA
BEACON: A Bayesian Optimization Inspired Strategy for Efficient Novelty Search
Jun 5, 2024Wei-Ting Tang, Ankush Chakrabarty, Joel A. PaulsonNovelty Search Black-Box Optimization
University of Wisconsin-Madison, Department of Chemical and Biological Engineering, Madison, WI, USA. · The Ohio State University, Department of Chemical and Biomolecular Engineering, Columbus, OH, USA. · Insulet Corporation, Acton, MA, USA.

Sep 27, 2026, cond-mat.mtrl-sciJ/K move · Enter open · S save

Let CSP Be Your ANCHOR: Adaptive Crystal Search over Frozen Structure Priors

Emma Lei Hovmand, Jonas Elsborg, Melih Kandemir, Arghya Bhowmik

Department of Energy Conversion and Storage, Technical University of Denmark · Department of Mathematics and Computer Science, University of Southern Denmark

Novelty Search Crystal Structure Prediction Crystal Structure Generation

Open paper AskPDF arXiv

De novo crystal generation (DNG) models decide where to search in composition space and how to generate structures with one set of weights. We argue that discovery is better served by separating the two. A crystal structure prediction (CSP) model is a physical prior that should be improved by likelihood training, while rewards, including novelty measured against the search's own history, should act on a search over compositions. We introduce ANCHOR, a GRPO composition policy trained with multi-objective rewards around a frozen CSP model, and continuous adaptive novelty (CAN), a graded novelty score against known structures and a growing discovery history. Using the frozen CSP model as a fixed ruler under one evaluator, we test where adaptation should act. Replacing DNG compositions with ANCHOR's policy on the same CSP backbone raises MSUN from 11.4% to 47.6% and SUN from 1.1% to 22.1% at 99.9% formula uniqueness. Fine-tuning DNG models directly on the same rewards instead moves their composition marginal without raising their on-hull fraction. We show that KL-regularized fine-tuning of a DNG model can only reweight chemistry the pretrained model already supports by a bounded factor, while unregularized DNG fine-tunes move toward known or less stable chemistry. Even a stability-only reward routed into ANCHOR's CSP backbone roughly halves SUN relative to the frozen backbone, whereas likelihood training on structures found during search can improve a CSP backbone. Under MatterGen's evaluation pipeline, ANCHOR raises state-of-the-art MSUN from 29.2% to 41.3%, transfers without retraining to two further CSP backbones, and reaches 47.1% after distillation into Crystalite-CSP. As with any model optimised against a potential, its on-hull rate depends on that potential.

Momentum

Related topics

Latest papers 10