Limits of CDCL learning via merge resolution
(2023) In LIPIcs 271. p.1-27- Abstract
- In their seminal work, Atserias et al. and independently Pipatsrisawat and Darwiche in 2009 showed that CDCL solvers can simulate resolution proofs with polynomial overhead. However, previous work does not address the tightness of the simulation, i.e., the question of how large this overhead needs to be. In this paper, we address this question by focusing on an important property of proofs generated by CDCL solvers that employ standard learning schemes, namely that the derivation of a learned clause has at least one inference where a literal appears in both premises (aka, a merge literal). Specifically, we show that proofs of this kind can simulate resolution proofs with at most a linear overhead, but there also exist formulas where such... (More)
- In their seminal work, Atserias et al. and independently Pipatsrisawat and Darwiche in 2009 showed that CDCL solvers can simulate resolution proofs with polynomial overhead. However, previous work does not address the tightness of the simulation, i.e., the question of how large this overhead needs to be. In this paper, we address this question by focusing on an important property of proofs generated by CDCL solvers that employ standard learning schemes, namely that the derivation of a learned clause has at least one inference where a literal appears in both premises (aka, a merge literal). Specifically, we show that proofs of this kind can simulate resolution proofs with at most a linear overhead, but there also exist formulas where such overhead is necessary or, more precisely, that there exist formulas with resolution proofs of linear length that require quadratic CDCL proofs. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/95f4df80-baa5-4e78-a91a-e9d3cdc8ee6d
- author
- Vinyals, Marc
; Li, Chunxiao
; Fleming, Noah
LU
; Kolokolova, Antonina
and Ganesh, Vijay
- publishing date
- 2023
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- The International Conferences on Theory and Applications of Satisfiability Testing
- series title
- LIPIcs
- volume
- 271
- pages
- 19 pages
- publisher
- Schloss Dagstuhl - Leibniz-Zentrum für Informatik
- external identifiers
-
- scopus:85170570603
- DOI
- 10.4230/LIPICS.SAT.2023.27
- language
- English
- LU publication?
- no
- id
- 95f4df80-baa5-4e78-a91a-e9d3cdc8ee6d
- date added to LUP
- 2025-11-05 15:49:39
- date last changed
- 2026-08-15 04:01:26
@inproceedings{95f4df80-baa5-4e78-a91a-e9d3cdc8ee6d,
abstract = {{In their seminal work, Atserias et al. and independently Pipatsrisawat and Darwiche in 2009 showed that CDCL solvers can simulate resolution proofs with polynomial overhead. However, previous work does not address the tightness of the simulation, i.e., the question of how large this overhead needs to be. In this paper, we address this question by focusing on an important property of proofs generated by CDCL solvers that employ standard learning schemes, namely that the derivation of a learned clause has at least one inference where a literal appears in both premises (aka, a merge literal). Specifically, we show that proofs of this kind can simulate resolution proofs with at most a linear overhead, but there also exist formulas where such overhead is necessary or, more precisely, that there exist formulas with resolution proofs of linear length that require quadratic CDCL proofs.}},
author = {{Vinyals, Marc and Li, Chunxiao and Fleming, Noah and Kolokolova, Antonina and Ganesh, Vijay}},
booktitle = {{The International Conferences on Theory and Applications of Satisfiability Testing}},
language = {{eng}},
pages = {{1--27}},
publisher = {{Schloss Dagstuhl - Leibniz-Zentrum für Informatik}},
series = {{LIPIcs}},
title = {{Limits of CDCL learning via merge resolution}},
url = {{http://dx.doi.org/10.4230/LIPICS.SAT.2023.27}},
doi = {{10.4230/LIPICS.SAT.2023.27}},
volume = {{271}},
year = {{2023}},
}