Issues

• Add calibrated uncertainty contract for decision-authorizing predictions

  enhancementpriority: high
  ContextualWisdomLab/fast-mlsirm#2315 · Filter by author seonghobae opened on Oct 1, 2026
  3

• GPU EM: f32 E-step loglik feeds the f64-epsilon monotonicity guard and tol convergence (bifactor #1931, two-tier #2291)

  bugpriority: high
  ContextualWisdomLab/fast-mlsirm#2292 · Filter by author seonghobae opened on Sep 30, 2026
  1

• M2/RMSEA2 for multigroup polytomous GRM (fixed-item focal groups) and multigroup bifactor GRM

  enhancementpriority: high
  ContextualWisdomLab/fast-mlsirm#2290 · Filter by author seonghobae opened on Sep 30, 2026

• CHANGELOG.md authoritative block is stale on main (license-evidence fragment; --update duplicates heading)

  bugpriority: high
  ContextualWisdomLab/fast-mlsirm#2289 · Filter by author seonghobae opened on Sep 30, 2026

• perf(gpu): wgpu path for the two-tier GRM EM E-step (reusing the bifactor reduced kernels)

  ContextualWisdomLab/fast-mlsirm#2282 · Filter by author seonghobae opened on Sep 29, 2026

• joint five-scale GPU 부트스트랩이 s1에서 Mac보다 4배 이상 느림

  ContextualWisdomLab/fast-mlsirm#2278 · Filter by author seonghobae opened on Sep 29, 2026
  2

• two-tier LW recursion: reject non-decreasing GRM thresholds in TwoTierItemParams::check_dims

  ContextualWisdomLab/fast-mlsirm#2277 · Filter by author seonghobae opened on Sep 29, 2026

• GPU path for two-tier expected-raw scoring (Lord-Wingersky 2.0, conditional on primary)

  ContextualWisdomLab/fast-mlsirm#2271 · Filter by author seonghobae opened on Sep 29, 2026
  2

• perf(gpu): wgpu path for polytomous_predictions (GRM/GPCM expected scores)

  ContextualWisdomLab/fast-mlsirm#2270 · Filter by author seonghobae opened on Sep 28, 2026
  1

• Two-tier fixed-item focal Gaussian calibration and person scoring

  ContextualWisdomLab/fast-mlsirm#2209 · Filter by author seonghobae opened on Sep 27, 2026
  32

• feat(inference): library-owned GRM report transforms and information condition diagnostics

  ContextualWisdomLab/fast-mlsirm#2200 · Filter by author seonghobae opened on Sep 27, 2026
  7

• Fix nonfinite high-node focal expected-score quadrature

  bugpriority: high
  ContextualWisdomLab/fast-mlsirm#2192 · Filter by author seonghobae opened on Sep 27, 2026

• license_inventory: Windows backslash notice paths never bind (GPL+GCC-exception label missing; stays HOLD)

  dependenciessecurity
  ContextualWisdomLab/fast-mlsirm#2153 · Filter by author seonghobae opened on Sep 24, 2026

• Add a CI license gate (cargo-deny + Python lock check) and close remaining license/lock gaps

  area: ci-cdarea: dependenciesdependencies
  ContextualWisdomLab/fast-mlsirm#2138 · Filter by author seonghobae opened on Sep 23, 2026

• Wheels ship no third-party license notices for 77-89 statically linked crates

  area: dependenciesdependenciessecurity
  ContextualWisdomLab/fast-mlsirm#2137 · Filter by author seonghobae opened on Sep 23, 2026

• NumPy wheels bundle LGPL-2.1+ libquadmath and GCC-exception runtime: owner decision and Fortran-free replacement path

  area: dependenciesdependenciespriority: highsecurity
  ContextualWisdomLab/fast-mlsirm#2136 · Filter by author seonghobae opened on Sep 23, 2026
  6

• test(changelog): cover the empty-inventory marker-stripping branch in render_changelog_fragments.py

  ContextualWisdomLab/fast-mlsirm#2127 · Filter by author seonghobae opened on Sep 22, 2026

• GPCM NumPy parity reduction: bounded-memory optimization with real benchmark evidence

  ContextualWisdomLab/fast-mlsirm#2125 · Filter by author seonghobae opened on Sep 22, 2026

• feat(polytomous): report final parameter change and optional param_tol stopping rule for EM fits

  ContextualWisdomLab/fast-mlsirm#2122 · Filter by author seonghobae opened on Sep 22, 2026
  2

• docs(polytomous): EM tol is relative (tol*(1+|loglik|)); document and test stopping contract

  ContextualWisdomLab/fast-mlsirm#2121 · Filter by author seonghobae opened on Sep 22, 2026

• feat(bifactor): joint multigroup Oakes observed information (items + group mean/variance), ML and MAP

  ContextualWisdomLab/fast-mlsirm#2113 · Filter by author seonghobae opened on Sep 21, 2026
  2

• fix(estimators): NumPy mmle/marginal Gauss-Hermite rules return NaN weights for q >= 371 (separate from #2109)

  ContextualWisdomLab/fast-mlsirm#2110 · Filter by author seonghobae opened on Sep 21, 2026

• test governance: migrate legacy deprecated API coverage to current public entrypoints

  ContextualWisdomLab/fast-mlsirm#2104 · Filter by author seonghobae opened on Sep 21, 2026

• test governance: remove remaining non-execution outcomes from required Python CI

  ContextualWisdomLab/fast-mlsirm#2103 · Filter by author seonghobae opened on Sep 21, 2026

• fit_bifactor_grm_multigroup (MML, no prior): EM observed-data log-likelihood decreases on a 12-person two-group fixture (pre-existing, reproduces on 0.11.4)

  ContextualWisdomLab/fast-mlsirm#2093 · Filter by author seonghobae opened on Sep 21, 2026
  1