Open Science & Auditability
Corrections & Benchmark Dispute Protocol
Maintaining 100% verifiable accuracy across all AI benchmark runs and DOI verification audits.
At DeepResearcher, we adhere to scientific standards of open reproducibility. If you are an academic researcher, tool developer, or student who identifies an inaccuracy in our benchmark reports, citation audits, or feature matrices, we welcome your formal audit submission.
Dispute & Correction Categories
1. Citation Audit InaccuracyA cited paper was flagged as hallucinated or unsupported, but you have proof of a valid matching DOI or peer-reviewed publication.
2. Pricing & Feature ChangesA tool has introduced new pricing tiers, student discounts, Zotero integrations, or updated free trial limits.
3. Model Architecture UpgradesA major model version (e.g., v3 release) has rendered an older benchmark run obsolete, and a standardized re-test is requested.
4. Prompt Replication FailureAn external laboratory failed to replicate the latency or retrieval metrics reported in an audited benchmark run.
How to Submit a Correction
Please send an email to [email protected] with the subject line format:[Benchmark Dispute] - [Task Code / Tool Name] - [Summary]
Please include:
- The Benchmark Run ID (e.g.,
RUN-MED-014) or Tool Profile URL. - A direct link to the DOI, CrossRef record, or official documentation supporting the correction.
- A raw reproduction log or timestamped output snippet.
Investigation SLA & Changelog Policy
All submissions are reviewed by our evaluation committee within 5 business days. If verified:
- The benchmark run and dimension scores are recalculated immediately using the deterministic scoring formula.
- A public correction note is permanently appended to the respective benchmark run detail page.
- The leaderboard version tag is incremented (e.g., from
2026.10-v1to2026.10-v2).