Compositional Generalization

Follow this topic Ask about this topic

Compositional generalization studies whether models can reuse learned concepts, skills, and rules in combinations they did not encounter during training. Approaches include modular and object-centric architectures, factorized or symbolic representations, and training methods such as curricula, structured reasoning traces, and reinforcement learning. Researchers test these abilities on controlled language, vision, arithmetic, planning, and robotics tasks, often holding out combinations, deeper structures, or longer sequences. Results remain uneven: models often recombine familiar parts more reliably than they discover new ones, and strong test accuracy can reflect shortcuts rather than reusable structure.

Latest papers

112 papers15 papers in the last four weeks, up 50% on the four weeks before. 0.1% of all new papers.

From Copying to Composition: How Language Models Develop In-Context Learning
Oct 13, 2026Minglai Yang, Xinyu Guo, Renxiong Wang +3In-Context Learning Compositional Generalization
Scale AI · University of Arizona
Generalizable Model Editing Could Be More Than Direct Generalization
Oct 11, 2026Xiaotian Ye, Chunyao Yang, Xiaohan Wang +4Model Editing Language Model Editing
University of Cambridge · NLPR, MAIS, Institute of Automation, Chinese Academy of Sciences · School of Advanced Interdisciplinary Sciences, University of the Chinese Academy of Sciences +3
Knowledge transfer between simple discrete worlds
Oct 10, 2026Silvère Gangloff, Jan HulaLogical Reasoning Compositional Generalization
University of Ostrava
Mu-DisCoCat: A Variational Pipeline for Compositional Generalization on Quantum Processors
Oct 6, 2026Mina Abbaszadeh, Matilda Karabina Moore, Raem Haq +2Quantum Machine Learning Variational Quantum Circuits
Department of Computer Science University College London London, United Kingdom · Institute for Logic, Language and Computation University of Amsterdam Amsterdam, Netherlands
Arm-wise Compositional Generalization in Dual-Arm Vision-Language-Action Models
Oct 5, 2026Zaibin Zhang, Binghao Ran, Yuhan Wu +10Bimanual Robotic Manipulation Vision-Language-Action Models
Dalian University of Technology · Shanghai Jiao Tong University · The Chinese University of Hong Kong, Shenzhen +1
Gated Target Propagation for Compositional Generalization in Continual Learning
Oct 3, 2026Abdel Mfougouon Njupoun, Colin Bredenberg, Blake Aaron Richards +1Continual Learning Compositional Generalization
Université de Montr´eal Mila Quebec AI Institute · University of Oregon Mila Quebec AI Institute · McGill University Mila Quebec AI Institute
Compound interpretation is based on analogy
Oct 1, 2026Tian Shen, Harald BaayenAnalogical Reasoning Distributional Semantics
School of Foreign Languages, Northwest University, Xi’an, China · Department of Quantitative Linguistics, University of Tübingen, Tübingen, Germany
Text-to-3D Policy: Fine-Grained Language-Behavior Alignment for Unseen Specification Generalization
Sep 30, 2026Xinhao Yang, Wenhao Wu, Ning Lv +5Language-Conditioned Robot Manipulation Generalization in Robotic Manipulation
Nanjing University · Australian National University · University of Technology Sydney
Agent Priors-guided Policy Learning
Sep 28, 2026Puming Jiang, Tianrun Hu, Haozhe Du +4Robot Skill Learning Robot Imitation Learning
National University of Singapore
SPOC-Net: Single-Primitive Online Composition Network for GNSS Jamming Set Recognition
Sep 28, 2026Zhihan Zeng, Kaihe Wang, José A. López-Salcedo +2Multi-Label Classification Compositional Representation Learning
National Key Laboratory of Wireless Communications, University of Electronic Science and Technology of China (UESTC), Chengdu, China · Shenzhen Institute for Advanced Study, UESTC, Shenzhen 518110, China · Universitat Aut`onoma de Barcelona, 08193 Barcelona, Spain
Do World Models Learn Global Understanding?
Sep 28, 2026Alexander Detkov, Matt ThomsonLLM World Models World Model Learning
California Institute of Technology Pasadena, CA 91125, USA
Witness: Discovery, Deciphering, and Epiphany in Interactive Puzzle Environments
Sep 26, 2026Guanghan Ning, Ping Liu, Linyi Li +6AI Agent Benchmarks RL for LLM Agents
Fleet AI · University of Nevada, Reno · Simon Fraser University +1
Amadeus: When Models of People Meet
Sep 24, 2026Karl HannaHuman Behavior Simulation Multi-Agent Systems
Queen’s University Belfast United Kingdom
Route-Verify-Vote: Procedure-Conditioned Self-Consistency for Mixed-Domain Reasoning
Sep 23, 2026Xinchen XiaoCommonsense Reasoning Self-Consistency in Language Models
Xinjiang University
Exactness at Inference: A Representational Criterion for Out-of-Distribution Generalization
Sep 21, 2026Filipe Marinho Rocha, Inês Dutra, Vítor Santos Costa +1OOD Generalization Compositional Generalization
Faculdade de Ciências da Universidade do Porto, Porto, Portugal · INESC TEC, Campus da FEUP, Porto, Portugal · Faculdade de Engenharia da Universidade do Porto, Rua Dr. Roberto Frias, Porto, Portugal +1
Compositional Reasoning in Language Models under Reinforcement Learning Post-Training
Sep 16, 2026Yu He, Yingxi Li, Yifei Wang +1Compositional Reasoning RL Post-Training
Stanford University · Amazon AGI Labs
What Do Current Systematic Generalization Tasks Miss? A Reasoning-Centered Analysis
Sep 16, 2026Chengwen Qi, Deheng Ye, Yatao BianLogical Reasoning Inductive Inference
National University of Singapore · Nanyang Technological University
What You Can't See Is Still What You Learn: A Preregistered Sixty-Society Confirmation That Evidence Masking Drives Compositional Generalization
Sep 15, 2026Narcis MarincatCompositional Generalization Modular Neural Networks
Independent researcher.
Type Diversity Enables Transformers to Generalise Compositionally
Sep 11, 2026Anssi Moisio, Mathias Creutz, Mikko KurimoTransformer Semantic Parsing
Aalto University · University of Helsinki
CoGe-GCD: Reframing Generalized Category Discovery with Compositional Generalization
Sep 9, 2026Luyao Tang, Jiewei Zheng, Kunze Huang +3Geometric Representation Learning Classification
The University of Hong Kong, Hong Kong SAR, China · Xiamen University, Xiamen, China · Shenzhen University, Shenzhen, China.
Reachability Is Not Generalization: Understanding Verb--Noun Decomposition in Assembly Action Recognition
Sep 4, 2026Changyi Li, Yu XiaoCompositional Generalization Human Activity Recognition
Department of Information and Communications Engineering Aalto University 02150 Espoo, Finland
TPR-Attention for Combinatorial Generalization
Aug 31, 2026Melisa Civelekoğlu, Isabeau Prémont-SchwarzAttention Mechanisms Compositional Representation Learning
Universit´e Laval · Institut intelligence et donn´ees · Mila – Qu´ebec Artificial Intelligence Institute
Learning Simple Test-Time Environments for LLM Web Agents
Aug 29, 2026Junxuan Li, Zijun Liu, Ziyi Huang +4LLM Agents Web Agents
College of Computer Science and Technology, Jilin University, Jilin, China · Dept. of Comp. Sci. & Tech., Tsinghua University, Beijing, China · School of Electronic and Information Engineering, Beijing Jiaotong University, Beijing, China +2
Compositional Generalization via Structural Identification in a Category-Theoretic Framework
Aug 26, 2026Akihiro Maeda, Thomas Seiller, Yohei OsekiCompositional Generalization
The University of Tokyo Tokyo, Japan · CNRS Paris, France
The Illusion of Control: Why Bare Classifier Inversion Silently Fails in Concept-Bottleneck Text Generation
Aug 24, 2026Qi Bing, Xiaowei ShaoControllable Text Generation Concept Bottleneck Models
Shanghai Jiao Tong University
Learning Generalizable Behaviors for Terminal Agents
Aug 23, 2026Yihang Yao, Bo Pang, Xuan Phi Nguyen +3Terminal Agents Generalization in RL
Salesforce AI Research · Carnegie Mellon University
Complexity Induction: Compositional Generalization via Structured Training Distortion
Aug 21, 2026Aleksandr V. AbramovCompositional Generalization Compositional Representation Learning
Independent Researcher
What You Can't See Is What You Learn: Slot-Selective Evidence Masking Favors Compositional Generalization in Shared-Genome Language-Model Societies
Aug 20, 2026Narcis MarincatCompositional Generalization Multi-Agent Communication
Independent researcher.
Beyond Multimodal Alignment: Shared Physical Representations Across Sensors and Action Orders
Aug 19, 2026Kaizhen Tan, Xin Xu, Siru Tao +4World Models Cross-Modal Alignment
New York University, New York, NY, USA · Carnegie Mellon University, Pittsburgh, PA, USA · Columbia University, New York, NY, USA
Quantum Models with Multi-Stage Training for Compositional Concept Generalization
Aug 16, 2026Mina Abbaszadeh, Matilda Karabina Moore, Mehrnoosh Sadrzadeh +1Compositional Generalization Compositional Representation Learning
Department of Computer Science University College London London, United Kingdom · Institute for Logic, Language and Computation University of Amsterdam Amsterdam, Netherlands

Oct 13, 2026J/K move · Enter open · S save

From Copying to Composition: How Language Models Develop In-Context Learning

Minglai Yang, Xinyu Guo, Renxiong Wang, Sunjie Hou+2

Scale AI · University of Arizona

In-Context Learning Compositional Generalization

Open paper AskPDF arXiv

In-context learning (ICL) allows language models to adapt their behavior from instructions and demonstrations in a prompt, but accuracy alone does not reveal how contextual information controls their predictions. We study the acquisition and trainability of ICL with a controlled diagnostic that separates copying a nearby demonstration from integrating information across the prompt. Across 17 models, about a third prefer the copied answer, a few are undecided, and the rest prefer the rule-consistent answer, which goes with higher accuracy and with control by the requested value rather than the closest conflicting example. Checkpoint analyses show that models begin prioritizing task-relevant prompt tokens over conflicting examples long before their accuracy improves beyond chance level. In matched training experiments, synthetic documents that define endings and then query them shift models toward compositional responses on the controlled diagnostic, whereas fine-tuning on definition exercises built from rewritten real documents improves accuracy when prompts reassign label meanings. The results suggest that ICL is not acquired as a single, uniform capability. Instead, prompt control, answer preference, construction without candidates, and transfer can develop along distinct trajectories.