arXiv AI By Joana Rosa, Pedro Santos, Valdemar Oliveira, Rom\~ao Silva, L. Miguel Silveira, Bruno Martins

Grounded Evaluation and Repair for NL-to-PDDL Problem Generation

Read the original on arXiv AI →

The paper presents an end‑to‑end pipeline for translating natural language planning descriptions into PDDL problem instances using large language models. It incorporates multiple checks—syntactic parsing, planner success, domain conformance, an LLM critic, and iterative repair—to ensure faithfulness to the original task. Experiments on Planetarium, AutoPlanBench, and curated PDDL~2.1 problems reveal that operational success can diverge from benchmark‑reference reconstruction, and that structured repair improves outcomes while PDDL~2.1 remains challenging for reference reconstruction.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

arXiv AI
Aug 19

LLM-Only PDDL Domain Repair with Open-Weight Models

The paper evaluates how well open-weight large language models can repair Planning Domain Definition Language (PDDL) models using only LLMs. Experiments show that while the best LLM achieves an F1 score of 0.87—an improvement of 0.38 over a symbolic baseline—it still fails to reliably satisfy test constraints, with a mean test pass rate of only 0.82 and as low as 0.06 on the Thoughtful domain. The study concludes that current open-weight models cannot guarantee the necessary test constraint satisfaction for dependable automated model repair.

By Nader Karimi Bavandpour, Pascal Bercher
arXiv AI
6d ago

Which LLM is Best for Translating Natural Language Goals to PDDL

The paper evaluates how well current Large Language Models can translate natural language goals, written by video game testers, into well‑formed PDDL targets for classical planning. Using a carefully designed prompt template, six state‑of‑the‑art LLMs were tested on correctness, speed, and error tendencies with real‑world benchmarks. All models achieved high correctness (>92%), with Gemini 2.5 Flash reaching 96% accuracy and the fewest false positives, while GPT‑4.1 was the fastest, yet differences in performance and occasional failures due to ambiguity and domain limits remain.

By Tomas Balyo, Lukas Chrpa, G. Michael Youngblood
arXiv Computation and Language
Aug 31

NL2AGBench: Benchmarking LLM Auto-Formalization for AlphaGeometry

NL2AGBench is a benchmark that evaluates how well large language models can translate English geometry problems into the formal language required by AlphaGeometry’s theorem‑proving engine. The study tests ten state‑of‑the‑art LLMs, comparing executable translation accuracy, syntactic correctness, and error types, and finds a large gap between closed‑source and open‑source models. The authors also propose an error taxonomy and test mitigation strategies such as few‑shot prompting, fine‑tuning, and human‑guided hinting, which improve performance across model families.

By Samuel Xiao, Judy Song, Rory Hu, Ziliang Zong