Search papers, labs, and topics across Lattice.
This paper introduces a framework that translates natural language descriptions of strategic requirements into formal ATL/ATL* specifications using Large Language Models (LLMs). By creating a novel expert-validated dataset for this translation task, the authors demonstrate that fine-tuned small open-weight models can achieve semantic accuracy comparable to proprietary API baselines, while maintaining data privacy. The findings reveal that the reliability of judges declines as generator strength increases, highlighting the nuanced performance dynamics between different model types in this context.
Fine-tuned open-weight models can match proprietary LLMs in translating strategic requirements into formal specifications, all while keeping sensitive data on-premises.
A rigorous formalization of system requirements is a fundamental prerequisite for the verification of Multi-Agent Systems (MAS). However, writing correct formal specifications is well known as an error-prone, time-consuming, and expertise-intensive task. This difficulty is further accentuated in MAS, where requirements must capture strategic abilities and temporal objectives. At present, there is no established methodology for deriving MAS specifications from natural language. We present a framework for translating Natural Language descriptions of strategic requirements into well-formed ATL/ATL* formulas using Large Language Models (LLMs). Since no available dataset supports supervised learning for the NL-to-ATL/ATL* translation task, we create and curate a novel expert-validated dataset, employed for training and evaluating fine-tuned models. On a held-out test set, evaluated under the LLM judge that best agrees with expert annotations, in-domain fine-tuning of small open-weight models (3 - 7B parameters) matches strong few-shot proprietary API baselines. Our best fine-tuned system reaches 0.84 semantic accuracy, statistically on par with 0.86 for the strongest few-shot proprietary baseline, while keeping requirements on-premises. We further find that judge reliability is inverse to generator strength. The open-weight Llama-3.3-70B tracks human verdicts most closely, whereas the strongest proprietary models are the least reliable judges, over-rejecting faithful paraphrases of the reference. To assess the practical applicability of the generated specifications, we embed our tool to an existing strategic logics model checker, enabling non-expert users to specify strategic properties in natural language.
