Quicopt Benchmarks

Public benchmark results for Quicopt, an optimization solver suite, on standard problem sets. The summary below has one row per problem family; the full per-instance results live in docs/.

• What this is: benchmark results across standard problem sets, with the solver identified by version (Quicopt vX.Y).
• Data: machine-readable in data/ — results.json (all problem types) and one <type>.csv per family. The summary and the per-problem pages are generated from those files by render.py; edit the data, not the prose tables.
• Per-instance columns vary by family: typically the instance and its size, the objective Quicopt achieved, our-hardware wall-time, and — where a confident literature value exists — the best-known / optimum, its source, and %_of_best (100 = matched). A family with no published reference at these sizes (e.g. LABS) reports the field's native quality metric instead.

Results by problem family

problem	instances	objective	Quicopt result
Gset — Maximum Cut	71	cut (max)	52/71 graded — median 100.0% of best-known
LABS — Low-Autocorrelation Binary Sequences	99	sidelobe energy (min)	99/99 graded — median 100.0% of best-known
Maximum Independent Set	50	set size (max)	50/50 graded — median 100.0% of best-known

Full per-instance tables are in docs/.

---

_{Summary and per-problem pages auto-generated from data/ by render.py (run in CI on every data change). Where a family carries reference values, they are sourced per row and drift as new records are published (see that family's data/<type>.csv). Reference values are attributions to third-party results, not Quicopt output.}