SlideShare a Scribd company logo
Unsatisfiability Proofs for
Parallel SAT Solver Portfolios with
Clause Sharing and Inprocessing
Tobias Philipp
International Center for Computational Logic
Technische Universität Dresden
Parallel Portfolio Approach
The Portfolio Approach
F0
Solver1 F0
Solver2 F0
Solver3 F0
Solver4 F0
F1
F2
F3
F4
SAT
1
The Cooperative Portfolio Approach
F0
Solver1 F0
Solver2 F0
Solver3 F0
Solver4 F0
F1
F2
F3
F4
SAT
2
The Cooperative Portfolio Approach
F0
Solver1 F0
Solver2 F0
Solver3 F0
Solver4 F0
F1
F2
F3
F4
SAT
SAT
2
Inprocessing in Clause Sharing SAT Solvers
ppfolio
Roussel, 2012
ManySAT
Hamadi et al, 2009
PLingeling
Biere, 2010
PLingeling
Biere, 2013
Inprocessing Clause Sharing
3
A Formal Model
Asymmetric Tautologies
Asymmetric literal addition
alaF (C) = C ∪ {L | {L1, . . . , Ln, L} ∈ F and {L1, . . . , Ln} ⊆ C}
Example
F = {{p, q}, {p, ¬q, r}, {¬r, ¬q}}
alaF ({p}) = {p, ¬q}
C is an asymmetric tautology
if there is n ∈ N st alan
F (C) is a tautology
Lemma
1. F ≡ F ∪ {C}, if C is an AT wrt F
2. {L1, . . . , Ln} is an AT wrt F iff F ∪ {L1} ∪ . . . ∪ {Ln} UP ⊥
3. Linear resolvents are AT
4. CDCL-learned clauses are AT
4
Resolution Asymmetric Tautology
Järvisalo et al.: Inprocessing Rules. In: IJCAR (2012).
C is a RAT upon L wrt F, if
C is an AT wrt F, or
L ∈ C res(C, D, L) is an AT wrt F for every D ∈ F with ¬L ∈ D.
{p}, {¬q} are RATS wrt F:
1. {p} p is pure
2. {¬q} one resolvent {p} that is an AT
Lemma
1. F ≡sat F ∪ {C}, if C is RAT in F
2. All known formula simplifications can be characterized as RATs
5
Portfolios can be Described as
State Transition Systems
State transition system (∆, ;)
∆ is the set of states
; ⊆ ∆ × ∆ is the state transition relation.
Local state for Solveri :
working formula Fi
melted literals Mi
State ((M1, F1), . . . , (Mn, Fn)), SAT, UNSAT
Initial state for init(F0, n) = ((∅, F0), . . . , (∅, F0))
Final states SAT, UNSAT
Transition relation
6
SAT Termination Rule
(M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn)
SAT
some Fi is satisfiable
7
UNSAT Termination Rule
(M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn)
UNSAT
∅ ∈ F0
8
AT Rule
(M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn)
(M0, F0) . . . (Mi , Fi ∪ {C}) . . . (Mn, Fn)
C is an AT wrt Fi
9
RAT Rule
(M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn)
(M0, F0) . . . (Mi ∪ {L}, Fi ∪ {C}) . . . (Mn, Fn)
C is RAT upon L wrt Fi
10
Clause Deletion Rule
(M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn)
(M0, F0) . . . (Mi , Fi  {C}) . . . (Mn, Fn)
Fi ≡sat Fi  {C}
11
Clause Sharing Rule
(M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn)
(M0, F0) . . . (Mi , Fi ∪ {C} . . . (Mn, Fn)
C ∈ Fj and C ∩ Mi = ∅, C ∩ Mj = ∅
12
Soundness and Completeness
Soundness:
(i) if init(n, F0)
∗
; SAT, then F0 is satisfiable, and
(ii) if init(n, F0)
∗
; UNSAT, then F0 is unsatisfiable
Completeness:
(i) if F0 is satisfiable, then init(n, F0)
∗
; SAT
(ii) if F0 is unsatisfiable, then init(n, F0)
∗
; UNSAT
Theorem: The formal model is sound and complete
Proof idea: Semantical equivalences are preserved wrt a signature
13
Parallel DRAT
A conservative extension of DRAT
Parallel DRAT
Labeled clause ( , j, C), where ∈ {a, d}, j ∈ N
PDRAT derivation D = (Di | 1 ≤ i ≤ n) in F
if there is a run that is represented by D
PDRAT refutation D = (Di | 1 ≤ i ≤ n) in F is a PDRAT
derivation in F in which the empty clause occurs
Theorem:
1. F is unsatisfiable iff there is a PDRAT refutation in F
2. We can efficiently check PDRAT refutations:
AT > CS > RAT
14
PDRAT Proof Construction
AT Rule
(M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn)
(M0, F0) . . . (Mi , Fi ∪ {C}) . . . (Mn, Fn)
C is an AT wrt Fi
Corresponding proof (a, i, C)
16
RAT Rule
(M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn)
(M0, F0) . . . (Mi ∪ {L}, Fi ∪ {C}) . . . (Mn, Fn)
C is RAT upon L wrt Fi
Corresponding proof (a, i, C)
17
Clause Deletion Rule
(M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn)
(M0, F0) . . . (Mi , Fi  {C}) . . . (Mn, Fn)
Fi ≡sat Fi  {C}
Corresponding proof (d, i, C)
18
Clause Sharing Rule
(M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn)
(M0, F0) . . . (Mi , Fi ∪ {C} . . . (Mn, Fn)
C ∈ Fj and C ∩ Mi = ∅, C ∩ Mj = ∅
Corresponding proof (a, i, C)
19
Example
Formula
F0 = {{p, q, r}, {p, ¬q, r}, {¬p, q, r}, {¬p, ¬q, r}}
PDRAT derivation
(a, 1, C1), (a, 1, C2), (a, 1, C3), (a, 2, C3)
Corresponding run
init(2, F0) ;AT ((∅, F0 ∪ {C1}), (∅, F0))
;AT ((∅, F0 ∪ {C1, C2}), (∅, F0))
;AT ((∅, F0 ∪ {C1, C2, C3}), (∅, F0))
;CS ((∅, F0 ∪ {C1, C2, C3}), (∅, F0 ∪ {C3}))
20
Example
Formula
F0 = {{p, q, r}, {p, ¬q, r}, {¬p, q, r}, {¬p, ¬q, r}}
PDRAT derivation
(a, 1, C1), (a, 1, C2), (a, 1, C3), (a, 2, C3)
Corresponding run
init(2, F0) ;AT ((∅, F0 ∪ {C1}), (∅, F0))
;AT ((∅, F0 ∪ {C1, C2}), (∅, F0))
;AT ((∅, F0 ∪ {C1, C2, C3}), (∅, F0))
;RAT ((∅, F0 ∪ {C1, C2, C3}), ({q}, F0 ∪ {C3}))
20
Open Issues
Extend ManySAT,Plingeling, PRiss by PDRAT
Develop a backward-checking procedure for PDRAT
Develop a mechanically verified checker for PDRAT
A mathematical, consistent and verified theory for RAT
Is drat-trim sound and complete?
A mechanically-verified and efficient checker for DRAT
21
Conclusion
A formal model for parallel portfolios with
restricted clause sharing
arbritrary inprocessing
Parallel DRAT can be used to certify answers from these solvers
A prototypical checker implementation is available
https://iccl.inf.tu-dresden.de/web/PDRAT
22
Unsatisfiability Proofs for
Parallel SAT Solver Portfolios with
Clause Sharing and Inprocessing
Tobias Philipp
Knowledge Representation and Reasoning Group
Technische Universität Dresden
Thank you for your attention.
UNSAT can be Incorrect: Applying
Clause Addition Techniques in Two Solvers
F =
{{x, ¬y}, {¬x, y}, {x, z}, {y, ¬z}}
C = {¬x, ¬z} is RAT in F upon ¬z
D = {¬y, z} is RAT in F upon z
F is satisfiable:
I = {x, y, z}
J = {x, y, ¬z}
I |= C, J |= D
F ∪ {C, D} is unsatisfiable
F F
F ∧ C F
F ∧ C F ∧ D
F ∧ C ∧ D F ∧ D
UNSAT
23
Rule Preference
Lemma (Monotonicity) If
((M1, F1), . . . (Sn, Fn))
∗
; ((M1, F1), . . . (Mn, Fn))
and Ti ⊆ Si , then:
((T1, F1), . . . (Tn, Fn))
∗
; ((T1, F1), . . . (Tn, Fn))
and Ti ⊆ Si
Strategy Prefer AT-, and CS-rule over RAT-rule.
24

More Related Content

PDF
An Expressive Model for Instance Decomposition Based Parallel SAT Solvers
PDF
Spherical Geodesic
PDF
Rolle's theorem, mean value theorem
PDF
Math Module 4
PPT
Al2ed chapter18
PDF
ARIC Team Seminar
An Expressive Model for Instance Decomposition Based Parallel SAT Solvers
Spherical Geodesic
Rolle's theorem, mean value theorem
Math Module 4
Al2ed chapter18
ARIC Team Seminar

What's hot (19)

PDF
Ch04
PPT
Roll's theorem
PDF
Εφαπτομένη Ευθεία ΕΠΑΛ
PDF
Ch06
PDF
Aho corasick-lecture
PDF
Ch03
PPT
Boolean Algebra DLD
PDF
IntrRSCode
PPTX
Compiler First Set Follow Set Brief
PDF
Hecke Curves and Moduli spcaes of Vector Bundles
PDF
Enumeration of 2-level polytopes
PDF
Lesson 21: Derivatives and the Shapes of Curves
PPTX
Vertex cover Problem
PDF
Object recognition with pictorial structures
PPT
Chap8 new
PDF
Bc0052 theory of computer science-mqp
PDF
smtlecture.5
PDF
Lesson 21: Derivatives and the Shapes of Curves
Ch04
Roll's theorem
Εφαπτομένη Ευθεία ΕΠΑΛ
Ch06
Aho corasick-lecture
Ch03
Boolean Algebra DLD
IntrRSCode
Compiler First Set Follow Set Brief
Hecke Curves and Moduli spcaes of Vector Bundles
Enumeration of 2-level polytopes
Lesson 21: Derivatives and the Shapes of Curves
Vertex cover Problem
Object recognition with pictorial structures
Chap8 new
Bc0052 theory of computer science-mqp
smtlecture.5
Lesson 21: Derivatives and the Shapes of Curves
Ad

Similar to Unsatisfiability Proofs for Parallel SAT Solver Portfolios with Clause Sharing and Inprocessing (20)

PPTX
Theorem proving 2018 2019
PDF
Theorem proving 2018 2019
PPT
C2.0 propositional logic
PPTX
LECTURE SERIES TOC
PPTX
Cfg to pda equivalnce 3
PPTX
Microsoft PROSE SDK: A Framework for Inductive Program Synthesis
PPTX
Unit4_CL_Unit_4_on Computation Logic_srm
PDF
Sentient Arithmetic and Godel's Incompleteness Theorems
PDF
Declare Your Language: Dynamic Semantics
PDF
smtlecture.3
PDF
Backdoors to Satisfiability
PDF
Compiler Construction | Lecture 9 | Constraint Resolution
PPT
Dac07
PDF
Efficient Solving Techniques for Answer Set Programming
PDF
10 - 29 Jan - Recursion Part 2
PPTX
Jarrar: First Order Logic
PPT
Ch09.ppt dilfidsnpfpdsvklnvskv vdv;dlvd k;vsvdvk
PPTX
DBMS Unit 3.pptx
Theorem proving 2018 2019
Theorem proving 2018 2019
C2.0 propositional logic
LECTURE SERIES TOC
Cfg to pda equivalnce 3
Microsoft PROSE SDK: A Framework for Inductive Program Synthesis
Unit4_CL_Unit_4_on Computation Logic_srm
Sentient Arithmetic and Godel's Incompleteness Theorems
Declare Your Language: Dynamic Semantics
smtlecture.3
Backdoors to Satisfiability
Compiler Construction | Lecture 9 | Constraint Resolution
Dac07
Efficient Solving Techniques for Answer Set Programming
10 - 29 Jan - Recursion Part 2
Jarrar: First Order Logic
Ch09.ppt dilfidsnpfpdsvklnvskv vdv;dlvd k;vsvdvk
DBMS Unit 3.pptx
Ad

More from Tobias Philipp (8)

PDF
Fuzzing and Verifying RAT Refutations with Deletion Information
PDF
A Verified Decision Procedure for Pseudo-Boolean Formulas
PDF
PBLib - A Library for Encoding Pseudo-Boolean Constraints into CNF
PDF
The Complexity of Contextual Abduction in Human Reasoning Tasks
PDF
Checking Unsatisfiability Proofs in Parallel
PDF
Anwendungen der Logik in der IT-Sicherheit
PDF
Formal Verification with Ada/SPARK
PDF
Formale Verifikation von Answer Set Programming
Fuzzing and Verifying RAT Refutations with Deletion Information
A Verified Decision Procedure for Pseudo-Boolean Formulas
PBLib - A Library for Encoding Pseudo-Boolean Constraints into CNF
The Complexity of Contextual Abduction in Human Reasoning Tasks
Checking Unsatisfiability Proofs in Parallel
Anwendungen der Logik in der IT-Sicherheit
Formal Verification with Ada/SPARK
Formale Verifikation von Answer Set Programming

Recently uploaded (20)

PDF
Review of recent advances in non-invasive hemoglobin estimation
PDF
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
PPTX
sap open course for s4hana steps from ECC to s4
PDF
Optimiser vos workloads AI/ML sur Amazon EC2 et AWS Graviton
PDF
Profit Center Accounting in SAP S/4HANA, S4F28 Col11
PDF
How UI/UX Design Impacts User Retention in Mobile Apps.pdf
PDF
Network Security Unit 5.pdf for BCA BBA.
PDF
Diabetes mellitus diagnosis method based random forest with bat algorithm
PDF
cuic standard and advanced reporting.pdf
PDF
KodekX | Application Modernization Development
PPTX
Detection-First SIEM: Rule Types, Dashboards, and Threat-Informed Strategy
PDF
Mobile App Security Testing_ A Comprehensive Guide.pdf
PDF
The Rise and Fall of 3GPP – Time for a Sabbatical?
PPTX
Cloud computing and distributed systems.
PDF
TokAI - TikTok AI Agent : The First AI Application That Analyzes 10,000+ Vira...
PDF
Electronic commerce courselecture one. Pdf
PDF
NewMind AI Weekly Chronicles - August'25 Week I
PPTX
ACSFv1EN-58255 AWS Academy Cloud Security Foundations.pptx
PPTX
Big Data Technologies - Introduction.pptx
PPTX
Programs and apps: productivity, graphics, security and other tools
Review of recent advances in non-invasive hemoglobin estimation
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
sap open course for s4hana steps from ECC to s4
Optimiser vos workloads AI/ML sur Amazon EC2 et AWS Graviton
Profit Center Accounting in SAP S/4HANA, S4F28 Col11
How UI/UX Design Impacts User Retention in Mobile Apps.pdf
Network Security Unit 5.pdf for BCA BBA.
Diabetes mellitus diagnosis method based random forest with bat algorithm
cuic standard and advanced reporting.pdf
KodekX | Application Modernization Development
Detection-First SIEM: Rule Types, Dashboards, and Threat-Informed Strategy
Mobile App Security Testing_ A Comprehensive Guide.pdf
The Rise and Fall of 3GPP – Time for a Sabbatical?
Cloud computing and distributed systems.
TokAI - TikTok AI Agent : The First AI Application That Analyzes 10,000+ Vira...
Electronic commerce courselecture one. Pdf
NewMind AI Weekly Chronicles - August'25 Week I
ACSFv1EN-58255 AWS Academy Cloud Security Foundations.pptx
Big Data Technologies - Introduction.pptx
Programs and apps: productivity, graphics, security and other tools

Unsatisfiability Proofs for Parallel SAT Solver Portfolios with Clause Sharing and Inprocessing

  • 1. Unsatisfiability Proofs for Parallel SAT Solver Portfolios with Clause Sharing and Inprocessing Tobias Philipp International Center for Computational Logic Technische Universität Dresden
  • 3. The Portfolio Approach F0 Solver1 F0 Solver2 F0 Solver3 F0 Solver4 F0 F1 F2 F3 F4 SAT 1
  • 4. The Cooperative Portfolio Approach F0 Solver1 F0 Solver2 F0 Solver3 F0 Solver4 F0 F1 F2 F3 F4 SAT 2
  • 5. The Cooperative Portfolio Approach F0 Solver1 F0 Solver2 F0 Solver3 F0 Solver4 F0 F1 F2 F3 F4 SAT SAT 2
  • 6. Inprocessing in Clause Sharing SAT Solvers ppfolio Roussel, 2012 ManySAT Hamadi et al, 2009 PLingeling Biere, 2010 PLingeling Biere, 2013 Inprocessing Clause Sharing 3
  • 8. Asymmetric Tautologies Asymmetric literal addition alaF (C) = C ∪ {L | {L1, . . . , Ln, L} ∈ F and {L1, . . . , Ln} ⊆ C} Example F = {{p, q}, {p, ¬q, r}, {¬r, ¬q}} alaF ({p}) = {p, ¬q} C is an asymmetric tautology if there is n ∈ N st alan F (C) is a tautology Lemma 1. F ≡ F ∪ {C}, if C is an AT wrt F 2. {L1, . . . , Ln} is an AT wrt F iff F ∪ {L1} ∪ . . . ∪ {Ln} UP ⊥ 3. Linear resolvents are AT 4. CDCL-learned clauses are AT 4
  • 9. Resolution Asymmetric Tautology Järvisalo et al.: Inprocessing Rules. In: IJCAR (2012). C is a RAT upon L wrt F, if C is an AT wrt F, or L ∈ C res(C, D, L) is an AT wrt F for every D ∈ F with ¬L ∈ D. {p}, {¬q} are RATS wrt F: 1. {p} p is pure 2. {¬q} one resolvent {p} that is an AT Lemma 1. F ≡sat F ∪ {C}, if C is RAT in F 2. All known formula simplifications can be characterized as RATs 5
  • 10. Portfolios can be Described as State Transition Systems State transition system (∆, ;) ∆ is the set of states ; ⊆ ∆ × ∆ is the state transition relation. Local state for Solveri : working formula Fi melted literals Mi State ((M1, F1), . . . , (Mn, Fn)), SAT, UNSAT Initial state for init(F0, n) = ((∅, F0), . . . , (∅, F0)) Final states SAT, UNSAT Transition relation 6
  • 11. SAT Termination Rule (M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn) SAT some Fi is satisfiable 7
  • 12. UNSAT Termination Rule (M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn) UNSAT ∅ ∈ F0 8
  • 13. AT Rule (M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn) (M0, F0) . . . (Mi , Fi ∪ {C}) . . . (Mn, Fn) C is an AT wrt Fi 9
  • 14. RAT Rule (M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn) (M0, F0) . . . (Mi ∪ {L}, Fi ∪ {C}) . . . (Mn, Fn) C is RAT upon L wrt Fi 10
  • 15. Clause Deletion Rule (M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn) (M0, F0) . . . (Mi , Fi {C}) . . . (Mn, Fn) Fi ≡sat Fi {C} 11
  • 16. Clause Sharing Rule (M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn) (M0, F0) . . . (Mi , Fi ∪ {C} . . . (Mn, Fn) C ∈ Fj and C ∩ Mi = ∅, C ∩ Mj = ∅ 12
  • 17. Soundness and Completeness Soundness: (i) if init(n, F0) ∗ ; SAT, then F0 is satisfiable, and (ii) if init(n, F0) ∗ ; UNSAT, then F0 is unsatisfiable Completeness: (i) if F0 is satisfiable, then init(n, F0) ∗ ; SAT (ii) if F0 is unsatisfiable, then init(n, F0) ∗ ; UNSAT Theorem: The formal model is sound and complete Proof idea: Semantical equivalences are preserved wrt a signature 13
  • 18. Parallel DRAT A conservative extension of DRAT
  • 19. Parallel DRAT Labeled clause ( , j, C), where ∈ {a, d}, j ∈ N PDRAT derivation D = (Di | 1 ≤ i ≤ n) in F if there is a run that is represented by D PDRAT refutation D = (Di | 1 ≤ i ≤ n) in F is a PDRAT derivation in F in which the empty clause occurs Theorem: 1. F is unsatisfiable iff there is a PDRAT refutation in F 2. We can efficiently check PDRAT refutations: AT > CS > RAT 14
  • 21. AT Rule (M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn) (M0, F0) . . . (Mi , Fi ∪ {C}) . . . (Mn, Fn) C is an AT wrt Fi Corresponding proof (a, i, C) 16
  • 22. RAT Rule (M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn) (M0, F0) . . . (Mi ∪ {L}, Fi ∪ {C}) . . . (Mn, Fn) C is RAT upon L wrt Fi Corresponding proof (a, i, C) 17
  • 23. Clause Deletion Rule (M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn) (M0, F0) . . . (Mi , Fi {C}) . . . (Mn, Fn) Fi ≡sat Fi {C} Corresponding proof (d, i, C) 18
  • 24. Clause Sharing Rule (M0, F0) . . . (Mi , Fi ) . . . (Mn, Fn) (M0, F0) . . . (Mi , Fi ∪ {C} . . . (Mn, Fn) C ∈ Fj and C ∩ Mi = ∅, C ∩ Mj = ∅ Corresponding proof (a, i, C) 19
  • 25. Example Formula F0 = {{p, q, r}, {p, ¬q, r}, {¬p, q, r}, {¬p, ¬q, r}} PDRAT derivation (a, 1, C1), (a, 1, C2), (a, 1, C3), (a, 2, C3) Corresponding run init(2, F0) ;AT ((∅, F0 ∪ {C1}), (∅, F0)) ;AT ((∅, F0 ∪ {C1, C2}), (∅, F0)) ;AT ((∅, F0 ∪ {C1, C2, C3}), (∅, F0)) ;CS ((∅, F0 ∪ {C1, C2, C3}), (∅, F0 ∪ {C3})) 20
  • 26. Example Formula F0 = {{p, q, r}, {p, ¬q, r}, {¬p, q, r}, {¬p, ¬q, r}} PDRAT derivation (a, 1, C1), (a, 1, C2), (a, 1, C3), (a, 2, C3) Corresponding run init(2, F0) ;AT ((∅, F0 ∪ {C1}), (∅, F0)) ;AT ((∅, F0 ∪ {C1, C2}), (∅, F0)) ;AT ((∅, F0 ∪ {C1, C2, C3}), (∅, F0)) ;RAT ((∅, F0 ∪ {C1, C2, C3}), ({q}, F0 ∪ {C3})) 20
  • 27. Open Issues Extend ManySAT,Plingeling, PRiss by PDRAT Develop a backward-checking procedure for PDRAT Develop a mechanically verified checker for PDRAT A mathematical, consistent and verified theory for RAT Is drat-trim sound and complete? A mechanically-verified and efficient checker for DRAT 21
  • 28. Conclusion A formal model for parallel portfolios with restricted clause sharing arbritrary inprocessing Parallel DRAT can be used to certify answers from these solvers A prototypical checker implementation is available https://iccl.inf.tu-dresden.de/web/PDRAT 22
  • 29. Unsatisfiability Proofs for Parallel SAT Solver Portfolios with Clause Sharing and Inprocessing Tobias Philipp Knowledge Representation and Reasoning Group Technische Universität Dresden Thank you for your attention.
  • 30. UNSAT can be Incorrect: Applying Clause Addition Techniques in Two Solvers F = {{x, ¬y}, {¬x, y}, {x, z}, {y, ¬z}} C = {¬x, ¬z} is RAT in F upon ¬z D = {¬y, z} is RAT in F upon z F is satisfiable: I = {x, y, z} J = {x, y, ¬z} I |= C, J |= D F ∪ {C, D} is unsatisfiable F F F ∧ C F F ∧ C F ∧ D F ∧ C ∧ D F ∧ D UNSAT 23
  • 31. Rule Preference Lemma (Monotonicity) If ((M1, F1), . . . (Sn, Fn)) ∗ ; ((M1, F1), . . . (Mn, Fn)) and Ti ⊆ Si , then: ((T1, F1), . . . (Tn, Fn)) ∗ ; ((T1, F1), . . . (Tn, Fn)) and Ti ⊆ Si Strategy Prefer AT-, and CS-rule over RAT-rule. 24