Compositional Reasoning

Follow this topic Ask about this topic

Compositional reasoning asks whether AI systems can combine learned pieces into reliable solutions, rather than merely recall facts or match familiar patterns. Approaches include symbolic decomposition and rule chaining, iterative planning and verification, structured neural representations, and training with synthetic examples or reinforcement learning. Researchers test these abilities in settings such as multi-hop question answering, ARC-style visual puzzles, traffic-rule conflicts, and scientific problems requiring combinations of properties. A central tension is that strong performance on individual skills does not guarantee transfer to new compositions, and more inference or training data does not consistently overcome this gap.

Latest papers

39 papers6 papers in the last four weeks, up 50% on the four weeks before. 0.0% of all new papers.

When Verification Fails: How Compositionally Infeasible Claims Escape Rejection
Apr 13, 2026Muxin Liu, Delip Rao, Grace Kim +1Scientific Claim Verification Scientific Reasoning in Language Models
University of Pennsylvania
PolyBench: A Benchmark for Compositional Reasoning in Polyphonic Audio
Mar 5, 2026Yuanjian Chen, Yang Xiao, Han Yin +3Audio Reasoning Audio-Language Model Evaluation
Harbin University of Science and Technology, China · The University of Melbourne, Australia · KAIST, South Korea +1
Learning from Synthetic Data Improves Multi-hop Reasoning
Mar 2, 2026Anmol Kabra, Yilun Yin, Albert Gong +6RL for Language Model Reasoning Multi-Hop Reasoning
Cornell University · University of Cambridge · Stanford University
DriveCombo: Benchmarking Compositional Traffic Rule Reasoning in Autonomous Driving
Mar 2, 2026Enhui Ma, Jiahuan Zhang, Guantian Zheng +10Autonomous Driving Benchmarks VLMs for Autonomous Driving
Autolab, Westlake University · Li Auto Inc · Tsinghua University +1
Predictive Reasoning with Augmented Anomaly Contrastive Learning for Compositional Visual Relations
Mar 1, 2026Chengtai Li, Yuting He, Jianfeng Ren +4Compositional Reasoning Visual Reasoning
School of Computer Science, University of Nottingham Ningbo China, Ningbo 315100, China. · Institute of Biomedical Engineering, Ningbo institute of materials technology and engineering, Chinese Academy of Sciences. · School of Electrical & Electronic Engineering, Nanyang Technological University, Nanyang Link, 639798 Singapore.
Emergent Manifold Separability during Reasoning in Large Language Models
Feb 23, 2026Chanwoo Chun, Alexandre Polo, SueYeon ChungRepresentation Geometry in Language Models CoT Reasoning
Department of Physics, Harvard University · Center for Data Science, New York University · Kempner Institute, Harvard University +1
Recursive Concept Evolution for Compositional Reasoning in Large Language Models
Feb 17, 2026Sarim ChaudhryCompositional Reasoning Test-Time Adaptation
Purdue University
Poly-attention: a general scheme for higher-order self-attention
Feb 2, 2026Sayak Chakrabarti, Toniann Pitassi, Josh AlmanSelf-Attention Compositional Reasoning
Computer Science, Columbia University, New York, NY 10027.
Explain Before You Answer: A Survey on Compositional Visual Reasoning
Aug 24, 2025Fucai Ke, Joy Hsu, Zhixi Cai +10Compositional Reasoning Visual Reasoning
Monash University · Stanford University · University of Washington +3

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

When Verification Fails: How Compositionally Infeasible Claims Escape Rejection

Muxin Liu, Delip Rao, Grace Kim, Chris Callison-Burch

University of Pennsylvania

Scientific Claim Verification Scientific Reasoning in Language Models Shortcut Learning Compositional Reasoning

Open paper AskPDF arXiv

Scientific claim verification, the task of determining whether claims are entailed by scientific evidence, is fundamental to establishing discoveries in evidence while preventing misinformation. This process involves evaluating each asserted constraint against validated evidence. Under the Closed-World Assumption (CWA), a claim is accepted if and only if all asserted constraints are positively supported. We show that existing verification benchmarks cannot distinguish models enforcing this standard from models applying a simpler shortcut called salient-constraint checking, which applies CWA's rejection criterion only to the most salient constraint and accepts when that constraint is supported. Because existing benchmarks construct infeasible claims by perturbing a single salient element they are insufficient at distinguishing between rigorous claim verification and simple salient-constraint reliance. To separate the two, we construct compositionally infeasible claims where the salient constraint is supported but a non-salient constraint is contradicted. Across model families and modalities, models that otherwise saturate existing benchmarks consistently over-accept these claims, confirming the prevalence of such shortcut reasoning. Via model context interventions, we show that different models and prompting strategies occupy distinct positions on a shared ROC curve, indicating that the gap between model families reflects differences in verification threshold rather than underlying reasoning ability, and that the compositional inference bottleneck is a structural property of current verification behavior that strategy guidance alone cannot overcome.