Codecov Report

❌ Patch coverage is 99.57627% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 97.64%. Comparing base (ccf12f4) to head (8926ad3).
⚠️ Report is 1 commits behind head on main.

@@            Coverage Diff             @@
##             main     #755      +/-   ##
==========================================
- Coverage   97.64%   97.64%   -0.01%     
==========================================
  Files         449      451       +2     
  Lines       55134    55370     +236     
==========================================
+ Hits        53834    54064     +230     
- Misses       1300     1306       +6     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Changes

• Added the new PartialFeedbackEdgeSet<G> satisfaction model, schema registration, bounded-cycle validation logic, and a canonical example matching the corrected issue discussion.
• Wired the model into library exports and the CLI create flow, including the new --max-cycle-length input.
• Added focused unit and CLI tests, and documented the model in the paper.
• Verified with make test, make clippy, and make paper.

Deviations from Plan

• I did not enforce max_cycle_length >= 3. For bounds below 3, the implementation treats the instance as vacuously satisfiable within budget because simple graphs have no cycles of length 1 or 2.

Open Questions

• None.

Agentic Review Report

Structural Check

Deterministic packet check: Whitelist FAIL, Completeness PASS.

Confirmed whitelist violations for a model PR:

• problemreductions-cli/src/cli.rs
• problemreductions-cli/src/commands/create.rs
• src/lib.rs

Build Status

• make test: PASS
• make clippy: PASS

Semantic Review

• evaluate() matches the decision problem and rejects invalid config length, non-binary entries, and over-budget removals in src/models/graph/partial_feedback_edge_set.rs:85.
• dims() returns the expected binary configuration space for each edge in src/models/graph/partial_feedback_edge_set.rs:105.
• The short-cycle detector is consistent with the bounded-cycle formulation and the simple-graph assumption in src/models/graph/partial_feedback_edge_set.rs:125.
• CLI creation wiring and help text are present in problemreductions-cli/src/commands/create.rs:1359 and problemreductions-cli/src/cli.rs:268.
• The canonical example and paper entry are aligned with the checked-in model example in docs/paper/reductions.typ:4821.

Issue Compliance

Summary

• 16/16 structural checks passed.
• 8/8 issue-compliance checks passed.
• Build is green: make test and make clippy both passed.
• No blacklisted generated files were committed.
• The only non-passing deterministic signal is the model-PR whitelist violation above.

Quality Check

Design Principles

• DRY: OK
• KISS: OK
• HC/LC: OK

HCI

• Error messages: ISSUE - The invalid-budget path omits the supplied value, so it is slightly less actionable than the strongest existing CLI errors. See problemreductions-cli/src/commands/create.rs:1371.
• Discoverability: OK - PartialFeedbackEdgeSet and --max-cycle-length are surfaced in pred create help and after-help. See problemreductions-cli/src/cli.rs:268 and problemreductions-cli/src/cli.rs:616.
• Consistency: OK
• Least surprise: OK
• Feedback: OK

Test Quality

• Naive test detection: ISSUE - test_create_partial_feedback_edge_set_json writes to a fixed path under std::env::temp_dir(), which is a potential collision point under concurrent test runs. See problemreductions-cli/src/commands/create.rs:7508.
• Boundary coverage: ISSUE - the new model tests do not exercise the documented max_cycle_length < 3 branch in has_cycle_with_length_at_most(). See src/models/graph/partial_feedback_edge_set.rs:130 and src/unit_tests/models/graph/partial_feedback_edge_set.rs:53.

Issues

Important

• The new create-path test uses a fixed temp filename, which is a possible flake risk in parallel execution. See problemreductions-cli/src/commands/create.rs:7508.

Minor

• The invalid-budget CLI error could include the supplied value for easier debugging. See problemreductions-cli/src/commands/create.rs:1371.
• The suite should cover the max_cycle_length < 3 boundary explicitly. See src/models/graph/partial_feedback_edge_set.rs:130.

Summary

• 0 critical issues.
• 1 important issue.
• 2 minor issues.

Agentic Feature Tests

PartialFeedbackEdgeSet was discoverable and usable end to end from the CLI in this review worktree. I exercised the requested user path: pred list, pred show PartialFeedbackEdgeSet, pred create --example PartialFeedbackEdgeSet, and pred solve on the generated example.

Per-Feature Details

• Role: downstream CLI user.
• Use case: discover the model, inspect it, create the canonical example instance, and solve that instance from docs/CLI alone.
• What I tried: pred list, pred show PartialFeedbackEdgeSet, pred create --example PartialFeedbackEdgeSet, and pred solve <generated-instance> --solver brute-force.
• Discoverability: yes. The model appears in pred list as PartialFeedbackEdgeSet/SimpleGraph, and pred show PartialFeedbackEdgeSet prints the model summary and fields.
• Setup: yes. The pred binary ran successfully from the review worktree.
• Functionality: yes. pred list showed PartialFeedbackEdgeSet/SimpleGraph with O(2^num_edges), pred show displayed the expected description and fields, pred create --example wrote a JSON instance, and pred solve returned evaluation: true with a valid solution vector.
• Expected vs Actual: matched.
• Blocked steps: none.
• Friction points: none observed in this run.
• Doc suggestions: none from this test.

Issues Found

• None.

Generated by review-pipeline