docs(readme): propagate Judge accuracy framing across all LongMemEval sections

[jaylfc]

Follow-up to #31 which clarified the intro headline. Three more README sections still labelled our 97.0% as Recall@5:

Section	Before	After
Benchmark Results table (line 158)	Single "Recall@5" column for all systems	Split into Score + Metric columns — per-row honest labelling
Fusion Strategy Comparison (line 181)	Column header "Recall@5"	"Judge accuracy" (all strategies use the Judge harness)
Key Features list (line 263)	"97.0% Recall@5"	"97.0% end-to-end Judge accuracy"

Also softened "MemPalace-equivalent" to "same algorithm as MemPalace" — that row describes the algorithm, not the metric, which differs between the two systems.

Remaining Recall@5 mentions in the README are intentional: competitors' published numbers, the paragraph explaining why direct comparison isn't apples-to-apples, and the code-block comment labelling benchmarks/longmemeval_recall.py (which is the Recall@5 reproduction script).

Summary by CodeRabbit

• Documentation
  • Updated benchmarking tables and narrative to report end-to-end Judge accuracy as the primary metric instead of Recall@5.
  • Clarified measurement differences across systems and added references to benchmark scripts for reproducibility.
  • Updated feature descriptions and fusion strategy comparison metrics accordingly.

[coderabbitai]

Caution

Review failed

Pull request was closed or merged during review

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 68770700-ed8d-45df-92ac-02dd3d962c14

📥 Commits

Reviewing files that changed from the base of the PR and between 116edab and a893cf1.

📒 Files selected for processing (1)

• README.md

---

📝 Walkthrough

Walkthrough

Updated README benchmarking tables and descriptions to report end-to-end Judge accuracy as the primary metric for taOSmd instead of Recall@5, clarified metric differences with other systems, and added references to specific benchmark scripts for reproducibility.

Changes

Cohort / File(s)	Summary
README Benchmark Documentation README.md	Shifted primary metric reporting from Recall@5 to end-to-end Judge accuracy for taOSmd; updated fusion strategy comparison table headers and metric labels; clarified benchmark methodology differences between taOSmd and other systems; added references to benchmark harness scripts; updated Key Features bullet point accordingly.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

Poem

🐰 Metrics shift like morning dew,

Judge accuracy shines bright and true,

From Recall to end-to-end flow,

The benchmarks' honest truth we show,

Clarity hops through every row! 📊

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and specifically summarizes the main change: propagating Judge accuracy framing across README sections, which matches the PR objectives of replacing misleading Recall@5 labels with accurate Judge accuracy descriptions.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/readme-judge-accuracy-consistency

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 files)

• README.md - 0 issues

---

_{Reviewed by grok-code-fast-1:optimized:free · 195,378 tokens}