Manuscript Intake & Processing
Connect a wallet or link ORCID in “Your account” to save a research profile, get a tailored RiBD, keep your piQ balance, and have diagnostics framed against your own goals. You can still assess manuscripts on the free trial without signing in.
Assess a Manuscript
Search open-access literature (via OpenAlex) and pick papers to feed straight into the pipeline below — no manual download needed.
Assessment Results
Explore
The journal
Papers that have been reviewed, and papers their authors then published. Review comes first: nothing here can be published without a completed peer review, so a published paper is always a reviewed one. The two are still shown in separate columns — a paper can be reviewed without being published, and the claims are not the same.
pi-Index (piX) Leaderboard [Top Papers]
Individual manuscripts ranked by their composite Pi-Index score. Select any row to open its full assessment report and dossier.
| # | Title | Author | piX | piQ | Logic | Date |
|---|
pi-Quotient (piQ) Leaderboard [Top Authors]
Researchers ranked by total piQ earned across all of their assessed manuscripts. Select a row to see that author's papers.
| # | Author | piQ | Papers | Avg piX |
|---|
Proof-of-Research Ledger Explorer
Performance & Analytics
The Global Map of Science
Loading the map…
Forecast
Where the corpus is heading, and how well the three engines behind it are performing. The chart projects how the eight Pi-Index criteria weights will shift in the next ledger epoch, learned from the weights recorded in every Proof-of-Research block.
Modifiers
These change how the projection is computed and drawn, not what was measured. The recorded blocks are the same in every view.
Projected Criteria Weights
Framework Architecture
Pi-Index filters noise and yields quantitative results strictly aligned with Responsible Research Assessment (RRA) and CoARA (Coalition for Advancing Research Assessment) guidelines.
End-to-end processing flow: intake and identity, extraction and deterministic scoring, the independent model panel, PiDN adjudication, and Proof-of-Research settlement on Sepolia.
Whitepaper
The full design rationale, the rubric in detail with every weight, the threat model, and an unmitigated limitations section.
Read it here
Loading…
Scoring Pipeline in Detail
Stage Reference
| Stage | Component | What Happens |
|---|---|---|
| 1 | Unified Intake & ZK-Identity Registry | ORCID iD / W3C DID verification, local PDF, Unpaywall DOI resolution and OpenAlex discovery, ZK double-blind assignment. |
| 2 | Extraction & Deterministic Scoring | PyMuPDF layout-aware text extraction, MDAR adherence, RRID validation, reproducibility markers, empirical density. |
| 3 | Independent Model Panel | Llama, Mistral, Qwen and Gemini each assess the manuscript separately, alongside the local SciM structural analyser. |
| 4 | PiDN Adjudication | Weighs the panel's independent verdicts into one judgement, and records the criteria weighting this paper implies — the input the PiDN forecast is built from. Cross-model evidence synthesis, inter-model agreement measurement, judgement-quality grading, adversarial logic integrity matrix. |
| 5 | Proof-of-Research & Tokenomics | Block validation, zk-SNARK proof generation, criteria-weight recording, soulbound piQ minting on Sepolia. |
| 6 | Dossiers, Cartography & Forecasting | CoARA/DORA dossiers, AI defence rebuttal strategy, global map of science, PiDN LSTM epoch forecasting. |
CoARA Compliance & Core Pillars
- Diverse Research Outputs (C5 & C8): structurally evaluates open datasets, code repositories, and containerized executable environments.
- Qualitative & Quantitative Balance (C1–C8): algorithms act as auditors, not replacements for peer review.
- Transparency & Researcher Sovereignty: evaluation weights, logic states, and criteria scores are hashed and stored on Ethereum (Sepolia). Researchers retain sovereign ownership via DID/ORCID.
Lab
Other projects, running inside ScholarPi. Each one keeps its own repository; the site builds and hosts it.
HAL-OS live in your browser
A bare-metal ternary network that boots into the framebuffer: 32,000 neurons, no OS, no compiler, no font. The real boot image runs on an x86 emulator in this page — feed it images, a camera, your screen or any file, and watch it learn to reproduce them. Nothing you inject leaves your browser.