The graph, off-chain
DKG nodes store each asset's statements and replicate them across the nodes that follow its graph. People and AI agents read them by the asset's permanent address, its UAL, or query them with SPARQL.
About LabsTrail
LabsTrail is an open registry that rates the quality of life-science papers, recording each paper and each of its ratings off-chain on the OriginTrail DKG and on-chain on Base.
How the score worksWhy LabsTrail exists
Many published results can't be repeated. In a 2016 Nature survey of more than 1,500 researchers, more than 70% had tried and failed to reproduce another scientist's experiments (Baker, Nature 533, 452-454).
Yet how far to trust a paper is still judged mostly by the journal it appeared in, and the reviews behind that judgement are rarely public. The link to a paper's file can also change or break after others have cited it.
AI agents now read, cite and build on research at a scale no person can. They have no better way to tell a paper that gives enough to be repeated from one that doesn't.
LabsTrail gives each paper a record whose provenance anyone can check, then rates its quality in the open: a machine reproducibility score first, then transparent human review. The paper and each of its ratings are recorded off-chain and on-chain, where people and agents can read and check them.
Built on the OriginTrail DKG
OriginTrail's Decentralized Knowledge Graph (DKG) is an open network of nodes that store knowledge as Knowledge Assets: sets of linked statements, each with a permanent address, an owner and a proof on a blockchain. An asset lives in two layers that check each other.
The proof, on-chain
Publishing mints the asset on Base as an ERC-721 token and anchors the merkle root of its statements, one hash computed from all of them. Anyone can check that the statements still match it, and only the token's holder can add a new version.
The paper itself, on IPFS
IPFS is a decentralized storage network that addresses each file by its CID, a hash of its content. The paper's PDF is stored there, off-chain, and its asset links that CID. A changed file would get another CID, so the link always leads to the file the submitter signed, and no one can swap it unnoticed.
Papers and ratings are Knowledge Assets
A paper
Its asset gives the title, the authors and the PDF's CID, with the submitter's address and their signature of that CID. Anyone can check who submitted which file without trusting LabsTrail.
A rating
Each rating is an asset of its own that points to the paper's, with every check, answer and quote behind its score. LabsTrail's RatingController contract on Base records the score. A new rating or review adds its own score, and the paper's asset and earlier scores stay as they were.
LabsTrail's node mints both kinds of asset, holds their tokens and pays their fees, so only LabsTrail can add a version to them. The rating grows in three phases, from a machine score for reproducibility readiness to wet-lab replication.
What comes next
- next
Transparent human review
A machine can check what a paper reports, not whether its content holds up. That takes people, and in traditional peer review, who reviews a paper is decided behind closed doors.
LabsTrail will draw reviewers at random with Chainlink VRF, which delivers each random number with a proof checked on-chain, so no one, LabsTrail included, can steer who is picked, and anyone can check the draw. The reviewers' scores join the paper's rating record, off-chain and on-chain.
The plan, not decided yet: reviewers stake to join the pool, are paid for their reviews, and lose part of their stake if they review in bad faith.
- planned
Payment, for people and AI agents
Today a requester pays only their gas, LabsTrail pays the rest, and caps hold spam back. The plan, not decided yet: a fee on publishing and on rating that a person or an AI agent can pay, covering the costs and the reviewers' pay.
- planned
An API for AI agents
Agents will read papers and ratings, and ask for a rating, without going through this website.
- planned
A lasting, decentralized copy of every paper
One pinning service holds each PDF on IPFS today. Its hash proves the file is unchanged, but only more copies keep it available, so the plan is to store each paper on a decentralized storage network too, before mainnet.
- later
Wet-lab replication
A lab repeats the experiments, and reviewers check its results before they are scored.
On testnets for now
Everything runs on Base Sepolia, a testnet, until a payment plan is in place; then LabsTrail moves to mainnet. Until then a requester pays only their gas, and LabsTrail pays to publish and to rate.
Visits are counted with Vercel Web Analytics, which sets no cookies.