Dataset-Specific Strategies for the E Theorem Prover

11 pages•Published: April 6, 2025

Abstract

The E automated theorem proving system has an “automatic” mode that analyzes the input problem in order to choose an effective proof search strategy. A strategy includes the term/literal orderings, given clause selection heuristics, and a number of other parameters. This paper investigates the idea of creating one strategy for a given dataset of problems by merging the strategies chosen by E’s automatic mode over all of the problems in the dataset. This strategy merging approach is evaluated on the MPTPTP2078, VBT, and SLH-29 datasets. Surprisingly, the merged strategies outperform E’s automatic mode over all three datasets.

Keyphrases: atp, eprover, heuristic, merging

In: Konstantin Korovin, Stephan Schulz and Michael Rawson (editors). Proceedings of the 14th and 15th International Workshops on the Implementation of Logics, vol 21, pages 34-44.

Links:	https://easychair.org/publications/paper/PXgt
	https://doi.org/10.29007/jn6j

BibTeX entry

@inproceedings{IWIL-2024:Dataset_Specific_Strategies_E,
  author    = {Jack McKeown and Geoff Sutcliffe},
  title     = {Dataset-Specific Strategies for the E Theorem Prover},
  booktitle = {Proceedings of the 14th and 15th International Workshops on the Implementation of Logics},
  editor    = {Konstantin Korovin and Stephan Schulz and Michael Rawson},
  series    = {Kalpa Publications in Computing},
  volume    = {21},
  publisher = {EasyChair},
  bibsource = {EasyChair, https://easychair.org},
  issn      = {2515-1762},
  url       = {/publications/paper/PXgt},
  doi       = {10.29007/jn6j},
  pages     = {34-44},
  year      = {2025}}

Download PDF Open PDF in browser