aiqbench analysis
This research demonstrates that a two-stage reinforcement learning curriculum, which transitions from symbolic data to natural language, significantly improves LLMs' ability to generalize procedures across different representations. The approach enables a compact 1.5B model to match the planning performance of much larger zero-shot models, suggesting that cross-representation generalization functions as a form of generative analogy. For practitioners, this offers a viable path to enhance model efficiency without relying solely on scaling parameters or extensive natural language-only training.
Can Large Language Models Generalize Procedures Across Representations?
arXiv:2602.03542v2 Announce Type: replace-cross Abstract: Large language models (LLMs) are trained and tested extensively on symbolic representations such as code and graphs, yet real-world user tasks are often specified in natural language. To what extent can LLMs generalize across these representations? Here, we approach this question by studying isomorphic tasks involving procedures represented in code, graphs, and natural language (e.g., scheduling steps in planning). We find that training LLMs with popular post-training methods on graphs or code data alone does not reliably generalize to corresponding natural language tasks, while training solely on natural language can lead to inefficient performance gains. To address this gap, we propose a two-stage reinforcement learning curriculum that first trains on symbolic, then natural language data. The curriculum substantially improves model performance across model families and tasks. Remarkably, a 1.5B Qwen model trained by our method can closely match zero-shot GPT-4o in naturalistic planning. Finally, our analysis suggests that successful cross-representation generalization can be interpreted as a form of generative analogy, which our curriculum effectively encourages. The dataset and code used in this paper can be found \href{https://github.com/fangru-lin/procedure_generalization_llm}{here}.
Title:Can Large Language Models Generalize Procedures Across Representations?
Abstract:Large language models (LLMs) are trained and tested extensively on symbolic representations such as code and graphs, yet real-world user tasks are often specified in natural language. To what extent can LLMs generalize across these representations? Here, we approach this question by studying isomorphic tasks involving procedures represented in code, graphs, and natural language (e.g., scheduling steps in planning). We find that training LLMs with popular post-training methods on graphs or code data alone does not reliably generalize to corresponding natural language tasks, while training solely on natural language can lead to inefficient performance gains. To address this gap, we propose a two-stage reinforcement learning curriculum that first trains on symbolic, then natural language data. The curriculum substantially improves model performance across model families and tasks. Remarkably, a 1.5B Qwen model trained by our method can closely match zero-shot GPT-4o in naturalistic planning. Finally, our analysis suggests that successful cross-representation generalization can be interpreted as a form of generative analogy, which our curriculum effectively encourages. The dataset and code used in this paper can be found \href{this https URL}{here}.
Computation and Language (cs.CL); Machine Learning (cs.LG)
arXiv:2602.03542 [cs.CL]
(or arXiv:2602.03542v2 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2602.03542 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 283 words
Related coverage
The Role of Fine-grained Harm Signals in LLM Safety
arXiv · Sep 18, 2026
The Generalized BEADS Representation: An Informationally Complete Visualization of Mixed Quantum States
arXiv · Oct 15, 2026
Generalized PINN-Based Inverse Estimation of Spatially Varying Heat Transfer Coefficients with Transfer Learning
arXiv · Oct 14, 2026
Differentiated Aggregation to Improve Generalization in Federated Learning
arXiv · Aug 27, 2026