Search papers, labs, and topics across Lattice.
This study investigates the effectiveness of looped language models in compositional tool-calling scenarios, focusing on their ability to manage multiple API calls and maintain state across interactions. By evaluating both native and retrofitted looped models against traditional models on various benchmarks, the research reveals that increased recurrent depth generally enhances accuracy in multi-step tool use. Notably, adaptive inference strategies provide a more efficient compute-performance balance by dynamically allocating resources, suggesting looped models could significantly enhance agentic systems' planning and execution capabilities.
Looping mechanisms in language models can dramatically improve the accuracy and efficiency of multi-step API interactions, reshaping agentic tool use.
Looped language models have shown promising results on reasoning benchmarks, yet their potential for agentic tool use remains largely unexplored. We study this question in compositional tool-calling settings, where models must coordinate multiple API calls, maintain intermediate state, and preserve dependencies across tool interactions. We evaluate native and retrofitted looped language models on API-Bank, BFCL, and NESTful, comparing looped and non-looped models trained under matched supervised fine-tuning recipes and varying recurrent depth at inference time. In controlled experiments, recurrent computation generally benefits compositional and dependency-aware tool use, while providing smaller and more model-dependent gains on isolated API invocation. Accuracy on multi-step tool use generally increases with recurrent depth; adaptive inference, however, achieves a more favorable compute-performance trade-off by allocating additional computation only when needed. Our results suggest that looped language models are a promising architecture for agentic systems that require reliable planning, coordination, and execution of compositional tool use workflows.
