Battam commited on
Commit
fa4e5c6
·
verified ·
1 Parent(s): b3f5983

Release release-2026-11-21.1, made by the table store from release-2026-11-20.1

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. campaign_files/2026-10-a-spreadsheets-with-answers-2/australia-government-workbooks-cc-by-3.0-au/LICENSE +5 -0
  2. campaign_files/2026-10-a-spreadsheets-with-answers-2/australia-government-workbooks-cc-by-3.0-au/README.md +50 -0
  3. campaign_files/2026-10-g-solver-instances-2/assembly_line_balancing_de/LICENSE +4 -0
  4. campaign_files/2026-10-g-solver-instances-2/assembly_line_balancing_de/README.md +36 -0
  5. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-dob38c_te2/LICENSE +3 -0
  6. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-dob38c_te2/README.md +23 -0
  7. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-mat18c_blue_te2/LICENSE +3 -0
  8. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-mat18c_blue_te2/README.md +23 -0
  9. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-mith_te3/LICENSE +3 -0
  10. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-mith_te3/README.md +23 -0
  11. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-monhard_te23/LICENSE +3 -0
  12. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-monhard_te23/README.md +23 -0
  13. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-paquita_te2/LICENSE +3 -0
  14. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-paquita_te2/README.md +23 -0
  15. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-ph2010_ser112plus/LICENSE +3 -0
  16. campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-ph2010_ser112plus/README.md +23 -0
  17. campaign_files/2026-10-g-solver-instances-2/epfl_combinational_benchmarks/LICENSE +3 -0
  18. campaign_files/2026-10-g-solver-instances-2/epfl_combinational_benchmarks/README.md +138 -0
  19. campaign_files/2026-10-g-solver-instances-2/fogleman-rush-hour-database/LICENSE +3 -0
  20. campaign_files/2026-10-g-solver-instances-2/fogleman-rush-hour-database/README.md +23 -0
  21. campaign_files/2026-10-g-solver-instances-2/github_jordipereiragude_dp_albp/LICENSE +7 -0
  22. campaign_files/2026-10-g-solver-instances-2/github_jordipereiragude_dp_albp/README.md +48 -0
  23. campaign_files/2026-10-g-solver-instances-2/github_jordipereiragude_workload_allocation_chains/LICENSE +5 -0
  24. campaign_files/2026-10-g-solver-instances-2/github_jordipereiragude_workload_allocation_chains/README.md +36 -0
  25. campaign_files/2026-10-g-solver-instances-2/github_lost-ufpb_clsp_pm_study/LICENSE +4 -0
  26. campaign_files/2026-10-g-solver-instances-2/github_lost-ufpb_clsp_pm_study/README.md +30 -0
  27. campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_cobot_albp2/LICENSE +3 -0
  28. campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_cobot_albp2/README.md +36 -0
  29. campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_ralsp/LICENSE +3 -0
  30. campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_ralsp/README.md +24 -0
  31. campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_two_sided_albp1/LICENSE +3 -0
  32. campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_two_sided_albp1/README.md +30 -0
  33. campaign_files/2026-10-g-solver-instances-2/iwls2022_ls_contest/LICENSE +3 -0
  34. campaign_files/2026-10-g-solver-instances-2/iwls2022_ls_contest/README.md +24 -0
  35. campaign_files/2026-10-g-solver-instances-2/iwls2023_ls_contest/LICENSE +3 -0
  36. campaign_files/2026-10-g-solver-instances-2/iwls2023_ls_contest/README.md +42 -0
  37. campaign_files/2026-10-g-solver-instances-2/iwls2024_ls_contest/LICENSE +3 -0
  38. campaign_files/2026-10-g-solver-instances-2/iwls2024_ls_contest/README.md +72 -0
  39. campaign_files/2026-10-g-solver-instances-2/iwls2025_ls_contest/LICENSE +3 -0
  40. campaign_files/2026-10-g-solver-instances-2/iwls2025_ls_contest/README.md +48 -0
  41. campaign_files/2026-10-g-solver-instances-2/jinnaiyuu-combinatorial_instances/LICENSE +3 -0
  42. campaign_files/2026-10-g-solver-instances-2/jinnaiyuu-combinatorial_instances/README.md +48 -0
  43. campaign_files/2026-10-g-solver-instances-2/kochetov_discrete_location_benchmarks/LICENSE +3 -0
  44. campaign_files/2026-10-g-solver-instances-2/kochetov_discrete_location_benchmarks/README.md +138 -0
  45. campaign_files/2026-10-g-solver-instances-2/lotsizingdatamodel_benchmarks/LICENSE +3 -0
  46. campaign_files/2026-10-g-solver-instances-2/lotsizingdatamodel_benchmarks/README.md +36 -0
  47. campaign_files/2026-10-g-solver-instances-2/neo_vrp_malaga/LICENSE +3 -0
  48. campaign_files/2026-10-g-solver-instances-2/neo_vrp_malaga/README.md +25 -0
  49. campaign_files/2026-10-g-solver-instances-2/rub_hgi_shorter_linear_slps_mds/LICENSE +3 -0
  50. campaign_files/2026-10-g-solver-instances-2/rub_hgi_shorter_linear_slps_mds/README.md +78 -0
campaign_files/2026-10-a-spreadsheets-with-answers-2/australia-government-workbooks-cc-by-3.0-au/LICENSE ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ Licence: NOASSERTION AND CC-BY-3.0-AU
2
+
3
+ - CC-BY-3.0-AU: https://creativecommons.org/licenses/by/3.0/au/ (publisher: Department of Finance)
4
+ - CC-BY-3.0-AU: https://creativecommons.org/licenses/by/3.0/au/ (publisher: Transport for NSW)
5
+ - CC-BY-3.0-AU: https://creativecommons.org/licenses/by/3.0/au/ (publisher: Victorian State Budget)
campaign_files/2026-10-a-spreadsheets-with-answers-2/australia-government-workbooks-cc-by-3.0-au/README.md ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # australia-government-workbooks-cc-by-3.0-au
2
+
3
+ **What this is:** Files published by Transport for NSW; Department of Finance; Office of the Australian Information Commissioner; Department of Primary Industries and Regional Development; Victorian State Budget.
4
+
5
+ **What one unit is:** One upstream file per unit. Units are kept under workbooks/.
6
+
7
+ **Where it comes from:** Upstream: https://data.gov.au. Kind: real files published upstream.
8
+
9
+ **Self-contained:** Some files are self-contained; the others are used together with other upstream files.
10
+
11
+ **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
12
+
13
+ **Licence (SPDX):** NOASSERTION AND CC-BY-3.0-AU
14
+
15
+ **Provenance:** https://data.gov.au
16
+
17
+ Files:
18
+
19
+ - `workbooks/au-tzp24-workforce-summary.xlsx`: Workforce Projections (TZP24 Workforce Summary).
20
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
21
+ - `workbooks/au-pbs-naa201516pbsexceltables.xlsx`: Budget 2015-16 and Portfolio Budget Statements (PBS) - Tables and Data (National Archives of Australia PBS 2015-16).
22
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
23
+ - `workbooks/au-pbs-comcare-2016-17-pbs.xlsx`: Budget 2016-17 and Portfolio Budget Statements (PBS) - Tables and Data (Comcare 2016-17 PBS.xlsx).
24
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
25
+ - `workbooks/au-pbs-nga-2016-17-pbs.xlsx`: Budget 2016-17 and Portfolio Budget Statements (PBS) - Tables and Data (NGA 2016-17 PBS.xlsx).
26
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
27
+ - `workbooks/au-pbs-acma-2016-17-pbs.xlsx`: Budget 2016-17 and Portfolio Budget Statements (PBS) - Tables and Data (ACMA 2016-17 PBS .xlsx).
28
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
29
+ - `workbooks/au-agency-foi-data-202425.xlsx`: Freedom of information statistics (Agency FOI data 2024-25).
30
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
31
+ - `workbooks/au-2019-published-1.xlsx`: Western Australia - Nominal Gross Regional Product (Nominal Gross Regional Product 2019).
32
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
33
+ - `workbooks/au-open-data-opal-distance-tables-2026-06.xlsx`: Opal Fares (Open Data Opal Distance Tables and Opal Fare Values).
34
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
35
+ - `workbooks/au-macroeconomic-indicators.xlsx`: State Budget 2017-18 - Macroeconomic Indicators (Macroeconomic indicators 2017-18).
36
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
37
+ - `workbooks/au-1-2-taxation-revenue-qtrly-gg-b17-18.xlsx`: State Budget 2017-18 - State Taxation Revenue (Quarterly) (State Taxation Revenue).
38
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
39
+ - `workbooks/au-macroeconomic-data-2024-25-budget-update.xlsx`: State Budget 2024-25 - Macroeconomic Indicators (Macroeconomic Data ).
40
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
41
+ - `workbooks/au-labour-force-and-unemployment-data.xlsx`: Western Australia Regional Labour Force and Employment (Regional Labour Force and Unemployment Data).
42
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
43
+ - `workbooks/au-regional-population-data.xlsx`: Western Australia Regional Population (Regional Population Data).
44
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
45
+ - `workbooks/au-nsw-trainlink-train-and-coach-patronage-jul-2021-to-may-2026.xlsx`: NSW TrainLink Regional Train And Coaches Statistics (NSW TrainLink Train And Coach Services).
46
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
47
+ - `workbooks/au-nsw-trainlink-train-and-coach-patronage-jan-2020-to-mar-2022.xlsx`: NSW TrainLink Regional Train And Coaches Statistics (NSW TrainLink Coach Services - Jan 2020 To Mar 2022).
48
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
49
+ - `workbooks/au-agency-foi-data-2023-24.xlsx`: Freedom of information statistics (Agency FOI data 2023-24.xlsx).
50
+ Tools: openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
campaign_files/2026-10-g-solver-instances-2/assembly_line_balancing_de/LICENSE ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://assembly-line-balancing.de/home/legal-notice/ (publisher: Armin Scholl and coauthors, assembly-line-balancing.de (Friedrich-Schiller-Universitaet Jena))
4
+ - NOASSERTION: https://assembly-line-balancing.de/home/legal-notice/ (publisher: Alena Otto, Christian Otto, Armin Scholl; assembly-line-balancing.de)
campaign_files/2026-10-g-solver-instances-2/assembly_line_balancing_de/README.md ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # assembly_line_balancing_de
2
+
3
+ **What this is:** Files published by Armin Scholl and coauthors, assembly-line-balancing.de (Friedrich-Schiller-Universitaet Jena); Alena Otto, Christian Otto, Armin Scholl; assembly-line-balancing.de; Armin Scholl, Nils Boysen, Malte Fliedner; assembly-line-balancing.de. Collection theme: Production planning and lot sizing instance libraries (lot sizing, capacity planning, assembly line balancing and sequencing models, with published optimal values or bounds).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. The format of each file is in the notes.
6
+
7
+ **Where it comes from:** Upstream: https://assembly-line-balancing.de/wp-content/uploads/2017/01/SALBP-data-sets.zip; https://assembly-line-balancing.de/wp-content/uploads/2017/01/SALBP_benchmark.zip; https://assembly-line-balancing.de/wp-content/uploads/2017/01/SDALBP-data-sets.zip. Kind: test or example files shipped with a tool. Classic benchmark data sets collected by Armin Scholl (Scholl 1993 report; Talbot et al. 1986, Hoffmann 1990 graphs), published by the authors on assembly-line-balancing.de; recorded tool_fixture; SALBP-1 data set generated with SALBPGen by Otto, Otto and Scholl (EJOR 228/1, 2013), published by the authors on assembly-line-balancing.de with SALOME-1 result tables; recorded tool_fixture; 16 SDALBP-1 data sets generated by Scholl, Boysen and Fliedner (OR Spectrum 30, 2008, 579-609; sdalbp page), published by the authors on assembly-line-balancing.de with SALOME-SDT1 solution files; recorded tool_fixture.
8
+
9
+ **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://assembly-line-balancing.de/wp-content/uploads/2017/01/SALBP-data-sets.zip https://assembly-line-balancing.de/wp-content/uploads/2017/01/SALBP_benchmark.zip https://assembly-line-balancing.de/wp-content/uploads/2017/01/SDALBP-data-sets.zip
16
+
17
+ Files:
18
+
19
+ - `wp-content/uploads/2017/01/SALBP-data-sets.zip`: SALBP benchmark data sets by Scholl (1993): 25 precedence graphs with SALBP-1, SALBP-2 and SALBP-E optimal or bounded results.
20
+ Format: zip (79 KB): precedence graphs/*.IN2 (ALB format: number of tasks, task times, precedence pairs, end marker -1,-1; format-ALB.pdf) plus precedence graphs/README.DOC and SALBP data sets.xlsx (sheets SALBP-1 data set, SALBP-2 data sets, SALBP-E data set); one case = one graph with one cycle time (SALBP-1) or one station count (SALBP-2) or one station range (SALBP-E)
21
+ Origin: classic benchmark data sets collected by Armin Scholl (Scholl 1993 report; Talbot et al. 1986, Hoffmann 1990 graphs), published by the authors on assembly-line-balancing.de; recorded tool_fixture
22
+ Check: Program: read the agent assignment task -> station; feasible when every task is assigned once, every precedence pair (i,j) of the .IN2 file has station(i) <= station(j), and every station load (sum of task times) is at most the cycle time; SALBP-1 passes when feasible and the number of used stations equals the published m* (fewer means a checker error, m* is proven); where only [LB,UB] is printed, pass when stations <= UB; SALBP-2 passes when the station count is the given m and the largest load equals c*
23
+ Companions (not included): wp-content/uploads/2017/01/format-ALB.pdf; wp-content/uploads/2017/01/Scholl-1993-ALBData.pdf; wp-content/uploads/2017/08/UALBP-1.data-set.xlsx; salbp/benchmark-data-sets-1993/index.html
24
+ Dependencies: no upstream code used (SALOME and BBR are code and were not run)
25
+ - `wp-content/uploads/2017/01/SALBP_benchmark.zip`: SALBP-1 benchmark data set by Otto et al. (2013): 6825 instances (n=20: 525, n=50: 525, n=50 permuted: 4725, n=100: 525, n=1000: 525 .alb files) and SALOME-1 optimum, LB and UB.
26
+ Format: zip (10.3 MB) holding small data set_n=20.zip, medium data set_n=50.zip, medium data set_n=50permuted.zip, large data set_n=100.zip, very large data set_n=1000.zip (ALB format .alb files, format-ALB.pdf) and one Details xlsx per size (sheet Summary); one case = one instance
27
+ Origin: SALBP-1 data set generated with SALBPGen by Otto, Otto and Scholl (EJOR 228/1, 2013), published by the authors on assembly-line-balancing.de with SALOME-1 result tables; recorded tool_fixture
28
+ Check: Program: feasible when every task sits on one station, precedence relations of the .alb file hold (station of a predecessor <= station of its successor) and every station load is at most the cycle time given in the file; pass when feasible and the station count is at most the upper bound in column I (equal to column G where the optimality flag J is 1); a count below the lower bound in column H means a checker error
29
+ Companions (not included): wp-content/uploads/2017/01/format-ALB.pdf; salbp/benchmark-data-sets-2013/index.html
30
+ Dependencies: SALOME and BBR not run
31
+ - `wp-content/uploads/2017/01/SDALBP-data-sets.zip`: SDALBP data sets: 4304 sequence-dependent assembly line balancing instances (.sdt) each with a SALOME-SDT1 solution file (.sol).
32
+ Format: zip (9.8 MB): folders <sr>_<ir>/ with <graph>_c=<cycle>_sr=<sr>_ir=<ir>.sdt (ALB format plus sequence-dependent time increments) and a .sol of the same name (sections <cycle time (given)>, <number m of stations found>, <optimality status>, <lower bound on m>, <upper bound on m>, <task assignments>), plus one pdf; one case = one .sdt with its .sol
33
+ Origin: 16 SDALBP-1 data sets generated by Scholl, Boysen and Fliedner (OR Spectrum 30, 2008, 579-609; sdalbp page), published by the authors on assembly-line-balancing.de with SALOME-SDT1 solution files; recorded tool_fixture
34
+ Check: Program: feasible when each task is on one station, precedence holds, and for each station the load computed with the sequence-dependent increments of the .sdt file for the agent order of tasks is at most the cycle time; pass when feasible and the station count is at most the <upper bound on m> of the .sol (equal to it where the status is optimal); the published assignment of the .sol is used to self-test the checker first (spec general rule 5)
35
+ Companions (not included): wp-content/uploads/2017/01/format-ALB.pdf; sdalbp/index.html
36
+ Dependencies: SALOME-SDT1 not run
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-dob38c_te2/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://raw.githubusercontent.com/denis-berthier/Sudoku-classif/15c9d217b08d2d3ebf0f147d4f7c0beb5c2fcdb2/README.md (publisher: Denis Berthier (puzzles: Mladen Dobrichev, [c38] in the README references))
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-dob38c_te2/README.md ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # berthier-sudoku-classif-dob38c_te2
2
+
3
+ **What this is:** Files published by Denis Berthier (puzzles: Mladen Dobrichev, [c38] in the README references). Collection theme: Public puzzle sets that ship with solutions (sudoku, nonogram, logic puzzle collections).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Plain text, 35 lines, LF: 81 puzzle characters, then an id after '#'; a case is one line, the puzzle is the first 81 characters.
6
+
7
+ **Where it comes from:** Upstream: github.com/denis-berthier/Sudoku-classif. Kind: real files published upstream. The T&E(2) subset of Mladen Dobrichev's 38-clue collection ([c38] in README section 7, section 5.3), republished by Denis Berthier.
8
+
9
+ **Self-contained:** The file is self-contained. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: No solver is needed at judge time.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://github.com/denis-berthier/Sudoku-classif
16
+
17
+ Files:
18
+
19
+ - `dob-2014078-38c/TE2-puzzles.txt`: The 35 T&E-depth 2 puzzles among Mladen Dobrichev's 2,014,078 minimal 38-clue sudoku puzzles, from Sudoku-classif.
20
+ Format: plain text, 35 lines, LF: 81 puzzle characters, then an id after '#'; a case is one line, the puzzle is the first 81 characters.
21
+ Origin: the T&E(2) subset of Mladen Dobrichev's 38-clue collection ([c38] in README section 7, section 5.3), republished by Denis Berthier
22
+ Check: No solver is needed at judge time.
23
+ Dependencies: not packed; path and commit filled when the script is committed, before the run; not packed
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-mat18c_blue_te2/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://raw.githubusercontent.com/denis-berthier/Sudoku-classif/15c9d217b08d2d3ebf0f147d4f7c0beb5c2fcdb2/README.md (publisher: Denis Berthier (puzzles: forum user Blue; folder of Jim White's 2,000,000 18-clue collection, [c18] in the README references))
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-mat18c_blue_te2/README.md ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # berthier-sudoku-classif-mat18c_blue_te2
2
+
3
+ **What this is:** Files published by Denis Berthier (puzzles: forum user Blue; folder of Jim White's 2,000,000 18-clue collection, [c18] in the README references). Collection theme: Public puzzle sets that ship with solutions (sudoku, nonogram, logic puzzle collections).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Plain text, 780 lines, LF: 81 puzzle characters, then the SER rating in brackets; a case is one line, the puzzle is the first 81 characters.
6
+
7
+ **Where it comes from:** Upstream: github.com/denis-berthier/Sudoku-classif. Kind: real files published upstream. 780 18-clue puzzles in T&E(2) due to the forum user Blue (README section 5.2), republished as-is by Denis Berthier.
8
+
9
+ **Self-contained:** The file is self-contained. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: No solver is needed at judge time.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://github.com/denis-berthier/Sudoku-classif
16
+
17
+ Files:
18
+
19
+ - `Mat-2000000-18c/TE2-780-Blue/puzzles.txt`: 780 18-clue sudoku puzzles at T&E-depth 2 found by Blue, from the Mat-2000000-18c folder of Sudoku-classif.
20
+ Format: plain text, 780 lines, LF: 81 puzzle characters, then the SER rating in brackets; a case is one line, the puzzle is the first 81 characters.
21
+ Origin: 780 18-clue puzzles in T&E(2) due to the forum user Blue (README section 5.2), republished as-is by Denis Berthier
22
+ Check: No solver is needed at judge time.
23
+ Dependencies: not packed; path and commit filled when the script is committed, before the run; not packed
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-mith_te3/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://raw.githubusercontent.com/denis-berthier/Sudoku-classif/15c9d217b08d2d3ebf0f147d4f7c0beb5c2fcdb2/README.md (publisher: Denis Berthier (puzzles: Philip Newman, alias mith, [mith-TE3] in the README references))
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-mith_te3/README.md ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # berthier-sudoku-classif-mith_te3
2
+
3
+ **What this is:** Files published by Denis Berthier (puzzles: Philip Newman, alias mith, [mith-TE3] in the README references). Collection theme: Public puzzle sets that ship with solutions (sudoku, nonogram, logic puzzle collections).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Plain text, 158,276 lines, LF: 81 puzzle characters followed by ';' separated numbers; a case is one line, the puzzle is the first 81 characters.
6
+
7
+ **Where it comes from:** Upstream: github.com/denis-berthier/Sudoku-classif. Kind: real files published upstream. Philip Newman's Nov 2022 collections of puzzles at T&E-depth 3 posted on the enjoysudoku forum ([mith-TE3] in README section 7: expanded forms, no minimals), republished as-is by Denis Berthier.
8
+
9
+ **Self-contained:** The file is self-contained. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: No solver is needed at judge time.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://github.com/denis-berthier/Sudoku-classif
16
+
17
+ Files:
18
+
19
+ - `mith-158276-TE3/puzzles.txt`: Philip Newman (mith) collection of 158,276 sudoku puzzles at T&E-depth 3 (Nov 2022), the hardest known class, as republished in Sudoku-classif.
20
+ Format: plain text, 158,276 lines, LF: 81 puzzle characters followed by ';' separated numbers; a case is one line, the puzzle is the first 81 characters.
21
+ Origin: Philip Newman's Nov 2022 collections of puzzles at T&E-depth 3 posted on the enjoysudoku forum ([mith-TE3] in README section 7: expanded forms, no minimals), republished as-is by Denis Berthier
22
+ Check: No solver is needed at judge time.
23
+ Dependencies: not packed; path and commit filled when the script is committed, before the run; not packed
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-monhard_te23/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://raw.githubusercontent.com/denis-berthier/Sudoku-classif/15c9d217b08d2d3ebf0f147d4f7c0beb5c2fcdb2/README.md (publisher: Denis Berthier (puzzles: Hendrik Monhard, [Monh] in the README references))
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-monhard_te23/README.md ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # berthier-sudoku-classif-monhard_te23
2
+
3
+ **What this is:** Files published by Denis Berthier (puzzles: Hendrik Monhard, [Monh] in the README references). Collection theme: Public puzzle sets that ship with solutions (sudoku, nonogram, logic puzzle collections).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Plain text, 55,088 lines, LF: 81 puzzle characters ('.' for an empty cell), then ratings and an id; a case is one line, the puzzle is the first 81 characters.
6
+
7
+ **Where it comes from:** Upstream: github.com/denis-berthier/Sudoku-classif. Kind: real files published upstream. Hendrik Monhard's Aug 2023 collection of minimal puzzles at T&E-depth 2 and 3, posted on the enjoysudoku forum ([Monh] in README section 7), republished as-is by Denis Berthier.
8
+
9
+ **Self-contained:** The file is self-contained. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: No solver is needed at judge time.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://github.com/denis-berthier/Sudoku-classif
16
+
17
+ Files:
18
+
19
+ - `Monhard-until-2023-08-15-TE23/puzzles.txt`: Hendrik Monhard's collection of 55,088 minimal sudoku puzzles at T&E-depth 2 and 3 (Aug 2023), as republished in Sudoku-classif.
20
+ Format: plain text, 55,088 lines, LF: 81 puzzle characters ('.' for an empty cell), then ratings and an id; a case is one line, the puzzle is the first 81 characters.
21
+ Origin: Hendrik Monhard's Aug 2023 collection of minimal puzzles at T&E-depth 2 and 3, posted on the enjoysudoku forum ([Monh] in README section 7), republished as-is by Denis Berthier
22
+ Check: No solver is needed at judge time.
23
+ Dependencies: not packed; path and commit filled when the script is committed, before the run; not packed
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-paquita_te2/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://raw.githubusercontent.com/denis-berthier/Sudoku-classif/15c9d217b08d2d3ebf0f147d4f7c0beb5c2fcdb2/README.md (publisher: Denis Berthier (puzzles: Franciska Ruessink, alias Paquita, [Paq-TE2] in the README references))
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-paquita_te2/README.md ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # berthier-sudoku-classif-paquita_te2
2
+
3
+ **What this is:** Files published by Denis Berthier (puzzles: Franciska Ruessink, alias Paquita, [Paq-TE2] in the README references). Collection theme: Public puzzle sets that ship with solutions (sudoku, nonogram, logic puzzle collections).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Plain text, 81,139 lines, LF: 81 puzzle characters, a space, ED=x/y/z (three difficulty ratings); a case is one line, the puzzle is the first 81 characters.
6
+
7
+ **Where it comes from:** Upstream: github.com/denis-berthier/Sudoku-classif. Kind: real files published upstream. Franciska Ruessink's collection of minimal T&E(2) puzzles ([Paq-TE2] in README section 7, shared on Google Drive), republished as-is by Denis Berthier.
8
+
9
+ **Self-contained:** The file is self-contained. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: No solver is needed at judge time.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://github.com/denis-berthier/Sudoku-classif
16
+
17
+ Files:
18
+
19
+ - `Paquita-2023-sept-dec-TE2/puzzles.txt`: Franciska Ruessink (Paquita) collection of 81,139 minimal sudoku puzzles at T&E-depth 2 (Sept to Dec 2023), as republished in Sudoku-classif.
20
+ Format: plain text, 81,139 lines, LF: 81 puzzle characters, a space, ED=x/y/z (three difficulty ratings); a case is one line, the puzzle is the first 81 characters.
21
+ Origin: Franciska Ruessink's collection of minimal T&E(2) puzzles ([Paq-TE2] in README section 7, shared on Google Drive), republished as-is by Denis Berthier
22
+ Check: No solver is needed at judge time.
23
+ Dependencies: not packed; path and commit filled when the script is committed, before the run; not packed
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-ph2010_ser112plus/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://raw.githubusercontent.com/denis-berthier/Sudoku-classif/15c9d217b08d2d3ebf0f147d4f7c0beb5c2fcdb2/README.md (publisher: Denis Berthier (puzzles: Gerard Penet, alias champagne, [ph2010] in the README references))
campaign_files/2026-10-g-solver-instances-2/berthier-sudoku-classif-ph2010_ser112plus/README.md ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # berthier-sudoku-classif-ph2010_ser112plus
2
+
3
+ **What this is:** Files published by Denis Berthier (puzzles: Gerard Penet, alias champagne, [ph2010] in the README references). Collection theme: Public puzzle sets that ship with solutions (sudoku, nonogram, logic puzzle collections).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Plain text, 47,795 lines, LF: 81 puzzle characters followed by ';' separated ratings and fields; a case is one line, the puzzle is the first 81 characters.
6
+
7
+ **Where it comes from:** Upstream: github.com/denis-berthier/Sudoku-classif. Kind: real files published upstream. The hardest-rated tier of Gerard Penet's database of 3,103,972 potentially hardest puzzles published on the enjoysudoku forum on 2020-10-14 ([ph2010] in README section 7), split by SER in Sudoku-classif.
8
+
9
+ **Self-contained:** The file is self-contained. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: No solver is needed at judge time.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://github.com/denis-berthier/Sudoku-classif
16
+
17
+ Files:
18
+
19
+ - `ph2010-3103972-TE2/PUZZLES/puzzles-11.9-11.2.txt`: SER 11.2 to 11.9 tier (47,795 puzzles) of Gerard Penet's ph2010 database of potentially hardest sudoku puzzles, as split in Sudoku-classif.
20
+ Format: plain text, 47,795 lines, LF: 81 puzzle characters followed by ';' separated ratings and fields; a case is one line, the puzzle is the first 81 characters.
21
+ Origin: the hardest-rated tier of Gerard Penet's database of 3,103,972 potentially hardest puzzles published on the enjoysudoku forum on 2020-10-14 ([ph2010] in README section 7), split by SER in Sudoku-classif
22
+ Check: No solver is needed at judge time.
23
+ Dependencies: not packed; path and commit filled when the script is committed, before the run; not packed
campaign_files/2026-10-g-solver-instances-2/epfl_combinational_benchmarks/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://github.com/lsils/benchmarks/blob/82d8cc6910419298e713a46644ed59fd3df53038/LICENSE (publisher: Integrated Systems Laboratory (LSI), EPFL)
campaign_files/2026-10-g-solver-instances-2/epfl_combinational_benchmarks/README.md ADDED
@@ -0,0 +1,138 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # epfl_combinational_benchmarks
2
+
3
+ **What this is:** Files published by Integrated Systems Laboratory (LSI), EPFL. Collection theme: Logic synthesis and optimisation benchmark instances (combinational circuits with best known node counts or levels: EPFL benchmarks, IWLS contests and the like).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. The format of each file is in the notes.
6
+
7
+ **Where it comes from:** Upstream: https://github.com/lsils/benchmarks. Kind: test or example files shipped with a tool. EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition).
8
+
9
+ **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/lsils/benchmarks
16
+
17
+ Files:
18
+
19
+ - `arithmetic/adder.aig`: EPFL combinational benchmark adder (arithmetic): AIG with 256 inputs, 129 outputs, 1020 AND nodes; best known LUT-6 size 129 LUTs, best known depth 5 levels.
20
+ Format: binary AIGER (.aig), header 'aig 1276 256 0 129 1020' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/adder_size_2022.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
21
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
22
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of arithmetic/adder.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 129. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/adder_size_2022.blif gives 129, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 5. Hide best_results/size/adder_size_2022.blif and best_results/README.md when posing.
23
+ Companions (not included): best_results/size/adder_size_2022.blif; best_results/README.md; README.md; LICENSE
24
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
25
+ - `arithmetic/bar.aig`: EPFL combinational benchmark bar (arithmetic): AIG with 135 inputs, 128 outputs, 3336 AND nodes; best known LUT-6 size 512 LUTs, best known depth 4 levels.
26
+ Format: binary AIGER (.aig), header 'aig 3471 135 0 128 3336' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/bar_size_2015.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
27
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
28
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of arithmetic/bar.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 512. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/bar_size_2015.blif gives 512, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 4. Hide best_results/size/bar_size_2015.blif and best_results/README.md when posing.
29
+ Companions (not included): best_results/size/bar_size_2015.blif; best_results/README.md; README.md; LICENSE
30
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
31
+ - `arithmetic/div.aig`: EPFL combinational benchmark div (arithmetic): AIG with 128 inputs, 128 outputs, 57247 AND nodes; best known LUT-6 size 3082 LUTs, best known depth 170 levels.
32
+ Format: binary AIGER (.aig), header 'aig 57375 128 0 128 57247' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/div_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
33
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
34
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of arithmetic/div.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 3082. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/div_size_2026.blif gives 3082, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 170. Hide best_results/size/div_size_2026.blif and best_results/README.md when posing.
35
+ Companions (not included): best_results/size/div_size_2026.blif; best_results/README.md; README.md; LICENSE
36
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
37
+ - `arithmetic/hyp.aig`: EPFL combinational benchmark hyp (arithmetic): AIG with 256 inputs, 128 outputs, 214335 AND nodes; best known LUT-6 size 36451 LUTs, best known depth 458 levels.
38
+ Format: binary AIGER (.aig), header 'aig 214591 256 0 128 214335' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/hyp_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
39
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
40
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of arithmetic/hyp.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 36451. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/hyp_size_2026.blif gives 36451, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 458. Hide best_results/size/hyp_size_2026.blif and best_results/README.md when posing.
41
+ Companions (not included): best_results/size/hyp_size_2026.blif; best_results/README.md; README.md; LICENSE
42
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
43
+ - `arithmetic/log2.aig`: EPFL combinational benchmark log2 (arithmetic): AIG with 32 inputs, 32 outputs, 32060 AND nodes; best known LUT-6 size 5986 LUTs, best known depth 51 levels.
44
+ Format: binary AIGER (.aig), header 'aig 32092 32 0 32 32060' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/log2_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
45
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
46
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of arithmetic/log2.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 5986. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/log2_size_2026.blif gives 5986, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 51. Hide best_results/size/log2_size_2026.blif and best_results/README.md when posing.
47
+ Companions (not included): best_results/size/log2_size_2026.blif; best_results/README.md; README.md; LICENSE
48
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
49
+ - `arithmetic/max.aig`: EPFL combinational benchmark max (arithmetic): AIG with 512 inputs, 130 outputs, 2865 AND nodes; best known LUT-6 size 511 LUTs, best known depth 6 levels.
50
+ Format: binary AIGER (.aig), header 'aig 3377 512 0 130 2865' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/max_size_2024.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
51
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
52
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of arithmetic/max.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 511. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/max_size_2024.blif gives 511, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 6. Hide best_results/size/max_size_2024.blif and best_results/README.md when posing.
53
+ Companions (not included): best_results/size/max_size_2024.blif; best_results/README.md; README.md; LICENSE
54
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
55
+ - `arithmetic/multiplier.aig`: EPFL combinational benchmark multiplier (arithmetic): AIG with 128 inputs, 128 outputs, 27062 AND nodes; best known LUT-6 size 4282 LUTs, best known depth 25 levels.
56
+ Format: binary AIGER (.aig), header 'aig 27190 128 0 128 27062' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/multiplier_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
57
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
58
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of arithmetic/multiplier.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 4282. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/multiplier_size_2026.blif gives 4282, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 25. Hide best_results/size/multiplier_size_2026.blif and best_results/README.md when posing.
59
+ Companions (not included): best_results/size/multiplier_size_2026.blif; best_results/README.md; README.md; LICENSE
60
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
61
+ - `arithmetic/sin.aig`: EPFL combinational benchmark sin (arithmetic): AIG with 24 inputs, 25 outputs, 5416 AND nodes; best known LUT-6 size 994 LUTs, best known depth 10 levels.
62
+ Format: binary AIGER (.aig), header 'aig 5440 24 0 25 5416' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/sin_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
63
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
64
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of arithmetic/sin.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 994. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/sin_size_2026.blif gives 994, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 10. Hide best_results/size/sin_size_2026.blif and best_results/README.md when posing.
65
+ Companions (not included): best_results/size/sin_size_2026.blif; best_results/README.md; README.md; LICENSE
66
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
67
+ - `arithmetic/sqrt.aig`: EPFL combinational benchmark sqrt (arithmetic): AIG with 128 inputs, 64 outputs, 24618 AND nodes; best known LUT-6 size 2958 LUTs, best known depth 150 levels.
68
+ Format: binary AIGER (.aig), header 'aig 24746 128 0 64 24618' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/sqrt_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
69
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
70
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of arithmetic/sqrt.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 2958. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/sqrt_size_2026.blif gives 2958, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 150. Hide best_results/size/sqrt_size_2026.blif and best_results/README.md when posing.
71
+ Companions (not included): best_results/size/sqrt_size_2026.blif; best_results/README.md; README.md; LICENSE
72
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
73
+ - `arithmetic/square.aig`: EPFL combinational benchmark square (arithmetic): AIG with 64 inputs, 128 outputs, 18484 AND nodes; best known LUT-6 size 2894 LUTs, best known depth 10 levels.
74
+ Format: binary AIGER (.aig), header 'aig 18548 64 0 128 18484' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/square_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
75
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
76
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of arithmetic/square.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 2894. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/square_size_2026.blif gives 2894, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 10. Hide best_results/size/square_size_2026.blif and best_results/README.md when posing.
77
+ Companions (not included): best_results/size/square_size_2026.blif; best_results/README.md; README.md; LICENSE
78
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
79
+ - `random_control/arbiter.aig`: EPFL combinational benchmark arbiter (random/control): AIG with 256 inputs, 129 outputs, 11839 AND nodes; best known LUT-6 size 259 LUTs, best known depth 5 levels.
80
+ Format: binary AIGER (.aig), header 'aig 12095 256 0 129 11839' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/arbiter_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
81
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
82
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of random_control/arbiter.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 259. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/arbiter_size_2026.blif gives 259, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 5. Hide best_results/size/arbiter_size_2026.blif and best_results/README.md when posing.
83
+ Companions (not included): best_results/size/arbiter_size_2026.blif; best_results/README.md; README.md; LICENSE
84
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
85
+ - `random_control/cavlc.aig`: EPFL combinational benchmark cavlc (random/control): AIG with 10 inputs, 11 outputs, 693 AND nodes; best known LUT-6 size 46 LUTs, best known depth 3 levels.
86
+ Format: binary AIGER (.aig), header 'aig 703 10 0 11 693' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/cavlc_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
87
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
88
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of random_control/cavlc.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 46. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/cavlc_size_2026.blif gives 46, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 3. Hide best_results/size/cavlc_size_2026.blif and best_results/README.md when posing.
89
+ Companions (not included): best_results/size/cavlc_size_2026.blif; best_results/README.md; README.md; LICENSE
90
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
91
+ - `random_control/ctrl.aig`: EPFL combinational benchmark ctrl (random/control): AIG with 7 inputs, 26 outputs, 174 AND nodes; best known LUT-6 size 25 LUTs, best known depth 2 levels.
92
+ Format: binary AIGER (.aig), header 'aig 181 7 0 26 174' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/ctrl_size_2023.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
93
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
94
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of random_control/ctrl.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 25. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/ctrl_size_2023.blif gives 25, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 2. Hide best_results/size/ctrl_size_2023.blif and best_results/README.md when posing.
95
+ Companions (not included): best_results/size/ctrl_size_2023.blif; best_results/README.md; README.md; LICENSE
96
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
97
+ - `random_control/dec.aig`: EPFL combinational benchmark dec (random/control): AIG with 8 inputs, 256 outputs, 304 AND nodes; best known LUT-6 size 264 LUTs, best known depth 2 levels.
98
+ Format: binary AIGER (.aig), header 'aig 312 8 0 256 304' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/dec_size_2018.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
99
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
100
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of random_control/dec.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 264. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/dec_size_2018.blif gives 264, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 2. Hide best_results/size/dec_size_2018.blif and best_results/README.md when posing.
101
+ Companions (not included): best_results/size/dec_size_2018.blif; best_results/README.md; README.md; LICENSE
102
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
103
+ - `random_control/i2c.aig`: EPFL combinational benchmark i2c (random/control): AIG with 147 inputs, 142 outputs, 1342 AND nodes; best known LUT-6 size 173 LUTs, best known depth 3 levels.
104
+ Format: binary AIGER (.aig), header 'aig 1489 147 0 142 1342' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/i2c_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
105
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
106
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of random_control/i2c.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 173. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/i2c_size_2026.blif gives 173, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 3. Hide best_results/size/i2c_size_2026.blif and best_results/README.md when posing.
107
+ Companions (not included): best_results/size/i2c_size_2026.blif; best_results/README.md; README.md; LICENSE
108
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
109
+ - `random_control/int2float.aig`: EPFL combinational benchmark int2float (random/control): AIG with 11 inputs, 7 outputs, 260 AND nodes; best known LUT-6 size 17 LUTs, best known depth 2 levels.
110
+ Format: binary AIGER (.aig), header 'aig 271 11 0 7 260' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/int2float_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
111
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
112
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of random_control/int2float.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 17. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/int2float_size_2026.blif gives 17, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 2. Hide best_results/size/int2float_size_2026.blif and best_results/README.md when posing.
113
+ Companions (not included): best_results/size/int2float_size_2026.blif; best_results/README.md; README.md; LICENSE
114
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
115
+ - `random_control/mem_ctrl.aig`: EPFL combinational benchmark mem_ctrl (random/control): AIG with 1204 inputs, 1231 outputs, 46836 AND nodes; best known LUT-6 size 1669 LUTs, best known depth 5 levels.
116
+ Format: binary AIGER (.aig), header 'aig 48040 1204 0 1231 46836' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/mem_ctrl_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
117
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
118
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of random_control/mem_ctrl.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 1669. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/mem_ctrl_size_2026.blif gives 1669, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 5. Hide best_results/size/mem_ctrl_size_2026.blif and best_results/README.md when posing.
119
+ Companions (not included): best_results/size/mem_ctrl_size_2026.blif; best_results/README.md; README.md; LICENSE
120
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
121
+ - `random_control/priority.aig`: EPFL combinational benchmark priority (random/control): AIG with 128 inputs, 8 outputs, 978 AND nodes; best known LUT-6 size 86 LUTs, best known depth 4 levels.
122
+ Format: binary AIGER (.aig), header 'aig 1106 128 0 8 978' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/priority_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
123
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
124
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of random_control/priority.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 86. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/priority_size_2026.blif gives 86, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 4. Hide best_results/size/priority_size_2026.blif and best_results/README.md when posing.
125
+ Companions (not included): best_results/size/priority_size_2026.blif; best_results/README.md; README.md; LICENSE
126
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
127
+ - `random_control/router.aig`: EPFL combinational benchmark router (random/control): AIG with 60 inputs, 30 outputs, 257 AND nodes; best known LUT-6 size 18 LUTs, best known depth 3 levels.
128
+ Format: binary AIGER (.aig), header 'aig 317 60 0 30 257' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/router_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
129
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
130
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of random_control/router.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 18. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/router_size_2026.blif gives 18, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 3. Hide best_results/size/router_size_2026.blif and best_results/README.md when posing.
131
+ Companions (not included): best_results/size/router_size_2026.blif; best_results/README.md; README.md; LICENSE
132
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
133
+ - `random_control/voter.aig`: EPFL combinational benchmark voter (random/control): AIG with 1001 inputs, 1 outputs, 13758 AND nodes; best known LUT-6 size 1145 LUTs, best known depth 11 levels.
134
+ Format: binary AIGER (.aig), header 'aig 14759 1001 0 1 13758' (max variable index, inputs, latches, outputs, AND nodes); one case per file; companions: best size BLIF best_results/size/voter_size_2026.blif (LUT-6 netlist, one .names per LUT), best_results/README.md (answer tables), README.md, LICENSE
135
+ Origin: EPFL Combinational Benchmark Suite (Amaru, Gaillardon, De Micheli, IWLS 2015), maintained by LSI EPFL in lsils/benchmarks with a yearly best-results competition; the circuits are intentionally sub-optimal (README.md line 7); recorded tool_fixture (spec 115 general rule 6: official instances of a standard suite or competition)
136
+ Check: Program: the agent returns a LUT-6 netlist in BLIF; pass when (1) ABC cec of random_control/voter.aig against the netlist reports the networks equivalent (the maintainers verify submissions by combinational equivalence checking, README.md line 62), (2) every .names node has at most 6 inputs, and (3) the LUT count, counted as .names nodes with 2 to 6 inputs (constants and single-input buffers or inverters not counted), is not above 1145. This counting rule reproduces all 20 size records of best_results/README.md exactly from the reference BLIFs (best_results/size/voter_size_2026.blif gives 1145, COMPUTED this round). Depth variant: the number of LUT levels on the longest input-to-output path is not above 11. Hide best_results/size/voter_size_2026.blif and best_results/README.md when posing.
137
+ Companions (not included): best_results/size/voter_size_2026.blif; best_results/README.md; README.md; LICENSE
138
+ Dependencies: ABC (berkeley-abc/abc, commit a3001b72edc5de22442e942165487fddf150e3d0 of 2026-10-05, UC Regents permissive licence in copyright.txt), used only for its cec command; not packed
campaign_files/2026-10-g-solver-instances-2/fogleman-rush-hour-database/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://www.michaelfogleman.com/rush/ (publisher: Michael Fogleman)
campaign_files/2026-10-g-solver-instances-2/fogleman-rush-hour-database/README.md ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # fogleman-rush-hour-database
2
+
3
+ **What this is:** Files published by Michael Fogleman. Collection theme: Public puzzle sets that ship with solutions (sudoku, nonogram, logic puzzle collections).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. A file whose upstream path has no folder is kept in a folder named after its unit. Format: Plain text, 1000 lines: moves board36 cluster_size (36 characters row-major, o empty, x wall, A the primary piece); delivered whole as the upstream file; a case is one line, split on LF.
6
+
7
+ **Where it comes from:** Upstream: https://www.michaelfogleman.com/static/rush/rush1000.txt. Kind: real files published upstream. Rush Hour position database published as-is by Michael Fogleman, made by his exhaustive enumeration program with recorded optimal move counts (was procedurally_generated).
8
+
9
+ **Self-contained:** The file is self-contained. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: Accept when legal; score length against the recorded optimum. No solver is needed at judge time.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://www.michaelfogleman.com/static/rush/rush1000.txt
16
+
17
+ Files:
18
+
19
+ - `g023-02/rush1000.txt`: Rush Hour hardest-1000 sample from the complete 6x6 Rush Hour database, with optimal move counts.
20
+ Format: plain text, 1000 lines: moves board36 cluster_size (36 characters row-major, o empty, x wall, A the primary piece); delivered whole as the upstream file; a case is one line, split on LF
21
+ Origin: Rush Hour position database published as-is by Michael Fogleman, made by his exhaustive enumeration program with recorded optimal move counts (was procedurally_generated)
22
+ Check: accept when legal; score length against the recorded optimum. No solver is needed at judge time.
23
+ Dependencies: not packed
campaign_files/2026-10-g-solver-instances-2/github_jordipereiragude_dp_albp/LICENSE ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://raw.githubusercontent.com/jordipereiragude/dp_albp/0ec45eb1bdaf85daa5c0ac2ecb7e3868cd736440/instances/ralbp/025_003_roszieg.ralbp (publisher: Jordi Pereira (GitHub jordipereiragude))
4
+ - NOASSERTION: https://raw.githubusercontent.com/jordipereiragude/dp_albp/0ec45eb1bdaf85daa5c0ac2ecb7e3868cd736440/instances/tsalbp/P148_4.talbp (publisher: Jordi Pereira (GitHub jordipereiragude))
5
+ - NOASSERTION: https://raw.githubusercontent.com/jordipereiragude/dp_albp/0ec45eb1bdaf85daa5c0ac2ecb7e3868cd736440/instances/sualbsp/025/arc111.10027.sualbsp (publisher: Jordi Pereira (GitHub jordipereiragude))
6
+ - NOASSERTION: https://raw.githubusercontent.com/jordipereiragude/dp_albp/0ec45eb1bdaf85daa5c0ac2ecb7e3868cd736440/instances/lr-albp/mertens.1.7.in2 (publisher: Jordi Pereira (GitHub jordipereiragude))
7
+ - NOASSERTION: https://raw.githubusercontent.com/jordipereiragude/dp_albp/0ec45eb1bdaf85daa5c0ac2ecb7e3868cd736440/instances/sr-albp/sawyer30-1-0.9.ssalbp (publisher: Jordi Pereira (GitHub jordipereiragude))
campaign_files/2026-10-g-solver-instances-2/github_jordipereiragude_dp_albp/README.md ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # github_jordipereiragude_dp_albp
2
+
3
+ **What this is:** Files published by Jordi Pereira (GitHub jordipereiragude). Collection theme: Production planning and lot sizing instance libraries (lot sizing, capacity planning, assembly line balancing and scheduling models, with published optimal values or bounds).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Text, xml-like tags in the Otto, Otto and Scholl (2013) scheme (README.md: "All these instances follow a xml-like scheme"): <number of tasks>, <number of workstations>, <number of robots>, <task times> as task robot time triples, <precedence relations> as i,j pairs; text, xml-like tags in the Otto, Otto and Scholl (2013) scheme (README.md: "All these instances follow a xml-like scheme"): <number of tasks>, <mated-station number>, <task times>, task side codes L, R or E, <precedence relations>; text, xml-like tags in the Otto, Otto and Scholl (2013) scheme (README.md: "All these instances follow a xml-like scheme"): <number of tasks>, <number of workstations>, <task times>, <precedence relations>, forward and backward setup times, and <optimal SALBP-1 value>; text, xml-like tags in the Otto, Otto and Scholl (2013) scheme (README.md: "All these instances follow a xml-like scheme"): <number of tasks>, <cycle time>, <confidence level>, <task times> as task, mean, standard deviation triples (the third column is 0.1 times the mean, cv 0.1), <precedence relations>; text, xml-like tags in the Otto, Otto and Scholl (2013) scheme (README.md: "All these instances follow a xml-like scheme"): <number of tasks>, <number of workstations>, <confidence level>, <task times> as task, mean, standard deviation triples (third column 0.1 times the mean), <precedence relations>.
6
+
7
+ **Where it comes from:** Upstream: github.com/jordipereiragude/dp_albp. Kind: test or example files shipped with a tool. Instances and per-instance literature results of the repository for the paper "A decomposition-based approach for assembly line balancing problems" (README.md line 1: "Code, instances and results for ..."; README.md: "There are six excel files with per-instance results from the literature"); recorded tool_fixture (spec 115 general rule 6).
8
+
9
+ **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://github.com/jordipereiragude/dp_albp
16
+
17
+ Files:
18
+
19
+ - `instances/ralbp/025_003_roszieg.ralbp`: Robotic assembly line balancing type 2 (RALBP-2) instance Roszieg, 25 tasks, 3 workstations, 3 robot types, with best published cycle time 503.
20
+ Format: text, xml-like tags in the Otto, Otto and Scholl (2013) scheme (README.md: "All these instances follow a xml-like scheme"): <number of tasks>, <number of workstations>, <number of robots>, <task times> as task robot time triples, <precedence relations> as i,j pairs
21
+ Origin: instances and per-instance literature results of the repository for the paper "A decomposition-based approach for assembly line balancing problems" (README.md line 1: "Code, instances and results for ..."; README.md: "There are six excel files with per-instance results from the literature"); recorded tool_fixture (spec 115 general rule 6)
22
+ Check: Program: read the agent assignment of each task to a station 1..3 and a robot type per station; feasible when every task is assigned once, precedence i,j has station(i) <= station(j); station time = sum of the times of its tasks on the robot of that station (each robot type used at most once, per the RALBP-2 model of Gao et al. 2009); cycle time = largest station time; passes when feasible and cycle time <= 503
23
+ Companions (not included): results/ralbp2.xlsx; README.md
24
+ Dependencies: no upstream code used
25
+ - `instances/tsalbp/P148_4.talbp`: Two-sided assembly line balancing type 2 (TALBP-2) instance P148 with 4 mated stations, proven optimal cycle time 641 (equals the lower bound).
26
+ Format: text, xml-like tags in the Otto, Otto and Scholl (2013) scheme (README.md: "All these instances follow a xml-like scheme"): <number of tasks>, <mated-station number>, <task times>, task side codes L, R or E, <precedence relations>
27
+ Origin: instances and per-instance literature results of the repository for the paper "A decomposition-based approach for assembly line balancing problems" (README.md line 1: "Code, instances and results for ..."; README.md: "There are six excel files with per-instance results from the literature"); recorded tool_fixture (spec 115 general rule 6)
28
+ Check: Program: read the agent assignment of each task to a mated station 1..4 and a side; feasible when L tasks sit on the left, R tasks on the right, E tasks on either, and every precedence pair respects station order with start times on a side (two-sided model: a task starts after its predecessors finish, also across the two sides of a mated station); cycle time = largest side finish time; passes when feasible and cycle time <= 641
29
+ Companions (not included): results/talbp2.xlsx; README.md
30
+ Dependencies: no upstream code used
31
+ - `instances/sualbsp/025/arc111.10027.sualbsp`: Assembly line balancing with sequence-dependent setup times type 2 (SUALBSP-2) instance Arcus111 (111 tasks, 15 workstations, setup variability alpha 0.25) with best published cycle time 10448.
32
+ Format: text, xml-like tags in the Otto, Otto and Scholl (2013) scheme (README.md: "All these instances follow a xml-like scheme"): <number of tasks>, <number of workstations>, <task times>, <precedence relations>, forward and backward setup times, and <optimal SALBP-1 value>
33
+ Origin: instances and per-instance literature results of the repository for the paper "A decomposition-based approach for assembly line balancing problems" (README.md line 1: "Code, instances and results for ..."; README.md: "There are six excel files with per-instance results from the literature"); recorded tool_fixture (spec 115 general rule 6)
34
+ Check: Program: read the agent assignment of tasks to 15 stations and the task order within each station; feasible when every task is assigned once and precedence holds across and within stations; station time = task times + forward setups between consecutive tasks + the backward setup from the last task to the first; cycle time = largest station time; passes when feasible and cycle time <= 10448
35
+ Companions (not included): results/sualbsp2.xlsx; README.md
36
+ Dependencies: no upstream code used
37
+ - `instances/lr-albp/mertens.1.7.in2`: Line-reliability chance-constrained assembly line balancing (LR-SALBP) instance Mertens, 7 tasks with normal task times (cv 0.1), cycle time 7, confidence 0.95, best published station count 6.
38
+ Format: text, xml-like tags in the Otto, Otto and Scholl (2013) scheme (README.md: "All these instances follow a xml-like scheme"): <number of tasks>, <cycle time>, <confidence level>, <task times> as task, mean, standard deviation triples (the third column is 0.1 times the mean, cv 0.1), <precedence relations>
39
+ Origin: instances and per-instance literature results of the repository for the paper "A decomposition-based approach for assembly line balancing problems" (README.md line 1: "Code, instances and results for ..."; README.md: "There are six excel files with per-instance results from the literature"); recorded tool_fixture (spec 115 general rule 6)
40
+ Check: Program: read the agent assignment of tasks to stations; feasible when every task is assigned once, precedence holds, and the product over stations of P(sum of normal task times <= 7) is at least 0.95 (independent normal times with the listed mean and standard deviation); passes when feasible and station count <= 6
41
+ Companions (not included): results/lr-salbp.xlsx; README.md
42
+ Dependencies: no upstream code used
43
+ - `instances/sr-albp/sawyer30-1-0.9.ssalbp`: Chance-constrained stochastic assembly line balancing type 2 (SR-ALBP) instance Sawyer30, 30 tasks, 6 workstations, cv 0.1, confidence 0.9, best published cycle time 57.78.
44
+ Format: text, xml-like tags in the Otto, Otto and Scholl (2013) scheme (README.md: "All these instances follow a xml-like scheme"): <number of tasks>, <number of workstations>, <confidence level>, <task times> as task, mean, standard deviation triples (third column 0.1 times the mean), <precedence relations>
45
+ Origin: instances and per-instance literature results of the repository for the paper "A decomposition-based approach for assembly line balancing problems" (README.md line 1: "Code, instances and results for ..."; README.md: "There are six excel files with per-instance results from the literature"); recorded tool_fixture (spec 115 general rule 6)
46
+ Check: Program: read the agent assignment of the 30 tasks to 6 stations and its claimed cycle time c; feasible when every task is assigned once, precedence holds and for every station mean + z(0.9) * sqrt(sum of variances) <= c (independent normal task times); passes when feasible and c <= 57.78 + 0.005
47
+ Companions (not included): results/ssalbp.xlsx; README.md
48
+ Dependencies: no upstream code used
campaign_files/2026-10-g-solver-instances-2/github_jordipereiragude_workload_allocation_chains/LICENSE ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://raw.githubusercontent.com/jordipereiragude/workload_allocation_chains/5764562b8b1af62e48e6020bfaa9c9c99ef18960/dat/Battarra%2Cetal/AP001.txt (publisher: Jordi Pereira and Marcus Ritt (jordipereiragude))
4
+ - NOASSERTION: https://raw.githubusercontent.com/jordipereiragude/workload_allocation_chains/5764562b8b1af62e48e6020bfaa9c9c99ef18960/dat/setII/ap-16-46-0.txt (publisher: Jordi Pereira and Marcus Ritt (jordipereiragude))
5
+ - NOASSERTION: https://raw.githubusercontent.com/jordipereiragude/workload_allocation_chains/5764562b8b1af62e48e6020bfaa9c9c99ef18960/dat/setIII/ap-23-73-2.txt (publisher: Jordi Pereira and Marcus Ritt (jordipereiragude))
campaign_files/2026-10-g-solver-instances-2/github_jordipereiragude_workload_allocation_chains/README.md ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # github_jordipereiragude_workload_allocation_chains
2
+
3
+ **What this is:** Files published by Jordi Pereira and Marcus Ritt (jordipereiragude). Collection theme: Production planning and lot sizing instance libraries (lot sizing, capacity planning, assembly line balancing and scheduling models, with published optimal values or bounds).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Text: line 1 number of operations, line 2 number of workers, then "Standard production times" (one row) and "Production times: workers row and ops columns" (a workers by operations matrix).
6
+
7
+ **Where it comes from:** Upstream: github.com/jordipereiragude/workload_allocation_chains. Kind: test or example files shipped with a tool. Instances and test logs of the repository for Pereira and Ritt, "Exact and heuristic methods for a workload allocation problem with chained precedence constraints", EJOR 2023, doi 10.1016/j.ejor.2022.12.035 (README.md lines 1 to 3); set of Battarra et al. (2020), 200 instances AP001 to AP200 in dat/Battarra,etal; recorded tool_fixture (spec 115 general rule 6); instances and test logs of the repository for Pereira and Ritt, "Exact and heuristic methods for a workload allocation problem with chained precedence constraints", EJOR 2023, doi 10.1016/j.ejor.2022.12.035 (README.md lines 1 to 3); set II of the paper, 200 instances in dat/setII; recorded tool_fixture (spec 115 general rule 6); instances and test logs of the repository for Pereira and Ritt, "Exact and heuristic methods for a workload allocation problem with chained precedence constraints", EJOR 2023, doi 10.1016/j.ejor.2022.12.035 (README.md lines 1 to 3); set III of the paper, 200 instances in dat/setIII; recorded tool_fixture (spec 115 general rule 6).
8
+
9
+ **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://github.com/jordipereiragude/workload_allocation_chains
16
+
17
+ Files:
18
+
19
+ - `dat/Battarra,etal/AP001.txt`: Workload allocation with chained precedence, Battarra et al. (2020) instance AP001 (18 operations, 15 workers) with proven optimal cycle time 13.44685.
20
+ Format: text: line 1 number of operations, line 2 number of workers, then "Standard production times" (one row) and "Production times: workers row and ops columns" (a workers by operations matrix)
21
+ Origin: instances and test logs of the repository for Pereira and Ritt, "Exact and heuristic methods for a workload allocation problem with chained precedence constraints", EJOR 2023, doi 10.1016/j.ejor.2022.12.035 (README.md lines 1 to 3); set of Battarra et al. (2020), 200 instances AP001 to AP200 in dat/Battarra,etal; recorded tool_fixture (spec 115 general rule 6)
22
+ Check: Program: read the agent allocation of the chain of operations to workers; feasible when every operation is assigned once, each worker gets one block of consecutive operations in chain order and each worker is used at most once; load of a worker = sum of its production times for its operations (matrix row of that worker); cycle time = largest load; passes when feasible and cycle time <= 13.44685 + 0.000005
23
+ Companions (not included): test/test030/solver.log; dat/Battarra,etal.dat; README.md
24
+ Dependencies: no upstream code used
25
+ - `dat/setII/ap-16-46-0.txt`: Workload allocation with chained precedence, set II instance ap-16-46-0 (46 operations, 16 workers) with proven optimal cycle time 27.75099.
26
+ Format: text: line 1 number of operations, line 2 number of workers, then "Standard production times" (one row) and "Production times: workers row and ops columns" (a workers by operations matrix)
27
+ Origin: instances and test logs of the repository for Pereira and Ritt, "Exact and heuristic methods for a workload allocation problem with chained precedence constraints", EJOR 2023, doi 10.1016/j.ejor.2022.12.035 (README.md lines 1 to 3); set II of the paper, 200 instances in dat/setII; recorded tool_fixture (spec 115 general rule 6)
28
+ Check: Program: read the agent allocation of the chain of operations to workers; feasible when every operation is assigned once, each worker gets one block of consecutive operations in chain order and each worker is used at most once; load of a worker = sum of its production times for its operations (matrix row of that worker); cycle time = largest load; passes when feasible and cycle time <= 27.75099 + 0.000005
29
+ Companions (not included): test/test034/solver.log; README.md
30
+ Dependencies: no upstream code used
31
+ - `dat/setIII/ap-23-73-2.txt`: Workload allocation with chained precedence, set III instance ap-23-73-2 (73 operations, 23 workers) with proven optimal cycle time 23.26103.
32
+ Format: text: line 1 number of operations, line 2 number of workers, then "Standard production times" (one row) and "Production times: workers row and ops columns" (a workers by operations matrix)
33
+ Origin: instances and test logs of the repository for Pereira and Ritt, "Exact and heuristic methods for a workload allocation problem with chained precedence constraints", EJOR 2023, doi 10.1016/j.ejor.2022.12.035 (README.md lines 1 to 3); set III of the paper, 200 instances in dat/setIII; recorded tool_fixture (spec 115 general rule 6)
34
+ Check: Program: read the agent allocation of the chain of operations to workers; feasible when every operation is assigned once, each worker gets one block of consecutive operations in chain order and each worker is used at most once; load of a worker = sum of its production times for its operations (matrix row of that worker); cycle time = largest load; passes when feasible and cycle time <= 23.26103 + 0.000005
35
+ Companions (not included): test/test104/solver.log; README.md
36
+ Dependencies: no upstream code used
campaign_files/2026-10-g-solver-instances-2/github_lost-ufpb_clsp_pm_study/LICENSE ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://raw.githubusercontent.com/lost-ufpb/clsp_pm_study/aed921a6b6f0206165b22866eec52b20d76285a6/1.instances/2-6/AAA0/AAA006.dat (publisher: LOST group, Universidade Federal da Paraiba (lost-ufpb))
4
+ - NOASSERTION: https://raw.githubusercontent.com/lost-ufpb/clsp_pm_study/aed921a6b6f0206165b22866eec52b20d76285a6/1.instances/2-6/NBB0/NBB006.dat (publisher: LOST group, Universidade Federal da Paraiba (lost-ufpb))
campaign_files/2026-10-g-solver-instances-2/github_lost-ufpb_clsp_pm_study/README.md ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # github_lost-ufpb_clsp_pm_study
2
+
3
+ **What this is:** Files published by LOST group, Universidade Federal da Paraiba (lost-ufpb). Collection theme: Production planning and lot sizing instance libraries (lot sizing, capacity planning, assembly line balancing and sequencing models, with published optimal values or bounds).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Text (format in 1.instances/readme.txt): items and periods, machines, capacity per machine, per item and machine production time, setup time, setup cost, production cost; inventory cost per item; demand per period and item.
6
+
7
+ **Where it comes from:** Upstream: github.com/lost-ufpb/clsp_pm_study. Kind: test or example files shipped with a tool. Instances of Toledo and Armentano (2006) (README.md line 20) with MILP results (600 s) of the paper "Combining Relax-and-Fix and Fix-and-Optimize Approaches to Solve the Multi-Item Capacitated Lot-Sizing Problem with Unrelated Parallel Machines" (README.md line 5); recorded tool_fixture (spec 115 general rule 6).
8
+
9
+ **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: Program: read the agent lot sizes x[i,m,t] and setups y[i,m,t]; feasible when demand of every item and period is met from production and inventory without backlog, x>0 only with a setup, and per machine and period production time plus setup time <= capacity; total cost = setup + production + inventory cost; passes when feasible and cost <= 14136.52 + 0.005; Program: read the agent lot sizes x[i,m,t] and setups y[i,m,t]; feasible when demand of every item and period is met from production and inventory without backlog, x>0 only with a setup, and per machine and period production time plus setup time <= capacity; total cost = setup + production + inventory cost; passes when feasible and cost <= 8081.53 + 0.005.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://github.com/lost-ufpb/clsp_pm_study
16
+
17
+ Files:
18
+
19
+ - `1.instances/2-6/AAA0/AAA006.dat`: Capacitated lot sizing with parallel machines instance AAA006 (6 items, 2 machines, 6 periods, tight capacity, high setup cost, high setup time) with MILP-proven optimal cost 14136.52.
20
+ Format: text (format in 1.instances/readme.txt): items and periods, machines, capacity per machine, per item and machine production time, setup time, setup cost, production cost; inventory cost per item; demand per period and item
21
+ Origin: instances of Toledo and Armentano (2006) (README.md line 20) with MILP results (600 s) of the paper "Combining Relax-and-Fix and Fix-and-Optimize Approaches to Solve the Multi-Item Capacitated Lot-Sizing Problem with Unrelated Parallel Machines" (README.md line 5); recorded tool_fixture (spec 115 general rule 6)
22
+ Check: Program: read the agent lot sizes x[i,m,t] and setups y[i,m,t]; feasible when demand of every item and period is met from production and inventory without backlog, x>0 only with a setup, and per machine and period production time plus setup time <= capacity; total cost = setup + production + inventory cost; passes when feasible and cost <= 14136.52 + 0.005
23
+ Companions (not included): 2.detailed_results/MIP_600/TCHSHT_MIP_600s.CSV; 1.instances/readme.txt; README.md
24
+ Dependencies: no upstream code used
25
+ - `1.instances/2-6/NBB0/NBB006.dat`: Capacitated lot sizing with parallel machines instance NBB006 (6 items, 2 machines, 6 periods, normal capacity, low setup cost, low setup time) with MILP-proven optimal cost 8081.53.
26
+ Format: text (format in 1.instances/readme.txt): items and periods, machines, capacity per machine, per item and machine production time, setup time, setup cost, production cost; inventory cost per item; demand per period and item
27
+ Origin: instances of Toledo and Armentano (2006) (README.md line 20) with MILP results (600 s) of the paper "Combining Relax-and-Fix and Fix-and-Optimize Approaches to Solve the Multi-Item Capacitated Lot-Sizing Problem with Unrelated Parallel Machines" (README.md line 5); recorded tool_fixture (spec 115 general rule 6)
28
+ Check: Program: read the agent lot sizes x[i,m,t] and setups y[i,m,t]; feasible when demand of every item and period is met from production and inventory without backlog, x>0 only with a setup, and per machine and period production time plus setup time <= capacity; total cost = setup + production + inventory cost; passes when feasible and cost <= 8081.53 + 0.005
29
+ Companions (not included): 2.detailed_results/MIP_600/NCLSLT_MIP_600s.CSV; 1.instances/readme.txt; README.md
30
+ Dependencies: no upstream code used
campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_cobot_albp2/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://raw.githubusercontent.com/zixiangliwust/Instances_Cobot_ALBP2/6a8cbee89e5c39515ae8bcfa8b7b52372c37b45d/LICENSE (publisher: Zixiang Li (Wuhan University of Science and Technology) and coauthors)
campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_cobot_albp2/README.md ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # github_zixiangliwust_instances_cobot_albp2
2
+
3
+ **What this is:** Files published by Zixiang Li (Wuhan University of Science and Technology) and coauthors. Collection theme: Production planning and lot sizing instance libraries (lot sizing, capacity planning, assembly line balancing and sequencing models, with published optimal values or bounds).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Text with tagged sections: <number of tasks>, <number of stations>, <order strength>, <type of the robots>, robot and collaboration flexibility, <number of robots>, <task times> (task, human time, robot time(s), collaboration time(s); 99999 means not possible), <precedence relations>, <end>; text with tagged sections: <number of tasks>, <number of stations>, <order strength>, <type of the robots>, robot and collaboration flexibility, <number of robots>, <task times> (task, human time, robot time(s), collaboration time(s); 10000 means not possible), <precedence relations>, <end>.
6
+
7
+ **Where it comes from:** Upstream name: github_zixiangliwust_Instances_Cobot_ALBP2. Upstream: github.com/zixiangliwust/Instances_Cobot_ALBP2. Kind: test or example files shipped with a tool. Instances of Weckenborg, Kieckhaefer, Mueller, Grunewald and Spengler 2020 (Business Research 13:93-132; README.md line 21) as used by Li and coauthors, with their result files; recorded tool_fixture (spec 115 general rule 6); instances re-generated by Li and coauthors following Mao, Sun, Fang, Huang and Zhang 2024 (IJPE 271:109207; README.md line 6 and line 23), multi-type cobots, with their result files; recorded tool_fixture (spec 115 general rule 6).
8
+
9
+ **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/zixiangliwust/Instances_Cobot_ALBP2
16
+
17
+ Files:
18
+
19
+ - `Instances_Singletype/instance_n=100_166_1.txt`: Cobot assembly line balancing type 2 instance instance_n=100_166_1.txt (single robot type, 100 tasks, 25 stations) with published cycle-time upper and lower bounds.
20
+ Format: text with tagged sections: <number of tasks>, <number of stations>, <order strength>, <type of the robots>, robot and collaboration flexibility, <number of robots>, <task times> (task, human time, robot time(s), collaboration time(s); 99999 means not possible), <precedence relations>, <end>
21
+ Origin: instances of Weckenborg, Kieckhaefer, Mueller, Grunewald and Spengler 2020 (Business Research 13:93-132; README.md line 21) as used by Li and coauthors, with their result files; recorded tool_fixture (spec 115 general rule 6)
22
+ Check: Program: read the agent assignment task -> station and mode (human, robot of a type, or collaboration) and start times; feasible when every task is assigned once in a mode whose time is not 99999, robots used do not exceed <number of robots>, precedence holds across and inside stations, and a station's human and robot do not overlap their own tasks; cycle time = latest finish over stations; passes when feasible and cycle time <= 883 (best known; below the lower bound 571 means a checker error)
23
+ Companions (not included): Instances_Singletype_Results/instance_n=100_166_1.txt; LICENSE; README.md
24
+ Dependencies: no upstream code used
25
+ - `Instances_Multitype_by_Weckenborg/instance_n=100_166_0.txt`: Cobot assembly line balancing type 2 instance instance_n=100_166_0.txt (four robot types, 100 tasks, 25 stations) with published cycle-time upper and lower bounds.
26
+ Format: text with tagged sections: <number of tasks>, <number of stations>, <order strength>, <type of the robots>, robot and collaboration flexibility, <number of robots>, <task times> (task, human time, robot time(s), collaboration time(s); 10000 means not possible), <precedence relations>, <end>
27
+ Origin: instances re-generated by Li and coauthors following Mao, Sun, Fang, Huang and Zhang 2024 (IJPE 271:109207; README.md line 6 and line 23), multi-type cobots, with their result files; recorded tool_fixture (spec 115 general rule 6)
28
+ Check: Program: read the agent assignment task -> station and mode (human, robot of a type, or collaboration) and start times; feasible when every task is assigned once in a mode whose time is not 10000, robots used do not exceed <number of robots>, precedence holds across and inside stations, and a station's human and robot do not overlap their own tasks; cycle time = latest finish over stations; passes when feasible and cycle time <= 943 (best known; below the lower bound 571 means a checker error)
29
+ Companions (not included): Instances_Multitype_by_Weckenborg_Results/instance_n=100_166_0.txt; LICENSE; README.md
30
+ Dependencies: no upstream code used
31
+ - `Instances_Singletype/instance_n=20_141_1.txt`: Cobot assembly line balancing type 2 instance instance_n=20_141_1.txt (single robot type, 20 tasks, 5 stations) with published cycle-time optimum.
32
+ Format: text with tagged sections: <number of tasks>, <number of stations>, <order strength>, <type of the robots>, robot and collaboration flexibility, <number of robots>, <task times> (task, human time, robot time(s), collaboration time(s); 99999 means not possible), <precedence relations>, <end>
33
+ Origin: instances of Weckenborg, Kieckhaefer, Mueller, Grunewald and Spengler 2020 (Business Research 13:93-132; README.md line 21) as used by Li and coauthors, with their result files; recorded tool_fixture (spec 115 general rule 6)
34
+ Check: Program: read the agent assignment task -> station and mode (human, robot of a type, or collaboration) and start times; feasible when every task is assigned once in a mode whose time is not 99999, robots used do not exceed <number of robots>, precedence holds across and inside stations, and a station's human and robot do not overlap their own tasks; cycle time = latest finish over stations; passes when feasible and cycle time <= 537 (the proven optimum)
35
+ Companions (not included): Instances_Singletype_Results/instance_n=20_141_1.txt; LICENSE; README.md
36
+ Dependencies: no upstream code used
campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_ralsp/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://raw.githubusercontent.com/zixiangliwust/Instances_RALSP/8ea1cb8610bc8fbdc772c8fa022482cc41aa7593/LICENSE (publisher: Zixiang Li (Wuhan University of Science and Technology) and coauthors)
campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_ralsp/README.md ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # github_zixiangliwust_instances_ralsp
2
+
3
+ **What this is:** Files published by Zixiang Li (Wuhan University of Science and Technology) and coauthors. Collection theme: Production planning and lot sizing instance libraries (lot sizing, capacity planning, assembly line balancing and sequencing models, with published optimal values or bounds).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Text with tagged sections: <number of products>, <product cycle> (product, demand), <number of workstations>, <station length>, <working hours> (product x station), <interval time>, <product switching cost> (i,j cost), <number of parts>, <optional frequency>, <product assembly parts> (0/1 matrix).
6
+
7
+ **Where it comes from:** Upstream name: github_zixiangliwust_Instances_RALSP. Upstream: github.com/zixiangliwust/Instances_RALSP. Kind: test or example files shipped with a tool. Illustrative example of the paper "Learning-based multi-objective hyper-heuristic algorithm for reconfigurable assembly line scheduling problems" (README.md line 7; journal under review, README.md line 252) with the model-based Pareto front of its Section 5; recorded tool_fixture (spec 115 general rule 6).
8
+
9
+ **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: Program: read the agent sequence of 16 product codes (0 to 3, each product as often as its demand); compute ReconfCost, Violations and LogisticsLeveling; passes when the vector is not dominated by any of the 13 front points (all three minimised), each compared within half a unit of the 6th decimal; matching a front point exactly also passes.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/zixiangliwust/Instances_RALSP
16
+
17
+ Files:
18
+
19
+ - `Instances/example_of_english_paper.txt`: Reconfigurable assembly line scheduling (RALSP) illustrative example: 4 products, 5 stations, 16 units, with the exact three-objective Pareto front of the MILP model.
20
+ Format: text with tagged sections: <number of products>, <product cycle> (product, demand), <number of workstations>, <station length>, <working hours> (product x station), <interval time>, <product switching cost> (i,j cost), <number of parts>, <optional frequency>, <product assembly parts> (0/1 matrix)
21
+ Origin: illustrative example of the paper "Learning-based multi-objective hyper-heuristic algorithm for reconfigurable assembly line scheduling problems" (README.md line 7; journal under review, README.md line 252) with the model-based Pareto front of its Section 5; recorded tool_fixture (spec 115 general rule 6)
22
+ Check: Program: read the agent sequence of 16 product codes (0 to 3, each product as often as its demand); compute ReconfCost, Violations and LogisticsLeveling; passes when the vector is not dominated by any of the 13 front points (all three minimised), each compared within half a unit of the 6th decimal; matching a front point exactly also passes
23
+ Companions (not included): results/Pareto_solutions_by_models in Section 5.txt; LICENSE; README.md
24
+ Dependencies: no upstream code run
campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_two_sided_albp1/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://raw.githubusercontent.com/zixiangliwust/Instances_Two_sided_ALBP1/34b2363598c2668dc384206845b22bd8852a1472/LICENSE (publisher: Zixiang Li (Wuhan University of Science and Technology) and coauthors)
campaign_files/2026-10-g-solver-instances-2/github_zixiangliwust_instances_two_sided_albp1/README.md ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # github_zixiangliwust_instances_two_sided_albp1
2
+
3
+ **What this is:** Files published by Zixiang Li (Wuhan University of Science and Technology) and coauthors. Collection theme: Production planning and lot sizing instance libraries (lot sizing, capacity planning, assembly line balancing and sequencing models, with published optimal values or bounds).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Text: <number of tasks>, <cycle time>, <task times> (task, time), <task directions> (task, L, R or E), <precedence relations> (i,j), <end>.
6
+
7
+ **Where it comes from:** Upstream name: github_zixiangliwust_Instances_Two_sided_ALBP1. Upstream: github.com/zixiangliwust/Instances_Two_sided_ALBP1. Kind: test or example files shipped with a tool. Benchmark instances collected from the TALBP literature (Li, Tang and Zhang 2017, C&OR 79:78-93, README.md line 29) with results of Li, Kucukkoc, Liu, Zhang and Zhang, Optimization and Engineering 2026, DOI 10.1007/s11081-025-10067-0 (README.md line 37); recorded tool_fixture (spec 115 general rule 6).
8
+
9
+ **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: Program: read the agent assignment task -> (mated station, side L or R) and start time; feasible when every task is assigned once, a task with direction L or R is on that side (E either side), every precedence pair (i,j) has i finished before j starts (on the same side) or i in an earlier mated station, and every task finishes within the cycle time on its side, start times respecting cross-side precedence inside a mated station; passes when feasible and the number of used stations (sides with work) equals the published optimum.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/zixiangliwust/Instances_Two_sided_ALBP1
16
+
17
+ Files:
18
+
19
+ - `Instances/P12_4.txt`: Two-sided assembly line balancing type I instance P12_4 (12 tasks, cycle time 4) with optimal number of stations 7.
20
+ Format: text: <number of tasks>, <cycle time>, <task times> (task, time), <task directions> (task, L, R or E), <precedence relations> (i,j), <end>
21
+ Origin: benchmark instances collected from the TALBP literature (Li, Tang and Zhang 2017, C&OR 79:78-93, README.md line 29) with results of Li, Kucukkoc, Liu, Zhang and Zhang, Optimization and Engineering 2026, DOI 10.1007/s11081-025-10067-0 (README.md line 37); recorded tool_fixture (spec 115 general rule 6)
22
+ Check: Program: read the agent assignment task -> (mated station, side L or R) and start time; feasible when every task is assigned once, a task with direction L or R is on that side (E either side), every precedence pair (i,j) has i finished before j starts (on the same side) or i in an earlier mated station, and every task finishes within the cycle time on its side, start times respecting cross-side precedence inside a mated station; passes when feasible and the number of used stations (sides with work) equals the published optimum
23
+ Companions (not included): Results in 10.1007s11081-025-10067-0/Comparative results/Results by published methods and developed methods.xlsx; LICENSE; README.md
24
+ Dependencies: no upstream code used
25
+ - `Instances/P65_381.txt`: Two-sided assembly line balancing type I instance P65_381 (65 tasks, cycle time 381) with optimal number of stations 14.
26
+ Format: text: <number of tasks>, <cycle time>, <task times> (task, time), <task directions> (task, L, R or E), <precedence relations> (i,j), <end>
27
+ Origin: benchmark instances collected from the TALBP literature (Li, Tang and Zhang 2017, C&OR 79:78-93, README.md line 29) with results of Li, Kucukkoc, Liu, Zhang and Zhang, Optimization and Engineering 2026, DOI 10.1007/s11081-025-10067-0 (README.md line 37); recorded tool_fixture (spec 115 general rule 6)
28
+ Check: Program: read the agent assignment task -> (mated station, side L or R) and start time; feasible when every task is assigned once, a task with direction L or R is on that side (E either side), every precedence pair (i,j) has i finished before j starts (on the same side) or i in an earlier mated station, and every task finishes within the cycle time on its side, start times respecting cross-side precedence inside a mated station; passes when feasible and the number of used stations (sides with work) equals the published optimum
29
+ Companions (not included): Results in 10.1007s11081-025-10067-0/Comparative results/Results by published methods and developed methods.xlsx; LICENSE; README.md
30
+ Dependencies: no upstream code used
campaign_files/2026-10-g-solver-instances-2/iwls2022_ls_contest/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://github.com/alanminko/iwls2022-ls-contest/blob/ee93f7ca742f172aa64a4a6e12a5b59944475bf0/LICENSE (publisher: Alan Mishchenko (UC Berkeley), IWLS 2022 contest chairs)
campaign_files/2026-10-g-solver-instances-2/iwls2022_ls_contest/README.md ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # iwls2022_ls_contest
2
+
3
+ **What this is:** Units published by Alan Mishchenko (UC Berkeley), IWLS 2022 contest chairs. Collection theme: Logic synthesis and optimisation benchmark instances (combinational circuits with best known node counts or levels: EPFL benchmarks, IWLS contests and the like).
4
+
5
+ **What one unit is:** Each unit is one upstream folder, delivered as all the files fetched from it, kept at its upstream path. Format: Directory of 100 text files exNN.truth; one file = one case; each line is one output, a string of 2^n characters '0'/'1', the leftmost character is the output for the all-ones input pattern and the rightmost for all zeros, input k is bit k of the pattern index (convention checked this round: the best AIG of all 100 cases simulates to the truth table exactly).
6
+
7
+ **Where it comes from:** Upstream: https://github.com/alanminko/iwls2022-ls-contest. Kind: test or example files shipped with a tool. IWLS 2022 Programming Contest (chairs Alan Mishchenko and Satrajit Chatterjee): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs of 7 teams; categories: random-looking, random decomposable, AES s-box and inverse, majority, sorters, Espresso benchmarks with permuted inputs, arithmetic functions with permuted inputs and dropped outputs, LogicNets neurons (README.md lines 22-31); recorded tool_fixture (official contest instances, spec 115 general rule 6).
8
+
9
+ **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. Each unit names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
+
11
+ **Tools to open and check:** Check: Program: the agent returns a binary AIGER file with only two-input AND nodes (README.md lines 6-8); pass when the number of inputs and outputs matches the .truth file, simulating every one of the 2^n input patterns reproduces every output line exactly, and the AND count (header field A) is not above the Best column of submissions/nodes.txt for that case. The simulator was checked this round on all 100 best AIGs (100 of 100 reproduce their truth tables, AND counts equal to Best). Note: in nodes.txt the UCAS entries of ex26, ex27 and ex40 are 1 below the header of the UCAS AIG; the Best column is not affected. Hide submissions/ when posing.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/alanminko/iwls2022-ls-contest
16
+
17
+ Units:
18
+
19
+ - `benchmarks/`: IWLS 2022 Programming Contest: 100 truth tables to synthesise into AIGs with the fewest two-input AND nodes, with the best submitted node counts.
20
+ Format: directory of 100 text files exNN.truth; one file = one case; each line is one output, a string of 2^n characters '0'/'1', the leftmost character is the output for the all-ones input pattern and the rightmost for all zeros, input k is bit k of the pattern index (convention checked this round: the best AIG of all 100 cases simulates to the truth table exactly)
21
+ Origin: IWLS 2022 Programming Contest (chairs Alan Mishchenko and Satrajit Chatterjee): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs of 7 teams; categories: random-looking, random decomposable, AES s-box and inverse, majority, sorters, Espresso benchmarks with permuted inputs, arithmetic functions with permuted inputs and dropped outputs, LogicNets neurons (README.md lines 22-31); recorded tool_fixture (official contest instances, spec 115 general rule 6)
22
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes (README.md lines 6-8); pass when the number of inputs and outputs matches the .truth file, simulating every one of the 2^n input patterns reproduces every output line exactly, and the AND count (header field A) is not above the Best column of submissions/nodes.txt for that case. The simulator was checked this round on all 100 best AIGs (100 of 100 reproduce their truth tables, AND counts equal to Best). Note: in nodes.txt the UCAS entries of ex26, ex27 and ex40 are 1 below the header of the UCAS AIG; the Best column is not affected. Hide submissions/ when posing.
23
+ Companions (not included): README.md; submissions/nodes.txt; submissions/scores.txt; submissions/levels.txt; IWLS_2022_Programming_Contest.pdf; LICENSE
24
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
campaign_files/2026-10-g-solver-instances-2/iwls2023_ls_contest/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://github.com/alanminko/iwls2023-ls-contest/blob/7dbfd510e2597cd07fdebcff91b5ac262d63c3f7/LICENSE (publisher: Alan Mishchenko (UC Berkeley), IWLS 2023 contest chairs)
campaign_files/2026-10-g-solver-instances-2/iwls2023_ls_contest/README.md ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # iwls2023_ls_contest
2
+
3
+ **What this is:** Files published by Alan Mishchenko (UC Berkeley), IWLS 2023 contest chairs. Collection theme: Logic synthesis and optimisation benchmark instances (combinational circuits with best known node counts or levels: EPFL benchmarks, IWLS contests and the like).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Text truth table: 7 lines, one per output, each a string of 2^6 = 64 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs); text truth table: 3 lines, one per output, each a string of 2^9 = 512 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs); text truth table: 3 lines, one per output, each a string of 2^11 = 2048 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs); text truth table: 14 lines, one per output, each a string of 2^10 = 1024 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs).
6
+
7
+ **Where it comes from:** Upstream: https://github.com/alanminko/iwls2023-ls-contest. Kind: test or example files shipped with a tool. IWLS 2023 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6).
8
+
9
+ **Self-contained:** Whether each file needs other upstream files is not stated. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/alanminko/iwls2023-ls-contest
16
+
17
+ Files:
18
+
19
+ - `benchmarks/ex41.truth`: IWLS 2023 Programming Contest ex41: truth table (6 inputs, 7 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 18.
20
+ Format: text truth table: 7 lines, one per output, each a string of 2^6 = 64 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
21
+ Origin: IWLS 2023 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
22
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 6 inputs and 7 outputs, simulating all 2^6 input patterns reproduces every output line of benchmarks/ex41.truth exactly, and the AND count (header field A) is not above 18. The simulator was checked this round on submissions/epfl/aig/ex41.aig (reproduces the truth table, AND count 18); the 4 team AIGs of this case all simulate to the truth table exactly (checked this round). Hide submissions/ when posing.
23
+ Companions (not included): README.md; LICENSE; IWLS_2023_Programming_Contest.pdf; submissions/epfl/aig/ex41.aig; submissions/google/aig/ex41.aig; submissions/nbu/aig/ex41.aig; submissions/tuw/aig/ex41.aig
24
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
25
+ - `benchmarks/ex23.truth`: IWLS 2023 Programming Contest ex23: truth table (9 inputs, 3 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 20.
26
+ Format: text truth table: 3 lines, one per output, each a string of 2^9 = 512 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
27
+ Origin: IWLS 2023 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
28
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 9 inputs and 3 outputs, simulating all 2^9 input patterns reproduces every output line of benchmarks/ex23.truth exactly, and the AND count (header field A) is not above 20. The simulator was checked this round on submissions/epfl/aig/ex23.aig (reproduces the truth table, AND count 20); the 4 team AIGs of this case all simulate to the truth table exactly (checked this round). Hide submissions/ when posing.
29
+ Companions (not included): README.md; LICENSE; IWLS_2023_Programming_Contest.pdf; submissions/epfl/aig/ex23.aig; submissions/google/aig/ex23.aig; submissions/nbu/aig/ex23.aig; submissions/tuw/aig/ex23.aig
30
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
31
+ - `benchmarks/ex57.truth`: IWLS 2023 Programming Contest ex57: truth table (11 inputs, 3 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 30.
32
+ Format: text truth table: 3 lines, one per output, each a string of 2^11 = 2048 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
33
+ Origin: IWLS 2023 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
34
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 11 inputs and 3 outputs, simulating all 2^11 input patterns reproduces every output line of benchmarks/ex57.truth exactly, and the AND count (header field A) is not above 30. The simulator was checked this round on submissions/google/aig/ex57.aig (reproduces the truth table, AND count 30); the 4 team AIGs of this case all simulate to the truth table exactly (checked this round). Hide submissions/ when posing.
35
+ Companions (not included): README.md; LICENSE; IWLS_2023_Programming_Contest.pdf; submissions/epfl/aig/ex57.aig; submissions/google/aig/ex57.aig; submissions/nbu/aig/ex57.aig; submissions/tuw/aig/ex57.aig
36
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
37
+ - `benchmarks/ex10.truth`: IWLS 2023 Programming Contest ex10: truth table (10 inputs, 14 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 38.
38
+ Format: text truth table: 14 lines, one per output, each a string of 2^10 = 1024 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
39
+ Origin: IWLS 2023 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
40
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 10 inputs and 14 outputs, simulating all 2^10 input patterns reproduces every output line of benchmarks/ex10.truth exactly, and the AND count (header field A) is not above 38. The simulator was checked this round on submissions/epfl/aig/ex10.aig (reproduces the truth table, AND count 38); the 4 team AIGs of this case all simulate to the truth table exactly (checked this round). Hide submissions/ when posing.
41
+ Companions (not included): README.md; LICENSE; IWLS_2023_Programming_Contest.pdf; submissions/epfl/aig/ex10.aig; submissions/google/aig/ex10.aig; submissions/nbu/aig/ex10.aig; submissions/tuw/aig/ex10.aig
42
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
campaign_files/2026-10-g-solver-instances-2/iwls2024_ls_contest/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://github.com/alanminko/iwls2024-ls-contest/blob/3d5ea875961db226a7fe40601cda2eb42d16c181/LICENSE (publisher: Alan Mishchenko (UC Berkeley), IWLS 2024 contest chairs)
campaign_files/2026-10-g-solver-instances-2/iwls2024_ls_contest/README.md ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # iwls2024_ls_contest
2
+
3
+ **What this is:** Files published by Alan Mishchenko (UC Berkeley), IWLS 2024 contest chairs. Collection theme: Logic synthesis and optimisation benchmark instances (combinational circuits with best known node counts or levels: EPFL benchmarks, IWLS contests and the like).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. The format of each file is in the notes.
6
+
7
+ **Where it comes from:** Upstream: https://github.com/alanminko/iwls2024-ls-contest. Kind: test or example files shipped with a tool. IWLS 2024 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6).
8
+
9
+ **Self-contained:** Whether each file needs other upstream files is not stated. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/alanminko/iwls2024-ls-contest
16
+
17
+ Files:
18
+
19
+ - `benchmarks/ex03.truth`: IWLS 2024 Programming Contest ex03: truth table (8 inputs, 3 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 15.
20
+ Format: text truth table: 3 lines, one per output, each a string of 2^8 = 256 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
21
+ Origin: IWLS 2024 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
22
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 8 inputs and 3 outputs, simulating all 2^8 input patterns reproduces every output line of benchmarks/ex03.truth exactly, and the AND count (header field A) is not above 15. XAIG variant: same check with XOR nodes allowed and the bound 14. The simulator was checked this round on submissions/ALCom_Lab/aig/ex03.aig (reproduces the truth table, AND count 15); the column order of results_aig.txt is inferred as the alphabetical order of the submission folders (ALCom_Lab, Gaiger, Kapenga, NBU, NUP, TU_Wien, Team_EPFL, Team_TUM, USTC_and_Huawei); confirmed on 6 cases by simulating the AIG of the team whose column holds the best value. Hide submissions/ when posing.
23
+ Companions (not included): README.md; LICENSE; IWLS_2024_Programming_Contest.pdf; submissions/results_aig.txt; submissions/results_xaig.txt; submissions/ALCom_Lab/aig/ex03.aig
24
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
25
+ - `benchmarks/ex38.truth`: IWLS 2024 Programming Contest ex38: truth table (12 inputs, 3 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 272.
26
+ Format: text truth table: 3 lines, one per output, each a string of 2^12 = 4096 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
27
+ Origin: IWLS 2024 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
28
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 12 inputs and 3 outputs, simulating all 2^12 input patterns reproduces every output line of benchmarks/ex38.truth exactly, and the AND count (header field A) is not above 272. XAIG variant: same check with XOR nodes allowed and the bound 226. The simulator was checked this round on submissions/TU_Wien/aig/ex38.aig (reproduces the truth table, AND count 272); the column order of results_aig.txt is inferred as the alphabetical order of the submission folders (ALCom_Lab, Gaiger, Kapenga, NBU, NUP, TU_Wien, Team_EPFL, Team_TUM, USTC_and_Huawei); confirmed on 6 cases by simulating the AIG of the team whose column holds the best value. Hide submissions/ when posing.
29
+ Companions (not included): README.md; LICENSE; IWLS_2024_Programming_Contest.pdf; submissions/results_aig.txt; submissions/results_xaig.txt; submissions/TU_Wien/aig/ex38.aig
30
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
31
+ - `benchmarks/ex48.truth`: IWLS 2024 Programming Contest ex48: truth table (10 inputs, 13 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 504.
32
+ Format: text truth table: 13 lines, one per output, each a string of 2^10 = 1024 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
33
+ Origin: IWLS 2024 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
34
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 10 inputs and 13 outputs, simulating all 2^10 input patterns reproduces every output line of benchmarks/ex48.truth exactly, and the AND count (header field A) is not above 504. XAIG variant: same check with XOR nodes allowed and the bound 410. The simulator was checked this round on submissions/USTC_and_Huawei/aig/ex48.aig (reproduces the truth table, AND count 504); the column order of results_aig.txt is inferred as the alphabetical order of the submission folders (ALCom_Lab, Gaiger, Kapenga, NBU, NUP, TU_Wien, Team_EPFL, Team_TUM, USTC_and_Huawei); confirmed on 6 cases by simulating the AIG of the team whose column holds the best value. Hide submissions/ when posing.
35
+ Companions (not included): README.md; LICENSE; IWLS_2024_Programming_Contest.pdf; submissions/results_aig.txt; submissions/results_xaig.txt; submissions/USTC_and_Huawei/aig/ex48.aig
36
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
37
+ - `benchmarks/ex40.truth`: IWLS 2024 Programming Contest ex40: truth table (11 inputs, 29 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 1202.
38
+ Format: text truth table: 29 lines, one per output, each a string of 2^11 = 2048 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
39
+ Origin: IWLS 2024 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
40
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 11 inputs and 29 outputs, simulating all 2^11 input patterns reproduces every output line of benchmarks/ex40.truth exactly, and the AND count (header field A) is not above 1202. XAIG variant: same check with XOR nodes allowed and the bound 1090. The simulator was checked this round on submissions/ALCom_Lab/aig/ex40.aig (reproduces the truth table, AND count 1202); the column order of results_aig.txt is inferred as the alphabetical order of the submission folders (ALCom_Lab, Gaiger, Kapenga, NBU, NUP, TU_Wien, Team_EPFL, Team_TUM, USTC_and_Huawei); confirmed on 6 cases by simulating the AIG of the team whose column holds the best value. Hide submissions/ when posing.
41
+ Companions (not included): README.md; LICENSE; IWLS_2024_Programming_Contest.pdf; submissions/results_aig.txt; submissions/results_xaig.txt; submissions/ALCom_Lab/aig/ex40.aig
42
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
43
+ - `benchmarks/ex79.truth`: IWLS 2024 Programming Contest ex79: truth table (12 inputs, 115 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 4671.
44
+ Format: text truth table: 115 lines, one per output, each a string of 2^12 = 4096 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
45
+ Origin: IWLS 2024 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
46
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 12 inputs and 115 outputs, simulating all 2^12 input patterns reproduces every output line of benchmarks/ex79.truth exactly, and the AND count (header field A) is not above 4671. XAIG variant: same check with XOR nodes allowed and the bound 4651. The simulator was checked this round on submissions/Team_EPFL/aig/ex79.aig (reproduces the truth table, AND count 4671); the column order of results_aig.txt is inferred as the alphabetical order of the submission folders (ALCom_Lab, Gaiger, Kapenga, NBU, NUP, TU_Wien, Team_EPFL, Team_TUM, USTC_and_Huawei); confirmed on 6 cases by simulating the AIG of the team whose column holds the best value. Hide submissions/ when posing.
47
+ Companions (not included): README.md; LICENSE; IWLS_2024_Programming_Contest.pdf; submissions/results_aig.txt; submissions/results_xaig.txt; submissions/Team_EPFL/aig/ex79.aig
48
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
49
+ - `benchmarks/ex18.truth`: IWLS 2024 Programming Contest ex18: truth table (13 inputs, 6 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 56.
50
+ Format: text truth table: 6 lines, one per output, each a string of 2^13 = 8192 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIG)
51
+ Origin: IWLS 2024 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
52
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 13 inputs and 6 outputs, simulating all 2^13 input patterns reproduces every output line of benchmarks/ex18.truth exactly, and the AND count (header field A) is not above 56. XAIG variant: same check with XOR nodes allowed and the bound 49. The simulator was checked this round on submissions/Kapenga/aig/ex18.aig (reproduces the truth table, AND count 56); the column order of results_aig.txt is the alphabetical order of the submission folders (ALCom_Lab, Gaiger, Kapenga, NBU, NUP, TU_Wien, Team_EPFL, Team_TUM, USTC_and_Huawei), inferred and confirmed again here by simulating the AIG of the team whose column holds the best value. Hide submissions/ when posing.
53
+ Companions (not included): README.md; LICENSE; IWLS_2024_Programming_Contest.pdf; submissions/results_aig.txt; submissions/results_xaig.txt; submissions/Kapenga/aig/ex18.aig
54
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
55
+ - `benchmarks/ex42.truth`: IWLS 2024 Programming Contest ex42: truth table (13 inputs, 3 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 40.
56
+ Format: text truth table: 3 lines, one per output, each a string of 2^13 = 8192 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIG)
57
+ Origin: IWLS 2024 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
58
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 13 inputs and 3 outputs, simulating all 2^13 input patterns reproduces every output line of benchmarks/ex42.truth exactly, and the AND count (header field A) is not above 40. XAIG variant: same check with XOR nodes allowed and the bound 31. The simulator was checked this round on submissions/NUP/aig/ex42.aig (reproduces the truth table, AND count 40); the column order of results_aig.txt is the alphabetical order of the submission folders (ALCom_Lab, Gaiger, Kapenga, NBU, NUP, TU_Wien, Team_EPFL, Team_TUM, USTC_and_Huawei), inferred and confirmed again here by simulating the AIG of the team whose column holds the best value. Hide submissions/ when posing.
59
+ Companions (not included): README.md; LICENSE; IWLS_2024_Programming_Contest.pdf; submissions/results_aig.txt; submissions/results_xaig.txt; submissions/NUP/aig/ex42.aig
60
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
61
+ - `benchmarks/ex53.truth`: IWLS 2024 Programming Contest ex53: truth table (12 inputs, 8 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 48.
62
+ Format: text truth table: 8 lines, one per output, each a string of 2^12 = 4096 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIG)
63
+ Origin: IWLS 2024 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
64
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 12 inputs and 8 outputs, simulating all 2^12 input patterns reproduces every output line of benchmarks/ex53.truth exactly, and the AND count (header field A) is not above 48. XAIG variant: same check with XOR nodes allowed and the bound 40. The simulator was checked this round on submissions/USTC_and_Huawei/aig/ex53.aig (reproduces the truth table, AND count 48); the column order of results_aig.txt is the alphabetical order of the submission folders (ALCom_Lab, Gaiger, Kapenga, NBU, NUP, TU_Wien, Team_EPFL, Team_TUM, USTC_and_Huawei), inferred and confirmed again here by simulating the AIG of the team whose column holds the best value. Hide submissions/ when posing.
65
+ Companions (not included): README.md; LICENSE; IWLS_2024_Programming_Contest.pdf; submissions/results_aig.txt; submissions/results_xaig.txt; submissions/USTC_and_Huawei/aig/ex53.aig
66
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
67
+ - `benchmarks/ex76.truth`: IWLS 2024 Programming Contest ex76: truth table (11 inputs, 15 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 46.
68
+ Format: text truth table: 15 lines, one per output, each a string of 2^11 = 2048 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIG)
69
+ Origin: IWLS 2024 Programming Contest (contest chairs Alan Mishchenko and Yukio Miyasaka, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables, with the AIGs submitted by the teams (README.md line 33: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
70
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 11 inputs and 15 outputs, simulating all 2^11 input patterns reproduces every output line of benchmarks/ex76.truth exactly, and the AND count (header field A) is not above 46. XAIG variant: same check with XOR nodes allowed and the bound 38. The simulator was checked this round on submissions/ALCom_Lab/aig/ex76.aig (reproduces the truth table, AND count 46); the column order of results_aig.txt is the alphabetical order of the submission folders (ALCom_Lab, Gaiger, Kapenga, NBU, NUP, TU_Wien, Team_EPFL, Team_TUM, USTC_and_Huawei), inferred and confirmed again here by simulating the AIG of the team whose column holds the best value. Hide submissions/ when posing.
71
+ Companions (not included): README.md; LICENSE; IWLS_2024_Programming_Contest.pdf; submissions/results_aig.txt; submissions/results_xaig.txt; submissions/ALCom_Lab/aig/ex76.aig
72
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
campaign_files/2026-10-g-solver-instances-2/iwls2025_ls_contest/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://github.com/alanminko/iwls2025-ls-contest/blob/defcf35c16381ead60776f08a3dd088b954b2ede/LICENSE (publisher: Alan Mishchenko (UC Berkeley), IWLS 2025 contest chairs)
campaign_files/2026-10-g-solver-instances-2/iwls2025_ls_contest/README.md ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # iwls2025_ls_contest
2
+
3
+ **What this is:** Files published by Alan Mishchenko (UC Berkeley), IWLS 2025 contest chairs. Collection theme: Logic synthesis and optimisation benchmark instances (combinational circuits with best known node counts or levels: EPFL benchmarks, IWLS contests and the like).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Text truth table: 7 lines, one per output, each a string of 2^7 = 128 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs); text truth table: 4 lines, one per output, each a string of 2^8 = 256 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs); text truth table: 16 lines, one per output, each a string of 2^8 = 256 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs); text truth table: 10 lines, one per output, each a string of 2^10 = 1024 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs); text truth table: 6 lines, one per output, each a string of 2^10 = 1024 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs).
6
+
7
+ **Where it comes from:** Upstream: https://github.com/alanminko/iwls2025-ls-contest. Kind: test or example files shipped with a tool. IWLS 2025 Programming Contest (contest chairs Alan Mishchenko and Alessandro Tempia Calvino, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables (the 2025 contest adds a new set ex100-ex199, README.md lines 19-22), with the AIGs submitted by the teams (README.md line 39: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6).
8
+
9
+ **Self-contained:** Whether each file needs other upstream files is not stated. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/alanminko/iwls2025-ls-contest
16
+
17
+ Files:
18
+
19
+ - `benchmarks/2025/ex153.truth`: IWLS 2025 Programming Contest ex153: truth table (7 inputs, 7 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 60.
20
+ Format: text truth table: 7 lines, one per output, each a string of 2^7 = 128 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
21
+ Origin: IWLS 2025 Programming Contest (contest chairs Alan Mishchenko and Alessandro Tempia Calvino, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables (the 2025 contest adds a new set ex100-ex199, README.md lines 19-22), with the AIGs submitted by the teams (README.md line 39: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
22
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 7 inputs and 7 outputs, simulating all 2^7 input patterns reproduces every output line of benchmarks/2025/ex153.truth exactly, and the AND count (header field A) is not above 60. The simulator was checked this round on submissions/UCB_NTU/2025/ex153.aig (reproduces the truth table, AND count 60); the 4 team AIGs of this case all simulate to the truth table exactly (checked this round). Hide submissions/ when posing.
23
+ Companions (not included): README.md; LICENSE; IWLS_2025_Programming_Contest.pdf; submissions/NUP_JetBrains/2025/ex153.aig; submissions/UCB_NTU/2025/ex153.aig; submissions/UCLA/2025/ex153.aig; submissions/UF_UL/2025/ex153.aig
24
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
25
+ - `benchmarks/2025/ex125.truth`: IWLS 2025 Programming Contest ex125: truth table (8 inputs, 4 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 67.
26
+ Format: text truth table: 4 lines, one per output, each a string of 2^8 = 256 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
27
+ Origin: IWLS 2025 Programming Contest (contest chairs Alan Mishchenko and Alessandro Tempia Calvino, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables (the 2025 contest adds a new set ex100-ex199, README.md lines 19-22), with the AIGs submitted by the teams (README.md line 39: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
28
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 8 inputs and 4 outputs, simulating all 2^8 input patterns reproduces every output line of benchmarks/2025/ex125.truth exactly, and the AND count (header field A) is not above 67. The simulator was checked this round on submissions/UCB_NTU/2025/ex125.aig (reproduces the truth table, AND count 67); the 4 team AIGs of this case all simulate to the truth table exactly (checked this round). Hide submissions/ when posing.
29
+ Companions (not included): README.md; LICENSE; IWLS_2025_Programming_Contest.pdf; submissions/NUP_JetBrains/2025/ex125.aig; submissions/UCB_NTU/2025/ex125.aig; submissions/UCLA/2025/ex125.aig; submissions/UF_UL/2025/ex125.aig
30
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
31
+ - `benchmarks/2025/ex140.truth`: IWLS 2025 Programming Contest ex140: truth table (8 inputs, 16 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 131.
32
+ Format: text truth table: 16 lines, one per output, each a string of 2^8 = 256 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
33
+ Origin: IWLS 2025 Programming Contest (contest chairs Alan Mishchenko and Alessandro Tempia Calvino, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables (the 2025 contest adds a new set ex100-ex199, README.md lines 19-22), with the AIGs submitted by the teams (README.md line 39: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
34
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 8 inputs and 16 outputs, simulating all 2^8 input patterns reproduces every output line of benchmarks/2025/ex140.truth exactly, and the AND count (header field A) is not above 131. The simulator was checked this round on submissions/UCB_NTU/2025/ex140.aig (reproduces the truth table, AND count 131); the 4 team AIGs of this case all simulate to the truth table exactly (checked this round). Hide submissions/ when posing.
35
+ Companions (not included): README.md; LICENSE; IWLS_2025_Programming_Contest.pdf; submissions/NUP_JetBrains/2025/ex140.aig; submissions/UCB_NTU/2025/ex140.aig; submissions/UCLA/2025/ex140.aig; submissions/UF_UL/2025/ex140.aig
36
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
37
+ - `benchmarks/2025/ex100.truth`: IWLS 2025 Programming Contest ex100: truth table (10 inputs, 10 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 156.
38
+ Format: text truth table: 10 lines, one per output, each a string of 2^10 = 1024 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
39
+ Origin: IWLS 2025 Programming Contest (contest chairs Alan Mishchenko and Alessandro Tempia Calvino, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables (the 2025 contest adds a new set ex100-ex199, README.md lines 19-22), with the AIGs submitted by the teams (README.md line 39: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
40
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 10 inputs and 10 outputs, simulating all 2^10 input patterns reproduces every output line of benchmarks/2025/ex100.truth exactly, and the AND count (header field A) is not above 156. The simulator was checked this round on submissions/NUP_JetBrains/2025/ex100.aig (reproduces the truth table, AND count 156); the 4 team AIGs of this case all simulate to the truth table exactly (checked this round). Hide submissions/ when posing.
41
+ Companions (not included): README.md; LICENSE; IWLS_2025_Programming_Contest.pdf; submissions/NUP_JetBrains/2025/ex100.aig; submissions/UCB_NTU/2025/ex100.aig; submissions/UCLA/2025/ex100.aig; submissions/UF_UL/2025/ex100.aig
42
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
43
+ - `benchmarks/2025/ex136.truth`: IWLS 2025 Programming Contest ex136: truth table (10 inputs, 6 outputs) to synthesise into an AIG with the fewest two-input AND nodes, best known 172.
44
+ Format: text truth table: 6 lines, one per output, each a string of 2^10 = 1024 characters '0'/'1'; the leftmost character is the output for the all-ones input pattern, input k is bit k of the pattern index (convention checked this round by simulating the published AIGs)
45
+ Origin: IWLS 2025 Programming Contest (contest chairs Alan Mishchenko and Alessandro Tempia Calvino, README.md line 13): 100 completely specified multi-output Boolean functions as truth tables (the 2025 contest adds a new set ex100-ex199, README.md lines 19-22), with the AIGs submitted by the teams (README.md line 39: published with the participants' permission); recorded tool_fixture (official contest instance, spec 115 general rule 6)
46
+ Check: Program: the agent returns a binary AIGER file with only two-input AND nodes; pass when it has 10 inputs and 6 outputs, simulating all 2^10 input patterns reproduces every output line of benchmarks/2025/ex136.truth exactly, and the AND count (header field A) is not above 172. The simulator was checked this round on submissions/UCB_NTU/2025/ex136.aig (reproduces the truth table, AND count 172); the 4 team AIGs of this case all simulate to the truth table exactly (checked this round). Hide submissions/ when posing.
47
+ Companions (not included): README.md; LICENSE; IWLS_2025_Programming_Contest.pdf; submissions/NUP_JetBrains/2025/ex136.aig; submissions/UCB_NTU/2025/ex136.aig; submissions/UCLA/2025/ex136.aig; submissions/UF_UL/2025/ex136.aig
48
+ Dependencies: optionally ABC commit a3001b72edc5de22442e942165487fddf150e3d0 (read_truth and cec) for a second check, not packed
campaign_files/2026-10-g-solver-instances-2/jinnaiyuu-combinatorial_instances/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://raw.githubusercontent.com/jinnaiyuu/combinatorial_instances/22c5395343ae7537e6d90bbea8e7e37cc8f8f929/LICENSE (publisher: Yuu Jinnai (repository owner; AUTHORS names Ethan Burns and Sofia Lemons as the 2014 PBNF authors))
campaign_files/2026-10-g-solver-instances-2/jinnaiyuu-combinatorial_instances/README.md ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # jinnaiyuu-combinatorial_instances
2
+
3
+ **What this is:** Files published by Yuu Jinnai (repository owner; AUTHORS names Ethan Burns and Sofia Lemons as the 2014 PBNF authors). Collection theme: Public puzzle sets that ship with solutions (sudoku, nonogram, logic puzzle collections).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Plain text, LF; 100 lines, each 16 space-separated tile numbers 0 to 15 (0 the blank), row-major; a case is one line; delivered whole as the upstream file plus LICENSE; plain text, LF; 1,000 lines of 16 tile numbers, row-major, 0 the blank; a case is one line; delivered whole as the upstream file plus LICENSE; plain text, LF; 100 lines of 25 tile numbers, row-major, 0 the blank; 97 distinct lines (3 repeats dropped at assembly); a case is one line; delivered whole as the upstream file plus LICENSE; plain text, LF; 1,000 lines of 25 tile numbers, row-major, 0 the blank; a case is one line; delivered whole as the upstream file plus LICENSE; plain text, LF; 50 lines of 25 tile numbers, row-major, 0 the blank; a case is one line; delivered whole as the upstream file plus LICENSE.
6
+
7
+ **Where it comes from:** Upstream: https://github.com/jinnaiyuu/combinatorial_instances. Kind: test or example files shipped with a tool. Korf's 100 15-puzzle instances (Korf 1985), a solver benchmark set; general rule 8 lets solver benchmark instances in for line G; randomly generated by the PBNF authors' instance generator (README: 15-puzzle includes randomly generated instances); randomly generated 24-puzzle instances (README: 24-puzzle includes randomly generated instances); randomly generated 24-puzzle instances (README); directory from_Korf; at least partly the Korf and Felner 24-puzzle benchmark instances, a solver benchmark set (general rule 8 lets these in for line G).
8
+
9
+ **Self-contained:** Each file is self-contained. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Check: Replay the agent's move sequence (U, D, L, R moves of the blank) from the start state; correct when every move is legal and the final state is the goal 0 1 2 ... with the blank in the top-left cell. Any legal sequence that reaches the goal counts (rule check by simulation, SPEC line 283), so no uniqueness check is needed.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/jinnaiyuu/combinatorial_instances
16
+
17
+ Files:
18
+
19
+ - `15puzzle/korf_instances`: korf_instances: 100 15-puzzle instances (Korf 1985 set, reordered).
20
+ Format: plain text, LF; 100 lines, each 16 space-separated tile numbers 0 to 15 (0 the blank), row-major; a case is one line; delivered whole as the upstream file plus LICENSE
21
+ Origin: Korf's 100 15-puzzle instances (Korf 1985), a solver benchmark set; general rule 8 lets solver benchmark instances in for line G
22
+ Check: Replay the agent's move sequence (U, D, L, R moves of the blank) from the start state; correct when every move is legal and the final state is the goal 0 1 2 ... with the blank in the top-left cell. Any legal sequence that reaches the goal counts (rule check by simulation, SPEC line 283), so no uniqueness check is needed.
23
+ Companions (not included): LICENSE
24
+ Dependencies: not packed; not packed
25
+ - `15puzzle/random_instances`: random_instances: 1,000 random 15-puzzle instances.
26
+ Format: plain text, LF; 1,000 lines of 16 tile numbers, row-major, 0 the blank; a case is one line; delivered whole as the upstream file plus LICENSE
27
+ Origin: randomly generated by the PBNF authors' instance generator (README: 15-puzzle includes randomly generated instances)
28
+ Check: Replay the agent's move sequence (U, D, L, R moves of the blank) from the start state; correct when every move is legal and the final state is the goal 0 1 2 ... with the blank in the top-left cell. Any legal sequence that reaches the goal counts (rule check by simulation, SPEC line 283), so no uniqueness check is needed.
29
+ Companions (not included): LICENSE
30
+ Dependencies: not packed; not packed
31
+ - `24puzzle/100_instances/24_100_instances`: 24_100_instances: 100 lines of 24-puzzle instances (97 distinct).
32
+ Format: plain text, LF; 100 lines of 25 tile numbers, row-major, 0 the blank; 97 distinct lines (3 repeats dropped at assembly); a case is one line; delivered whole as the upstream file plus LICENSE
33
+ Origin: randomly generated 24-puzzle instances (README: 24-puzzle includes randomly generated instances)
34
+ Check: Replay the agent's move sequence (U, D, L, R moves of the blank) from the start state; correct when every move is legal and the final state is the goal 0 1 2 ... with the blank in the top-left cell. Any legal sequence that reaches the goal counts (rule check by simulation, SPEC line 283), so no uniqueness check is needed.
35
+ Companions (not included): LICENSE
36
+ Dependencies: not packed; not packed
37
+ - `24puzzle/24_random_2000`: 24_random_2000: 1,000 random 24-puzzle instances.
38
+ Format: plain text, LF; 1,000 lines of 25 tile numbers, row-major, 0 the blank; a case is one line; delivered whole as the upstream file plus LICENSE
39
+ Origin: randomly generated 24-puzzle instances (README)
40
+ Check: Replay the agent's move sequence (U, D, L, R moves of the blank) from the start state; correct when every move is legal and the final state is the goal 0 1 2 ... with the blank in the top-left cell. Any legal sequence that reaches the goal counts (rule check by simulation, SPEC line 283), so no uniqueness check is needed.
41
+ Companions (not included): LICENSE
42
+ Dependencies: not packed; not packed
43
+ - `24puzzle/from_Korf/24_small_first_instances`: 24_small_first_instances: 50 24-puzzle instances (42 new).
44
+ Format: plain text, LF; 50 lines of 25 tile numbers, row-major, 0 the blank; a case is one line; delivered whole as the upstream file plus LICENSE
45
+ Origin: directory from_Korf; at least partly the Korf and Felner 24-puzzle benchmark instances, a solver benchmark set (general rule 8 lets these in for line G)
46
+ Check: Replay the agent's move sequence (U, D, L, R moves of the blank) from the start state; correct when every move is legal and the final state is the goal 0 1 2 ... with the blank in the top-left cell. Any legal sequence that reaches the goal counts (rule check by simulation, SPEC line 283), so no uniqueness check is needed.
47
+ Companions (not included): LICENSE
48
+ Dependencies: not packed; not packed
campaign_files/2026-10-g-solver-instances-2/kochetov_discrete_location_benchmarks/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: http://old.math.nsc.ru/AP/benchmarks/gen.html (publisher: Sobolev Institute of Mathematics, Novosibirsk (Yu. Kochetov and coauthors), Discrete Location Problems Benchmarks Library)
campaign_files/2026-10-g-solver-instances-2/kochetov_discrete_location_benchmarks/README.md ADDED
@@ -0,0 +1,138 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # kochetov_discrete_location_benchmarks
2
+
3
+ **What this is:** Files published by Sobolev Institute of Mathematics, Novosibirsk (Yu. Kochetov and coauthors), Discrete Location Problems Benchmarks Library. Collection theme: Loading and location logistics instance libraries (bin packing and container loading, facility location, network design, with best known solution or optimal value).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Plain text, CRLF; header lines then "n fixed_cost", then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities; plain text, CRLF; header lines then "n p" (Dimension, Median), then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of at most p open facilities; plain text, CRLF; line 1 code, line 3 "facilities paths clients, fixed cost", line 4 "Pathes" then 100 lines each listing the facilities of one path, line 105 "Transportation", line 106 "Path Client Cost", then sparse triples "path client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities.
6
+
7
+ **Where it comes from:** Upstream origins: named in the provenance line. Kind: test or example files shipped with a tool.
8
+
9
+ **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** http://old.math.nsc.ru/
16
+
17
+ Files:
18
+
19
+ - `AP/benchmarks/UFLP/P_CODES/334PCodes.txt`: UFLP instance 334PCodes (polynomially solvable class with exponential number of strong local optima, n=128).
20
+ Format: plain text, CRLF; header lines then "n fixed_cost", then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
21
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
22
+ Check: Program: read the agent solution (a list of open facility indices); feasible when every client j=1..128 has a listed finite cost g_ij from at least one open facility (a pair absent from the sparse file is forbidden, ben_eng.html file format); objective = 3000 x (number of open facilities) + sum over clients of the least listed g_ij over open facilities (UFLP model of uflp_eng.html, opening cost 3000 for every facility, ben_eng.html line 70); pass when feasible and the integer objective is at most 48232, the published optimal value (proven optimal, so a lower value means a checker error)
23
+ Companions (not included): AP/benchmarks/UFLP/Engl/uflp_pc_eng.html; AP/benchmarks/UFLP/Engl/uflp_eng.html; AP/benchmarks/UFLP/Engl/ben_eng.html
24
+ Dependencies: no upstream code used
25
+ - `AP/benchmarks/UFLP/Chess_board/334Chess.txt`: UFLP instance 334Chess (chess board class, n=144).
26
+ Format: plain text, CRLF; header lines then "n fixed_cost", then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
27
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
28
+ Check: Program: read the agent solution (a list of open facility indices); feasible when every client j=1..144 has a listed finite cost g_ij from at least one open facility (a pair absent from the sparse file is forbidden, ben_eng.html file format); objective = 3000 x (number of open facilities) + sum over clients of the least listed g_ij over open facilities (UFLP model of uflp_eng.html, opening cost 3000 for every facility, ben_eng.html line 70); pass when feasible and the integer objective is at most 48258, the published optimal value (proven optimal, so a lower value means a checker error)
29
+ Companions (not included): AP/benchmarks/UFLP/Engl/uflp_ch_eng.html; AP/benchmarks/UFLP/Engl/uflp_eng.html; AP/benchmarks/UFLP/Engl/ben_eng.html
30
+ Dependencies: no upstream code used
31
+ - `AP/benchmarks/UFLP/Gap-A/332GapA.txt`: UFLP instance 332GapA (large duality gap class Gap-A, n=100).
32
+ Format: plain text, CRLF; header lines then "n fixed_cost", then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
33
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
34
+ Check: Program: read the agent solution (a list of open facility indices); feasible when every client j=1..100 has a listed finite cost g_ij from at least one open facility (a pair absent from the sparse file is forbidden, ben_eng.html file format); objective = 3000 x (number of open facilities) + sum over clients of the least listed g_ij over open facilities (UFLP model of uflp_eng.html, opening cost 3000 for every facility, ben_eng.html line 70); pass when feasible and the integer objective is at most 36154, the published optimal value (proven optimal, so a lower value means a checker error)
35
+ Companions (not included): AP/benchmarks/UFLP/Engl/uflp_dg_eng.html; AP/benchmarks/UFLP/Engl/uflp_eng.html; AP/benchmarks/UFLP/Engl/ben_eng.html
36
+ Dependencies: no upstream code used
37
+ - `AP/benchmarks/UFLP/Euclid/111Eucl.txt`: UFLP instance 111Eucl (Euclidean class, n=100, dense).
38
+ Format: plain text, CRLF; header lines then "n fixed_cost", then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
39
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
40
+ Check: Program: read the agent solution (a list of open facility indices); feasible when every client j=1..100 has a listed finite cost g_ij from at least one open facility (a pair absent from the sparse file is forbidden, ben_eng.html file format); objective = 3000 x (number of open facilities) + sum over clients of the least listed g_ij over open facilities (UFLP model of uflp_eng.html, opening cost 3000 for every facility, ben_eng.html line 70); pass when feasible and the integer objective is at most 96116, the published optimal value (proven optimal, so a lower value means a checker error)
41
+ Companions (not included): AP/benchmarks/UFLP/Engl/uflp_eucl_eng.html; AP/benchmarks/UFLP/Engl/uflp_eng.html; AP/benchmarks/UFLP/Engl/ben_eng.html
42
+ Dependencies: no upstream code used
43
+ - `AP/benchmarks/UFLP/Uniform/123Unif.txt`: UFLP instance 123Unif (uniform random class, n=100, dense).
44
+ Format: plain text, CRLF; header lines then "n fixed_cost", then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
45
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
46
+ Check: Program: read the agent solution (a list of open facility indices); feasible when every client j=1..100 has a listed finite cost g_ij from at least one open facility (a pair absent from the sparse file is forbidden, ben_eng.html file format); objective = 3000 x (number of open facilities) + sum over clients of the least listed g_ij over open facilities (UFLP model of uflp_eng.html, opening cost 3000 for every facility, ben_eng.html line 70); pass when feasible and the integer objective is at most 71342, the published optimal value (proven optimal, so a lower value means a checker error)
47
+ Companions (not included): AP/benchmarks/UFLP/Engl/uflp_unif-eng.html; AP/benchmarks/UFLP/Engl/uflp_eng.html; AP/benchmarks/UFLP/Engl/ben_eng.html
48
+ Dependencies: no upstream code used
49
+ - `AP/benchmarks/UFLP/FPP_11/1FPP_11.txt`: UFLP instance 1FPP_11 (finite projective plane class k=11, n=133).
50
+ Format: plain text, CRLF; header lines then "n fixed_cost", then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
51
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
52
+ Check: Program: read the agent solution (a list of open facility indices); feasible when every client j=1..133 has a listed finite cost g_ij from at least one open facility (a pair absent from the sparse file is forbidden, ben_eng.html file format); objective = 3000 x (number of open facilities) + sum over clients of the least listed g_ij over open facilities (UFLP model of uflp_eng.html, opening cost 3000 for every facility, ben_eng.html line 70); pass when feasible and the integer objective is at most 36230, the published best found value (not proven optimal; lower is an improvement and also passes)
53
+ Companions (not included): AP/benchmarks/UFLP/Engl/uflp_fp_en.html; AP/benchmarks/UFLP/Engl/uflp_eng.html; AP/benchmarks/UFLP/Engl/ben_eng.html
54
+ Dependencies: no upstream code used
55
+ - `AP/benchmarks/P-median/Pm_Pc/334Pm_Pc.txt`: P-median instance 334Pm_Pc (perfect codes class, n=128, p=16).
56
+ Format: plain text, CRLF; header lines then "n p" (Dimension, Median), then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of at most p open facilities
57
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
58
+ Check: Program: read the agent solution (a list of open facility indices); feasible when at most p=16 facilities are open (line 3 of 334Pm_Pc.txt, "Dimension, Median") and every client j=1..128 has a listed finite cost g_ij from an open facility (absent pair = forbidden); objective = sum over clients of the least listed g_ij over open facilities (p-median model of p-med_eng.html, no opening cost); pass when feasible and the integer objective is at most 232, the published optimal value (proven optimal, so a lower value means a checker error)
59
+ Companions (not included): AP/benchmarks/P-median/p-med_pc_eng.html; AP/benchmarks/P-median/p-med_eng.html
60
+ Dependencies: no upstream code used
61
+ - `AP/benchmarks/P-median/PM_Chess/334Pm-Chess.txt`: P-median instance 334Pm-Chess (chess board class, n=144, p=16).
62
+ Format: plain text, CRLF; header lines then "n p" (Dimension, Median), then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of at most p open facilities
63
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
64
+ Check: Program: read the agent solution (a list of open facility indices); feasible when at most p=16 facilities are open (line 3 of 334Pm-Chess.txt, "Dimension, Median") and every client j=1..144 has a listed finite cost g_ij from an open facility (absent pair = forbidden); objective = sum over clients of the least listed g_ij over open facilities (p-median model of p-med_eng.html, no opening cost); pass when feasible and the integer objective is at most 258, the published optimal value (proven optimal, so a lower value means a checker error)
65
+ Companions (not included): AP/benchmarks/P-median/p-med_ch_eng.html; AP/benchmarks/P-median/p-med_eng.html
66
+ Dependencies: no upstream code used
67
+ - `AP/benchmarks/P-median/PM_FP_11/1PM_FP_11.txt`: P-median instance 1PM_FP_11 (finite projective plane class k=11, n=133, p=12).
68
+ Format: plain text, CRLF; header lines then "n p" (Dimension, Median), then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of at most p open facilities
69
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
70
+ Check: Program: read the agent solution (a list of open facility indices); feasible when at most p=12 facilities are open (line 3 of 1PM_FP_11.txt, "Dimension, Median") and every client j=1..133 has a listed finite cost g_ij from an open facility (absent pair = forbidden); objective = sum over clients of the least listed g_ij over open facilities (p-median model of p-med_eng.html, no opening cost); pass when feasible and the integer objective is at most 230, the published optimal value (proven optimal, so a lower value means a checker error)
71
+ Companions (not included): AP/benchmarks/P-median/p-med_fp_en.html; AP/benchmarks/P-median/p-med_eng.html
72
+ Dependencies: no upstream code used
73
+ - `AP/benchmarks/P-median/PM_Gap-A/332PM_GapA.txt`: P-median instance 332PM_GapA (large duality gap class Gap-A, n=100, p=12).
74
+ Format: plain text, CRLF; header lines then "n p" (Dimension, Median), then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of at most p open facilities
75
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
76
+ Check: Program: read the agent solution (a list of open facility indices); feasible when at most p=12 facilities are open (line 3 of 332PM_GapA.txt, "Dimension, Median") and every client j=1..100 has a listed finite cost g_ij from an open facility (absent pair = forbidden); objective = sum over clients of the least listed g_ij over open facilities (p-median model of p-med_eng.html, no opening cost); pass when feasible and the integer objective is at most 154, the published optimal value (proven optimal, so a lower value means a checker error)
77
+ Companions (not included): AP/benchmarks/P-median/p-med_dg_eng.html; AP/benchmarks/P-median/p-med_eng.html
78
+ Dependencies: no upstream code used
79
+ - `AP/benchmarks/MS/R4-gapA/471R4-A.txt`: Multi Stage UFLP instance 471 of class R4-GapA (50 facilities, 100 paths, 100 clients).
80
+ Format: plain text, CRLF; line 1 code, line 3 "facilities paths clients, fixed cost", line 4 "Pathes" then 100 lines each listing the facilities of one path, line 105 "Transportation", line 106 "Path Client Cost", then sparse triples "path client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
81
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
82
+ Check: Program: read the agent solution (a list of open facility indices 1..50); a path is usable when every facility on it (its line among lines 5 to 104 of the file) is open; feasible when every client 1..100 has at least one listed (path, client, cost) triple on a usable path (absent pairs are forbidden); objective = sum of fixed costs of the open facilities (3000 for every facility, line 3 of the file) + sum over clients of the least listed cost over usable paths (objective (1) of ms_eng.html: total cost of opening facilities and servicing the customers); pass when feasible and the integer objective is at most 69122, the published optimal value (proven optimal, so a lower value means a checker error)
83
+ Companions (not included): AP/benchmarks/MS/r4-gap-a_eng.html; AP/benchmarks/MS/ms_eng.html
84
+ Dependencies: no upstream code used
85
+ - `AP/benchmarks/MS/St4-GapC/471St4-C.txt`: Multi Stage UFLP instance 471 of class St4-GapC (50 facilities, 100 paths, 100 clients).
86
+ Format: plain text, CRLF; line 1 code, line 3 "facilities paths clients, fixed cost", line 4 "Pathes" then 100 lines each listing the facilities of one path, line 105 "Transportation", line 106 "Path Client Cost", then sparse triples "path client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
87
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
88
+ Check: Program: read the agent solution (a list of open facility indices 1..50); a path is usable when every facility on it (its line among lines 5 to 104 of the file) is open; feasible when every client 1..100 has at least one listed (path, client, cost) triple on a usable path (absent pairs are forbidden); objective = sum of fixed costs of the open facilities (3000 for every facility, line 3 of the file) + sum over clients of the least listed cost over usable paths (objective (1) of ms_eng.html: total cost of opening facilities and servicing the customers); pass when feasible and the integer objective is at most 78174, the published optimal value (proven optimal, so a lower value means a checker error)
89
+ Companions (not included): AP/benchmarks/MS/st4-gap-c_eng.html; AP/benchmarks/MS/ms_eng.html
90
+ Dependencies: no upstream code used
91
+ - `AP/benchmarks/MS/Uniform/471Unif.txt`: Multi Stage UFLP instance 471 of class R4-Unif (50 facilities, 100 paths, 100 clients).
92
+ Format: plain text, CRLF; line 1 code, line 3 "facilities paths clients, fixed cost", line 4 "Pathes" then 100 lines each listing the facilities of one path, line 105 "Transportation", line 106 "Path Client Cost", then sparse triples "path client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
93
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
94
+ Check: Program: read the agent solution (a list of open facility indices 1..50); a path is usable when every facility on it (its line among lines 5 to 104 of the file) is open; feasible when every client 1..100 has at least one listed (path, client, cost) triple on a usable path (absent pairs are forbidden); objective = sum of fixed costs of the open facilities (3000 for every facility, line 3 of the file) + sum over clients of the least listed cost over usable paths (objective (1) of ms_eng.html: total cost of opening facilities and servicing the customers); pass when feasible and the integer objective is at most 38891, the published optimal value (proven optimal, so a lower value means a checker error)
95
+ Companions (not included): AP/benchmarks/MS/r4-unif_eng.html; AP/benchmarks/MS/ms_eng.html
96
+ Dependencies: no upstream code used
97
+ - `AP/benchmarks/MS/Galax/471Galax.txt`: Multi Stage UFLP instance 471 of class 3-Galax (50 facilities, 100 paths, 100 clients).
98
+ Format: plain text, CRLF; line 1 code, line 3 "facilities paths clients, fixed cost", line 4 "Pathes" then 100 lines each listing the facilities of one path, line 105 "Transportation", line 106 "Path Client Cost", then sparse triples "path client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
99
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
100
+ Check: Program: read the agent solution (a list of open facility indices 1..50); a path is usable when every facility on it (its line among lines 5 to 104 of the file) is open; feasible when every client 1..100 has at least one listed (path, client, cost) triple on a usable path (absent pairs are forbidden); objective = sum of fixed costs of the open facilities (facilities 1 to 48 cost 300 each, facilities 49 and 50 cost 20000 each, line 3 of the file) + sum over clients of the least listed cost over usable paths (objective (1) of ms_eng.html: total cost of opening facilities and servicing the customers); pass when feasible and the integer objective is at most 35789, the published optimal value (proven optimal, so a lower value means a checker error)
101
+ Companions (not included): AP/benchmarks/MS/3galax_eng.html; AP/benchmarks/MS/ms_eng.html
102
+ Dependencies: no upstream code used
103
+ - `AP/benchmarks/UFLP/Gap-B/331GapB.txt`: UFLP instance 331GapB (class Gap-B, n=100).
104
+ Format: plain text, CRLF; header lines then "n fixed_cost", then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
105
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
106
+ Check: Program: read the agent solution (a list of open facility indices); feasible when every client j=1..100 has a listed finite cost g_ij from at least one open facility (a pair absent from the sparse file is forbidden); objective = 3000 x (number of open facilities) + sum over clients of the least listed g_ij over open facilities (UFLP model of uflp_eng.html, opening cost 3000, ben_eng.html line 70); pass when feasible and the integer objective is at most 45123, the published optimal value (proven optimal, so a lower value means a checker error)
107
+ Companions (not included): AP/benchmarks/UFLP/Engl/uflp_dg_eng.html; AP/benchmarks/UFLP/Engl/uflp_eng.html; AP/benchmarks/UFLP/Engl/ben_eng.html
108
+ Dependencies: no upstream code used
109
+ - `AP/benchmarks/UFLP/Gap-C/333GapC.txt`: UFLP instance 333GapC (class Gap-C, n=100).
110
+ Format: plain text, CRLF; header lines then "n fixed_cost", then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
111
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
112
+ Check: Program: read the agent solution (a list of open facility indices); feasible when every client j=1..100 has a listed finite cost g_ij from at least one open facility (a pair absent from the sparse file is forbidden); objective = 3000 x (number of open facilities) + sum over clients of the least listed g_ij over open facilities (UFLP model of uflp_eng.html, opening cost 3000, ben_eng.html line 70); pass when feasible and the integer objective is at most 42147, the published optimal value (proven optimal, so a lower value means a checker error)
113
+ Companions (not included): AP/benchmarks/UFLP/Engl/uflp_dg_eng.html; AP/benchmarks/UFLP/Engl/uflp_eng.html; AP/benchmarks/UFLP/Engl/ben_eng.html
114
+ Dependencies: no upstream code used
115
+ - `AP/benchmarks/UFLP/FPP_17/1FPP_17.txt`: UFLP instance 1FPP_17 (class FPP_17 (finite projective plane, k=17), n=307).
116
+ Format: plain text, CRLF; header lines then "n fixed_cost", then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of open facilities
117
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
118
+ Check: Program: read the agent solution (a list of open facility indices); feasible when every client j=1..307 has a listed finite cost g_ij from at least one open facility (a pair absent from the sparse file is forbidden); objective = 3000 x (number of open facilities) + sum over clients of the least listed g_ij over open facilities (UFLP model of uflp_eng.html, opening cost 3000, ben_eng.html line 70); pass when feasible and the integer objective is at most 54548, the published best found value
119
+ Companions (not included): AP/benchmarks/UFLP/Engl/uflp_fp_en.html; AP/benchmarks/UFLP/Engl/uflp_eng.html; AP/benchmarks/UFLP/Engl/ben_eng.html
120
+ Dependencies: no upstream code used
121
+ - `AP/benchmarks/P-median/PM_Gap-B/431PM_GapB.txt`: P-median instance 431PM_GapB (class PM_Gap-B, n=100, p=15).
122
+ Format: plain text, CRLF; header lines then "n p" (Dimension, Median), then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of at most p open facilities
123
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
124
+ Check: Program: read the agent solution (a list of open facility indices); feasible when at most p=15 facilities are open (line 3 of 431PM_GapB.txt, "Dimension, Median") and every client j=1..100 has a listed finite cost g_ij from an open facility (absent pair = forbidden); objective = sum over clients of the least listed g_ij over open facilities (p-median model of p-med_eng.html, no opening cost); pass when feasible and the integer objective is at most 132, the published optimal value (proven optimal, so a lower value means a checker error)
125
+ Companions (not included): AP/benchmarks/P-median/p-med_dg_eng.html; AP/benchmarks/P-median/p-med_eng.html
126
+ Dependencies: no upstream code used
127
+ - `AP/benchmarks/P-median/PM_Gap-C/433PM_GapC.txt`: P-median instance 433PM_GapC (class PM_Gap-C, n=100, p=14).
128
+ Format: plain text, CRLF; header lines then "n p" (Dimension, Median), then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of at most p open facilities
129
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
130
+ Check: Program: read the agent solution (a list of open facility indices); feasible when at most p=14 facilities are open (line 3 of 433PM_GapC.txt, "Dimension, Median") and every client j=1..100 has a listed finite cost g_ij from an open facility (absent pair = forbidden); objective = sum over clients of the least listed g_ij over open facilities (p-median model of p-med_eng.html, no opening cost); pass when feasible and the integer objective is at most 145, the published optimal value (proven optimal, so a lower value means a checker error)
131
+ Companions (not included): AP/benchmarks/P-median/p-med_dg_eng.html; AP/benchmarks/P-median/p-med_eng.html
132
+ Dependencies: no upstream code used
133
+ - `AP/benchmarks/P-median/PM_FP_17/2PM_FP_17.txt`: P-median instance 2PM_FP_17 (class PM_FP_17 (finite projective plane, k=17), n=307, p=18).
134
+ Format: plain text, CRLF; header lines then "n p" (Dimension, Median), then sparse triples "facility client cost" (pairs not listed are forbidden); one case = one instance; the answer is a set of at most p open facilities
135
+ Origin: published benchmark instance of the Sobolev Institute library, generated by its authors for testing location algorithms; recorded tool_fixture
136
+ Check: Program: read the agent solution (a list of open facility indices); feasible when at most p=18 facilities are open (line 3 of 2PM_FP_17.txt, "Dimension, Median") and every client j=1..307 has a listed finite cost g_ij from an open facility (absent pair = forbidden); objective = sum over clients of the least listed g_ij over open facilities (p-median model of p-med_eng.html, no opening cost); pass when feasible and the integer objective is at most 531, the published optimal value (proven optimal, so a lower value means a checker error)
137
+ Companions (not included): AP/benchmarks/P-median/p-med_fp_en.html; AP/benchmarks/P-median/p-med_eng.html
138
+ Dependencies: no upstream code used
campaign_files/2026-10-g-solver-instances-2/lotsizingdatamodel_benchmarks/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://raw.githubusercontent.com/Lemoine-OR/LotSizingDataModel.Benchmarks/a00ec691da5c3546f732a28ec9c6e7ba5d7137f7/LICENSE.md (publisher: Lemoine-OR (LotSizingDataModel.Benchmarks), converted from the original authors files)
campaign_files/2026-10-g-solver-instances-2/lotsizingdatamodel_benchmarks/README.md ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # lotsizingdatamodel_benchmarks
2
+
3
+ **What this is:** Files published by Lemoine-OR (LotSizingDataModel.Benchmarks), converted from the original authors files. Collection theme: Production planning and lot sizing instance libraries (lot sizing, capacity planning, assembly line balancing and sequencing models, with published optimal values or bounds).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: XML (lotSizingInstance, formatVersion 1.0): items, horizon 30, demand, capacity, setup times and costs, holding costs; one case = one instance; the answer is a production plan per item and period; XML (lotSizingInstance, formatVersion 1.0): 50 items, horizon 8, demand, capacity, setup and holding data; one case = one instance; XML (lotSizingInstance, formatVersion 1.0): items with BOM levels, horizon 12, demand, setup and holding costs; companion solution XML gives the plan; one case = one instance.
6
+
7
+ **Where it comes from:** Upstream: https://github.com/Lemoine-OR/LotSizingDataModel.Benchmarks. Kind: test or example files shipped with a tool. Trigeiro, Thomas and McClain (1989) CLSP instance G62, converted by Lemoine-OR from the original .trig file (sourceInformation of the xml gives the .trig sha256); reference and bound from Belvaux and Wolsey 2000 (CORE DP 2000-39) per the trust csv; recorded tool_fixture; Cattrysse et al. (1990) CLSP test set, converted by Lemoine-OR from the original workbook; reference values LITERATURE_BEST_REPORTED (best of the heuristics reported in the workbook); recorded tool_fixture; Dellaert and Jeunet (2000) phase-1 MLLP instance ph1in01 (Coleman and McKnew lineage, family.json lines 18 to 26), converted by Lemoine-OR; complete CPLEX-type solution file checked by the repository checker (96/96 valid); recorded tool_fixture.
8
+
9
+ **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://github.com/Lemoine-OR/LotSizingDataModel.Benchmarks
16
+
17
+ Files:
18
+
19
+ - `benchmarks/TRIGEIRO1989/instances/LSDM_TRIGEIRO1989_CLSP_6items_30periods_G62.xml`: Trigeiro et al. (1989) CLSP instance G62 (6 items, 30 periods) with reference objective 61746 and lower bound 60946.
20
+ Format: XML (lotSizingInstance, formatVersion 1.0): items, horizon 30, demand, capacity, setup times and costs, holding costs; one case = one instance; the answer is a production plan per item and period
21
+ Origin: Trigeiro, Thomas and McClain (1989) CLSP instance G62, converted by Lemoine-OR from the original .trig file (sourceInformation of the xml gives the .trig sha256); reference and bound from Belvaux and Wolsey 2000 (CORE DP 2000-39) per the trust csv; recorded tool_fixture
22
+ Check: Program: read the agent plan (production and setup per item and period); feasible when inventory balance holds with zero initial and no backlog, every period uses at most the capacity for production plus setup times, and production only where a setup is made; objective = setup costs + holding costs computed from the xml; pass when feasible and objective <= 61746 (the literature reference); a value below the lower bound 60946 means a checker error
23
+ Companions (not included): benchmarks/TRIGEIRO1989/RESULTS.md; benchmarks/TRIGEIRO1989/README.md; catalog/global/GLOBAL-NORMALIZED-TRUST-v1.0.0.csv; LICENSE.md
24
+ Dependencies: LotSizingDataModel.Checker and any solver not run
25
+ - `benchmarks/CATTRYSSE1990/instances/LSDM_CATTRYSSE1990_CLSP_50items_8periods_TEST1.xml`: Cattrysse et al. (1990) CLSP instance TEST1 (50 items, 8 periods) with best reported objective 152601.
26
+ Format: XML (lotSizingInstance, formatVersion 1.0): 50 items, horizon 8, demand, capacity, setup and holding data; one case = one instance
27
+ Origin: Cattrysse et al. (1990) CLSP test set, converted by Lemoine-OR from the original workbook; reference values LITERATURE_BEST_REPORTED (best of the heuristics reported in the workbook); recorded tool_fixture
28
+ Check: Program: read the agent plan; feasible when inventory balance holds without backlog, capacity per period holds for production plus setup times, and production only after a setup; objective from the xml costs; pass when feasible and objective <= 152601 (matches or beats the best reported heuristic)
29
+ Companions (not included): benchmarks/CATTRYSSE1990/RESULTS.md; benchmarks/CATTRYSSE1990/README.md; catalog/global/GLOBAL-NORMALIZED-TRUST-v1.0.0.csv; LICENSE.md
30
+ Dependencies: LotSizingDataModel.Checker and any solver not run
31
+ - `benchmarks/DJ2000/Phase1/instances/LSDM_DJ2000_MLLP_12_unknown_ph1in01st1de1mh0ms0.xml`: Dellaert and Jeunet (2000) MLLP phase-1 instance ph1in01 (5 items, 12 periods, serial BOM) with optimum 336.547875 and a complete solution.
32
+ Format: XML (lotSizingInstance, formatVersion 1.0): items with BOM levels, horizon 12, demand, setup and holding costs; companion solution XML gives the plan; one case = one instance
33
+ Origin: Dellaert and Jeunet (2000) phase-1 MLLP instance ph1in01 (Coleman and McKnew lineage, family.json lines 18 to 26), converted by Lemoine-OR; complete CPLEX-type solution file checked by the repository checker (96/96 valid); recorded tool_fixture
34
+ Check: Program: read the agent plan; feasible when, for every item and period, inventory balance holds with dependent demand from the BOM (parents need components in the same period) and production only after a setup; objective = setup costs + holding costs from the xml; the companion solution xml is used first to self-test the checker (it must give 336.547875); pass when feasible and objective <= 336.547875 + 1e-6
35
+ Companions (not included): benchmarks/DJ2000/Phase1/solutions/LSDM_DJ2000_MLLP_12_unknown_ph1in01st1de1mh0ms0.solver-4fb74dc1-a084-4aa2-9471-5e55bbbfd8cc.solution.xml; benchmarks/DJ2000/Phase1/README.md; benchmarks/DJ2000/metadata/family.json; catalog/global/GLOBAL-NORMALIZED-TRUST-v1.0.0.csv; LICENSE.md
36
+ Dependencies: LotSizingDataModel.Checker and any solver not run
campaign_files/2026-10-g-solver-instances-2/neo_vrp_malaga/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: NOASSERTION
2
+
3
+ - NOASSERTION: https://neo.lcc.uma.es/vrp/wp-content/data/instances/cordeau/C-pvrptw.zip (publisher: NEO research group, University of Malaga (VRP web); instance and solution by Cordeau, Laporte and Mercier (2001), periodic VRP with time windows)
campaign_files/2026-10-g-solver-instances-2/neo_vrp_malaga/README.md ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # neo_vrp_malaga
2
+
3
+ **What this is:** Units published by NEO research group, University of Malaga (VRP web); instance and solution by Cordeau, Laporte and Mercier (2001), periodic VRP with time windows. Collection theme: Vehicle routing and logistics benchmark libraries (CVRP, VRP with time windows and similar, with best known solutions).
4
+
5
+ **What one unit is:** Each unit is one file read out of an upstream archive (the archive's address is in the notes), kept at its upstream path. Format: Cordeau text format: line 1 type m n t, then t lines D Q (one per day), then one line per customer i x y d q f a list e l (row 0 is the depot); one case = one instance; the answer is the day by day routes and their total distance.
6
+
7
+ **Where it comes from:** Upstream: https://neo.lcc.uma.es/vrp/wp-content/data/instances/cordeau/C-pvrptw.zip. Kind: test or example files shipped with a tool. Published benchmark instance (Cordeau, Laporte and Mercier 2001, PVRPTW set) as served by the NEO VRP web; recorded tool_fixture; unit and answer are zip members (spec item 4).
8
+
9
+ **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. Each unit names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
+
11
+ **Tools to open and check:** Check: Program: read the agent plan (Cordeau .res layout: per route day, vehicle, then the customer sequence starting and ending at depot 0); feasible when each of the 48 customers is visited on exactly the days of one of its allowed visit combinations and once on each of those days, at most 3 routes per day, each route load at most 200, service at each customer starts within its window e to l (waiting allowed) and lasts d, the vehicle is back by the depot window end 1000, and each route duration from depot departure (which may be delayed) to return is at most 500; travel time equals unrounded Euclidean distance; cost is the total travel distance excluding service time (index.html line 167); pass when feasible and the total is at most 3007.84 plus 0.005, the best known value (a lower total also passes).
12
+
13
+ **Licence (SPDX):** NOASSERTION
14
+
15
+ **Provenance:** https://neo.lcc.uma.es/vrp/wp-content/data/instances/cordeau/C-pvrptw.zip
16
+
17
+ Units:
18
+
19
+ - `vrp/wp-content/data/instances/cordeau/C-pvrptw/pr01`: Periodic VRP with time windows instance pr01 (Cordeau) with best known solution 3007.84.
20
+ Read out of the archive: https://neo.lcc.uma.es/vrp/wp-content/data/instances/cordeau/C-pvrptw.zip
21
+ Format: Cordeau text format: line 1 type m n t, then t lines D Q (one per day), then one line per customer i x y d q f a list e l (row 0 is the depot); one case = one instance; the answer is the day by day routes and their total distance
22
+ Origin: published benchmark instance (Cordeau, Laporte and Mercier 2001, PVRPTW set) as served by the NEO VRP web; recorded tool_fixture; unit and answer are zip members (spec item 4)
23
+ Check: Program: read the agent plan (Cordeau .res layout: per route day, vehicle, then the customer sequence starting and ending at depot 0); feasible when each of the 48 customers is visited on exactly the days of one of its allowed visit combinations and once on each of those days, at most 3 routes per day, each route load at most 200, service at each customer starts within its window e to l (waiting allowed) and lasts d, the vehicle is back by the depot window end 1000, and each route duration from depot departure (which may be delayed) to return is at most 500; travel time equals unrounded Euclidean distance; cost is the total travel distance excluding service time (index.html line 167); pass when feasible and the total is at most 3007.84 plus 0.005, the best known value (a lower total also passes)
24
+ Companions (not included): vrp/wp-content/data/instances/cordeau/C-pvrptw-sol/pr01.res; vrp/vrp-instances/description-for-files-of-cordeaus-instances/index.html
25
+ Dependencies: no upstream code used
campaign_files/2026-10-g-solver-instances-2/rub_hgi_shorter_linear_slps_mds/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://github.com/rub-hgi/shorter_linear_slps_for_mds_matrices/blob/1de21b4ea01676c1bb8fb6632a7358884911e952/LICENSE (publisher: Thorsten Kranz, Gregor Leander, Ko Stoffelen, Friedrich Wiemer (Ruhr-Universitaet Bochum, Horst Goertz Institute))
campaign_files/2026-10-g-solver-instances-2/rub_hgi_shorter_linear_slps_mds/README.md ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # rub_hgi_shorter_linear_slps_mds
2
+
3
+ **What this is:** Files published by Thorsten Kranz, Gregor Leander, Ko Stoffelen, Friedrich Wiemer (Ruhr-Universitaet Bochum, Horst Goertz Institute). Collection theme: Logic synthesis and optimisation benchmark instances (combinational circuits with best known node counts or levels: EPFL benchmarks, IWLS contests and the like).
4
+
5
+ **What one unit is:** One upstream file per unit, kept at its path in the upstream repository or site. Format: Plain text: first line '1', second line '32 32' (rows, columns), then 32 lines of 32 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP); plain text: first line '1', second line '64 64' (rows, columns), then 64 lines of 64 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP); plain text: first line '1', second line '16 16' (rows, columns), then 16 lines of 16 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP).
6
+
7
+ **Where it comes from:** Upstream: https://github.com/rub-hgi/shorter_linear_slps_for_mds_matrices. Kind: test or example files shipped with a tool. Details per file in the notes.
8
+
9
+ **Self-contained:** Whether each file needs other upstream files is not stated. Each file names companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
+
11
+ **Tools to open and check:** Tools are named in the check written per file in the notes.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/rub-hgi/shorter_linear_slps_for_mds_matrices
16
+
17
+ Files:
18
+
19
+ - `matrices_bp_format/AES.txt`: Shorter linear SLPs for MDS matrices: AES MixColumns as a 32x32 binary matrix to implement with the fewest two-input XOR gates, published 97 XORs.
20
+ Format: plain text: first line '1', second line '32 32' (rows, columns), then 32 lines of 32 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
21
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the AES MixColumns from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
22
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x31 with outputs y0..y31 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 32-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 97. Checked this round on matrices_slp_implementations/AES.txt (0 wrong rows, 97 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/AES.txt when posing.
23
+ Companions (not included): matrices_slp_implementations/AES.txt; README.md; LICENSE
24
+ - `matrices_bp_format/Anubis.txt`: Shorter linear SLPs for MDS matrices: Anubis diffusion layer as a 32x32 binary matrix to implement with the fewest two-input XOR gates, published 113 XORs.
25
+ Format: plain text: first line '1', second line '32 32' (rows, columns), then 32 lines of 32 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
26
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the Anubis diffusion layer from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
27
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x31 with outputs y0..y31 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 32-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 113. Checked this round on matrices_slp_implementations/Anubis.txt (0 wrong rows, 113 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/Anubis.txt when posing.
28
+ Companions (not included): matrices_slp_implementations/Anubis.txt; README.md; LICENSE
29
+ - `matrices_bp_format/Clefia_M0.txt`: Shorter linear SLPs for MDS matrices: CLEFIA diffusion matrix M0 as a 32x32 binary matrix to implement with the fewest two-input XOR gates, published 106 XORs.
30
+ Format: plain text: first line '1', second line '32 32' (rows, columns), then 32 lines of 32 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
31
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the CLEFIA diffusion matrix M0 from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
32
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x31 with outputs y0..y31 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 32-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 106. Checked this round on matrices_slp_implementations/Clefia_M0.txt (0 wrong rows, 106 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/Clefia_M0.txt when posing.
33
+ Companions (not included): matrices_slp_implementations/Clefia_M0.txt; README.md; LICENSE
34
+ - `matrices_bp_format/Grostl.txt`: Shorter linear SLPs for MDS matrices: Groestl MixBytes as a 64x64 binary matrix to implement with the fewest two-input XOR gates, published 475 XORs.
35
+ Format: plain text: first line '1', second line '64 64' (rows, columns), then 64 lines of 64 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
36
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the Groestl MixBytes from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
37
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x63 with outputs y0..y63 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 64-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 475. Checked this round on matrices_slp_implementations/Grostl.txt (0 wrong rows, 475 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/Grostl.txt when posing.
38
+ Companions (not included): matrices_slp_implementations/Grostl.txt; README.md; LICENSE
39
+ - `matrices_bp_format/Joltik.txt`: Shorter linear SLPs for MDS matrices: Joltik MixColumns as a 16x16 binary matrix to implement with the fewest two-input XOR gates, published 48 XORs.
40
+ Format: plain text: first line '1', second line '16 16' (rows, columns), then 16 lines of 16 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
41
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the Joltik MixColumns from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
42
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x15 with outputs y0..y15 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 16-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 48. Checked this round on matrices_slp_implementations/Joltik.txt (0 wrong rows, 48 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/Joltik.txt when posing.
43
+ Companions (not included): matrices_slp_implementations/Joltik.txt; README.md; LICENSE
44
+ - `matrices_bp_format/Khazad.txt`: Shorter linear SLPs for MDS matrices: Khazad diffusion layer as a 64x64 binary matrix to implement with the fewest two-input XOR gates, published 507 XORs.
45
+ Format: plain text: first line '1', second line '64 64' (rows, columns), then 64 lines of 64 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
46
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the Khazad diffusion layer from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
47
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x63 with outputs y0..y63 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 64-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 507. Checked this round on matrices_slp_implementations/Khazad.txt (0 wrong rows, 507 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/Khazad.txt when posing.
48
+ Companions (not included): matrices_slp_implementations/Khazad.txt; README.md; LICENSE
49
+ - `matrices_bp_format/PRINCE_M_0.txt`: Shorter linear SLPs for MDS matrices: PRINCE M0 block as a 16x16 binary matrix to implement with the fewest two-input XOR gates, published 24 XORs.
50
+ Format: plain text: first line '1', second line '16 16' (rows, columns), then 16 lines of 16 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
51
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the PRINCE M0 block from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
52
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x15 with outputs y0..y15 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 16-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 24. Checked this round on matrices_slp_implementations/PRINCE_M_0.txt (0 wrong rows, 24 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/PRINCE_M_0.txt when posing.
53
+ Companions (not included): matrices_slp_implementations/PRINCE_M_0.txt; README.md; LICENSE
54
+ - `matrices_bp_format/QARMA128.txt`: Shorter linear SLPs for MDS matrices: QARMA-128 MixColumns as a 32x32 binary matrix to implement with the fewest two-input XOR gates, published 48 XORs.
55
+ Format: plain text: first line '1', second line '32 32' (rows, columns), then 32 lines of 32 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
56
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the QARMA-128 MixColumns from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
57
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x31 with outputs y0..y31 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 32-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 48. Checked this round on matrices_slp_implementations/QARMA128.txt (0 wrong rows, 48 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/QARMA128.txt when posing.
58
+ Companions (not included): matrices_slp_implementations/QARMA128.txt; README.md; LICENSE
59
+ - `matrices_bp_format/Twofish.txt`: Shorter linear SLPs for MDS matrices: Twofish MDS matrix as a 32x32 binary matrix to implement with the fewest two-input XOR gates, published 129 XORs.
60
+ Format: plain text: first line '1', second line '32 32' (rows, columns), then 32 lines of 32 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
61
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the Twofish MDS matrix from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
62
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x31 with outputs y0..y31 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 32-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 129. Checked this round on matrices_slp_implementations/Twofish.txt (0 wrong rows, 129 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/Twofish.txt when posing.
63
+ Companions (not included): matrices_slp_implementations/Twofish.txt; README.md; LICENSE
64
+ - `matrices_bp_format/Whirlpool.txt`: Shorter linear SLPs for MDS matrices: Whirlpool MixRows as a 64x64 binary matrix to implement with the fewest two-input XOR gates, published 465 XORs.
65
+ Format: plain text: first line '1', second line '64 64' (rows, columns), then 64 lines of 64 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
66
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the Whirlpool MixRows from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
67
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x63 with outputs y0..y63 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 64-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 465. Checked this round on matrices_slp_implementations/Whirlpool.txt (0 wrong rows, 465 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/Whirlpool.txt when posing.
68
+ Companions (not included): matrices_slp_implementations/Whirlpool.txt; README.md; LICENSE
69
+ - `matrices_bp_format/Fox_Mu4.txt`: Shorter linear SLPs for MDS matrices: FOX mu4 diffusion as a 32x32 binary matrix to implement with the fewest two-input XOR gates, published 137 XORs.
70
+ Format: plain text: first line '1', second line '32 32' (rows, columns), then 32 lines of 32 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
71
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the FOX mu4 diffusion from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
72
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x31 with outputs y0..y31 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 32-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 137. Checked this round on matrices_slp_implementations/Fox_Mu4.txt (0 wrong rows, 137 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/Fox_Mu4.txt when posing.
73
+ Companions (not included): matrices_slp_implementations/Fox_Mu4.txt; README.md; LICENSE
74
+ - `matrices_bp_format/SmallScale_AES.txt`: Shorter linear SLPs for MDS matrices: small-scale AES MixColumns as a 16x16 binary matrix to implement with the fewest two-input XOR gates, published 47 XORs.
75
+ Format: plain text: first line '1', second line '16 16' (rows, columns), then 16 lines of 16 space-separated 0/1 entries; row i gives output y_i as the XOR of the inputs x_j whose entry is 1 (the input format of the authors' slp_heuristic, README.md lines 10-13; row and column convention checked this round by evaluating the published SLP)
76
+ Origin: matrix file shipped with the paper 'Shorter Linear Straight-Line Programs for MDS Matrices' (Kranz, Leander, Stoffelen, Wiemer, IACR ToSC 2017 issue 4; README.md line 3) as the input of the authors' slp_heuristic program (README.md lines 10-13); the binary matrix of the small-scale AES MixColumns from the cipher specification; recorded tool_fixture (benchmark instance shipped with a published optimisation tool, spec 115 general rule 6)
77
+ Check: Program: the agent returns a straight-line program of lines 'v = a + b' (two-input XOR over GF(2)) on inputs x0..x15 with outputs y0..y15 (plain copies 'y = t' cost nothing); pass when every output equals the XOR of the inputs selected by its matrix row (each output compared as a 16-bit vector, which covers all inputs at once because the map is linear) and the number of XOR lines is not above 47. Checked this round on matrices_slp_implementations/SmallScale_AES.txt (0 wrong rows, 47 XORs). Later papers report fewer XORs for some of these matrices; the bound here is this source's published value. Hide matrices_slp_implementations/SmallScale_AES.txt when posing.
78
+ Companions (not included): matrices_slp_implementations/SmallScale_AES.txt; README.md; LICENSE