Evidence
How established and verifiable the result is. Rewards work that has cleared peer review in a selective venue, is reported by independent citation indices and corroborated by other kinds of evidence, and can be read and checked in the open. A retraction collapses the whole score.
| Sub-metric | Source signal | How it maps to 0 to 100 | Weight |
|---|---|---|---|
| Peer review & venueWhether the work has cleared peer review, and how selective its venue is. Peer review is necessary but not sufficient: before citations accrue, a fresh result in one of the five general flagships (Nature, Science, Cell, NEJM, the Lancet) is a stronger evidence signal than one in a selective specialist or society journal, which in turn outranks one in a legitimate but very high-volume mega-journal, and any of them outranks a preprint that has not been reviewed at all. | OpenAlextype / primary_location / source | Peer-reviewed article graded by venue tier: flagship journal (Nature, Science, Cell, NEJM, the Lancet) 100; selective journal (every other journal the index admits: the specialist journals of those families, JAMA, PNAS, Physical Review Letters and X) 90; high-volume mega-journal (Nature Communications, Science Advances) 78; venue not recorded 88. Other records: review 90, book 70, dataset 60, preprint 45, an article that has not been peer reviewed 45, other 55. | 0.30 |
| CorroborationIndependent corroboration that the result is on the record and reaching beyond a single source, read as the BREADTH of independent records rather than the size of the citation pile. Counts how many of the five independent citation indices (OpenAlex, Crossref, Semantic Scholar, OpenCitations, Europe PMC) report a citation count for the work at all, zero included, plus orthogonal, non-citation lines of corroboration: citation in Wikipedia, registered clinical trials or clinical citations, released datasets or software or a repository copy, public funding on record, and technical-community discussion. Citation MAGNITUDE is deliberately judged under Impact, not here, so Evidence and Impact measure genuinely different things instead of both rewarding the same citation pile twice. | OpenAlex, Crossref, Semantic Scholar, OpenCitations, Europe PMC, DataCite, NIH RePORTER, Wikipedia, Hacker Newscount of indices that report a count for the work + orthogonal reach | Starts at 28 and rises by 9 for each of the five independent citation indices that returns a citation count for the work (zero counts; OpenCitations returns zero even for a DOI it does not hold) and by 6 for each orthogonal non-citation line (Wikipedia, clinical trials or clinical citations, open datasets, software or a repository copy, public funding, community discussion), up to the top of the 0-100 scale. | 0.30 |
| VerifiabilityHow openly the result can be read, reused, and reproduced. Rewards open access, a Creative Commons BY or CC0 license, a green repository copy anyone can archive, openly minable full text, and released datasets or software a reader can actually run. Cross-checked across OpenAlex, Unpaywall, Europe PMC, and DataCite. | OpenAlex, Unpaywall, Europe PMC, DataCiteis_oa / license / has_repository_copy / open datasets | Closed access 45; open access 88; any Creative Commons BY license (including the NC, ND and SA variants) or CC0 100. A repository copy raises the floor to 85, openly minable full text to 90, and released data or software to 92. | 0.20 |
| IntegrityRetraction and post-publication correction status. A retracted result is not evidence of anything and collapses the whole Frontier Score; a correction or expression of concern is a softer flag. A correction the publisher labels as its own error (a Publisher Correction) is not a flag on the work. Read across OpenAlex and Crossref (both update-to and updated-by notices). | OpenAlex, Crossrefis_retracted / update-to / updated-by | Clean 100 (also when the only notice is a Publisher Correction); a correction or expression of concern 75; retracted 0. Flagged if OpenAlex OR Crossref records the notice. | 0.20 |
Impact
How much the result matters, judged field-relative and by current momentum. Anchored on field-weighted citation impact (so a small field and a large field are judged fairly) and on how fast the work is being taken up right now. Raw lifetime citation volume is deliberately not rewarded: on the frontier, a large pile signals a result is already established.
| Sub-metric | Source signal | How it maps to 0 to 100 | Weight |
|---|---|---|---|
| Field-normalized impactImpact relative to the world average for the same field, year, and type, so a small field and a large field are judged fairly. Triangulates three independent field-normalized measures: OpenAlex's Field-Weighted Citation Impact, its citation percentile within the exact field-and-year cohort, and the NIH iCite Relative Citation Ratio. 1.0x is average; the percentile is the share of same-field, same-year work it out-cites. | OpenAlex, NIH iCitefwci / citation_normalized_percentile / relative_citation_ratio | Each ratio (1.0 = field average) maps 1.0 to 50, 3.0 to 75, 9.0 to 90; the field+year percentile maps directly (top 1% -> ~99). The available measures are averaged. When none is available yet (very recent work), this sub-metric is left out and Impact is computed from the others. | 0.45 |
| Influential citationsCitations that genuinely build on the work rather than mention it in passing (Semantic Scholar's influential-citation measure). A sharper signal of real impact than a raw count. | Semantic ScholarinfluentialCitationCount | Saturating: 15 influential citations maps to 50, and none maps to 0. When Semantic Scholar has no measure for the work, this sub-metric is left out and Impact is computed from the others. | 0.20 |
| Citation velocityCitations received in the trailing twelve months (estimated from yearly citation counts: this year's so far plus the share of last year's that falls inside the window), judged as a rate rather than a lifetime total. Momentum is what a frontier index cares about: a result being taken up fast right now is landing, whether its lifetime pile is large or still small. Where an expected field citation rate is known, the momentum is also read relative to it, so a fast-moving result in a quiet field is not overlooked. | OpenAlex, NIH iCitecounts_by_year / field_citation_rate | Saturating: 30 citations in the last twelve months maps to 50. When NIH iCite provides the field's expected rate, this is blended evenly with the ratio of observed to expected momentum. | 0.35 |
Novelty
How genuinely new the result is, and how fast-moving. Its lead signal is intrinsic: how novel the CONTRIBUTION is, measured from the work's own content at publication, so a recent-but-commodity result no longer reads as frontier just for being recent. Recency, rising attention and the edge of peer review add to it.
| Sub-metric | Source signal | How it maps to 0 to 100 | Weight |
|---|---|---|---|
| Conceptual noveltyHow genuinely new the contribution is, measured at publication from the work's OWN content, not its date: how atypical its combination of research areas is versus all prior work, and whether it builds on a broad, deep base rather than only the newest papers in a fast-moving area. This is what separates a real advance from a recent-but-commodity result, and it needs no forward citations, so it is stable for brand-new work. An honestly noisy proxy (even state-of-the-art novelty measures agree only moderately with expert judgment), so it carries bounded weight and never drives the ranking alone. | OpenAlextopics + referenced_works (publication-time) | Two publication-time signals, each 0-100. The primary one is atypical combination: how unusual the work's median pairing of research topics is versus the prior literature, by pointwise mutual information (a pairing made far more often than chance scores low). The second, foundational reach (the median age of the work's references and the share older than five years), can only lift it: when reach is higher, the score rises by 35% of the gap, and it is never lowered. When only one signal is measurable it is used alone; when neither is (a work too thin in topics or references), this sub-metric is left out and Novelty is computed from the other three. A review is damped to 60%. | 0.45 |
| RecencyHow recently the work appeared. The frontier is now, so newer scores higher. | OpenAlexpublication_date | Exponential decay with a 12-month half-life from the publication date. | 0.30 |
| Attention accelerationWhether attention is rising: citations in the latest full year versus the year before. Accelerating interest signals an active, opening frontier. | OpenAlexcounts_by_year | Latest/prior-year ratio: 1x maps to 50, 2x to 75, 4x to 100. The prior year counts as at least 3 citations, so a jump from one citation to a few is not read as explosive growth. With no citations in the latest full year it reads as unknown, and unknown is neutral (50). | 0.10 |
| Edge of reviewPreprints sit ahead of the peer-review process and brand-new papers have only just cleared it. Both earn novelty here (while a preprint is discounted under Evidence), because that is where the frontier forms first. | OpenAlextype / publication_date | Preprint 95; any other record 80 when under 6 months old, 65 under 18 months, 45 after that. | 0.15 |
Grounded in peer-reviewed method
The score does not invent its own idea of merit. Its field-normalized impact uses OpenAlex's Field-Weighted Citation Impact and the NIH iCite Relative Citation Ratio (Hutchins et al., PLoS Biology, 2016), both established measures that judge a work against its own field and year. Its conceptual-novelty signal adapts the atypical-combination idea of Uzzi, Mukherjee, Stringer and Jones (Science, 2013), which measures how unusually a work pairs prior research areas: where that study compared the journals a paper cites, Frontier reads how often the research topics OpenAlex assigns the work had been paired in earlier work. Frontier's contribution is not a new metric but a transparent, decomposable way of combining these, so every number stays traceable to a peer-reviewed basis and a primary record.
Frontier recency: why the edge is now
A result's standing decays as it ages into established practice, so the active edge stays on top. It is full when new; the part above a floor of 0.5 halves every 8 months, so a result keeps 75% of its standing at 8 months and about 63% at 16, and it never falls below 0.5 (a settled landmark is retained at half weight, never erased). This multiplier, applied on top of the weighted pillars, is what makes Frontier a frontier index rather than an all-time-greats list: a result loses standing as it ages, however heavily cited, so genuinely new work stays on top. A retraction overrides everything, collapsing the score to zero.
Tiers
The composite maps to a plain-language tier. An Early-signal result is capped at Notable and a Firming-up one at Major, whatever its score, so a tier never claims more certainty than its evidence carries.
| From | Tier | What it means |
|---|---|---|
| 60+ | Landmark | A defining advance at the current edge. |
| 50+ | Major | Strong, high-signal work with real momentum. |
| 40+ | Notable | Credible and gaining attention on the frontier. |
| 30+ | Emerging | New and promising, still proving itself. |
| 0+ | Early | Very early signal. Watch this space. |
Confidence, not just a number
A score says how a work rates; confidence says how settled that rating is. A landmark with thousands of citations across agreeing indices, cited on Wikipedia and taken into clinical trials, is “Settled”. A two-week-old preprint is an “Early signal” whose impact is still an estimate. Confidence is a separate 0 to 100 reading: 36% citation maturity (the median citation count across the indices, where 40 citations reads 50), 20% time since publication (full at 12 months), 16% how closely the citation indices agree, 18% how many independent kinds of evidence corroborate the work (two or more citation indices, four or more, field-normalized impact, citation in Wikipedia, clinical trials or clinical citations, released code, data or a repository copy, public funding, community discussion; 6 of these read full) and 10% whether a field-normalized measure exists yet. 68 or more is Settled, 44 or more Firming up, and anything lower an Early signal. It is kept separate from the score, so novelty is never mistaken for certainty: the score itself is not lowered for being new, but its tier label is capped until the evidence firms up (an Early signal shows at most Notable, Firming up at most Major). You will see it on every breakthrough, next to its score.
What we designed around
Known ways a naive score goes wrong, and how the Frontier Score avoids each.
- 01
A single citation index can be wrong, incomplete, or gamed.
Citations are cross-checked across up to five independent indices (OpenAlex, Crossref, Semantic Scholar, OpenCitations, and Europe PMC): the Evidence pillar counts how many of them report a count for the work, and confidence uses the median of their counts and how closely they agree, so a gap or an outlier in one index shows up instead of passing silently. Impact reads OpenAlex's field-weighted impact and yearly counts, NIH iCite where it applies, and Semantic Scholar's influential-citation count, the citations that genuinely build on the work rather than mention it in passing. - 02
Citations only measure what OTHER PAPERS did, not whether a result reached the world.
Frontier pulls thirteen sources, not just citation indices. A result's reach is cross-checked against entry into Wikipedia, technical-community discussion, registered clinical trials, released datasets and code, and the public grants that funded it. Each of these lines adds a few points to the Evidence pillar's Corroboration sub-metric (released data or code also lift Verifiability) and raises a work's CONFIDENCE, because each is an independent line of evidence that the result is real and landing. - 03
Raw citation counts favor old, re-catalogued classics, and large fields over small ones.
Impact is anchored on Field-Weighted Citation Impact (FWCI), which compares a work only to others in the same field, year, and type. The field-and-year percentile and, for biomedicine, iCite's relative citation ratio join it, and momentum is read against the field's expected rate wherever iCite provides one; elsewhere momentum is an absolute count. Novelty separately rewards recency, so an established result cannot masquerade as a current frontier. - 04
Very recent work has almost no citations yet.
Impact leans on recent citation velocity and field-normalized measures rather than lifetime totals, a field-normalized measure or influential-citation count not yet available is left out (its weight redistributed) rather than scored as zero, and Novelty rewards recency and rising attention. New work is not punished for being new. - 05
A retracted result can still be highly cited.
Retraction applies a hard integrity multiplier that zeroes the whole score. A retracted result is not a frontier, whatever its citation count. - 06
Preprints are unreviewed but often the leading edge.
Preprints lose points under Evidence (not yet reviewed) and gain them under Novelty (ahead of review), rather than being treated as if peer-reviewed.
What the Frontier Score cannot tell you
A single number is a lens, not a verdict. These are the limits of what it can measure, stated plainly.
- 01
Absence is not a verdict.
The index can only rank what it has ingested. A result missing from Frontier has not been judged and found wanting; it may simply not have been swept in yet. Read a low or missing entry as incomplete coverage, never as a conclusion. - 02
The newest work is an estimate.
Very recent work has little citation or field-normalized evidence yet, so its impact reading is provisional and will move as uptake accrues. That is exactly why every entry also carries a Confidence, shown separately from the score, and why an Early-signal result is capped below the Major and Landmark tiers and a Firming-up one below Landmark. - 03
Citation signals carry field and language bias.
We normalize against the field-and-year cohort wherever a field-relative measure exists (FWCI, the citation percentile, iCite's RCR), but not every signal has one. A citation-sparse field, a very new subfield, or non-English work can still be undercounted. - 04
Fields are assigned automatically.
Each breakthrough is classified into one field by an automated keyword heuristic over its title and its OpenAlex topic, and misfilings found in review are corrected by hand. A genuinely cross-disciplinary result can land in the wrong box, or in Other. - 05
The novelty-led weighting is a choice.
Novelty carries half the composite on purpose: Frontier ranks the active edge, not the most-cited work of the decade. A different weighting would rank differently. That is an editorial judgment about what a frontier is, stated openly, not an objective fact.
How Frontier compares
Why a scored, decomposable, novelty-led frontier index is not the same thing as a citation count, an attention score, or a year-in-review.
- How is this different from Google Scholar or a raw citation count?
- Raw citation counts favor settled classics and large fields, and they say nothing about how new a result is. The Frontier Score anchors impact on measures normalized against the same field and year (field-weighted citation impact, the field-and-year citation percentile and the iCite relative citation ratio), weights novelty at half the composite, and then discounts a result's standing as it ages. So a genuinely new 2026 result can outrank a decade-old, heavily-cited paper, which is the opposite of what a citation count does.
- How is it different from Altmetric or an attention score?
- Attention scores measure buzz: news mentions, posts, and shares. That tracks what is being talked about, not what is significant or sound. Frontier scores evidence, impact, and novelty, and counts attention (Wikipedia, community discussion) only as a corroboration line: a few points inside the Evidence pillar and a boost to confidence, never the score itself.
- How is it different from a 'state of the field' annual report?
- Those are periodic, usually single-field, and written as narrative. Frontier is live, spans the natural sciences, medicine, computer science, engineering and mathematics in one ranking, and is fully decomposable: every entry carries a score you can trace, number by number, to the primary record it came from.
- Why trust the number?
- Because it is not a black box. The entire scoring recipe is public, each breakthrough page lists the source figures behind its score, each linked to its record, and the method builds on peer-reviewed science-of-science metrics, combined with weights and curves that are Frontier's own and published in full. You can check the inputs, the normalization, and the weights yourself.
Changes to the method
Scores move every week as evidence accrues. From version 1.3.0 on, when one moves because the method itself changed, the change is recorded here with what it did to the index.
- v1.3.1
A correction that the publisher labels as its own error (a Publisher Correction, used for errors the publisher introduced, such as a wrong author order) no longer counts as an integrity flag. Crossref records every correction notice the same way, so the index now reads each notice's own title. Author corrections, errata, corrigenda and expressions of concern are still flagged.
Effect on the index: Two of the seven entries with a correction notice had only a Publisher Correction: the Helios 98-qubit trapped-ion computer (author list) and the hippocampal semantics study (author order). Each rose by one point, to 60 and 72, and neither changed tier.
- v1.3.0
PNAS, Physical Review Letters and Physical Review X are now graded as selective journals (90 on the venue sub-metric), alongside the specialist journals of the Nature, Cell and Science families, rather than as flagships (100) alongside Nature, Science, Cell, NEJM and the Lancet. PNAS and Physical Review Letters each publish several times as many papers a year as Cell or NEJM, and Physical Review X is a specialist physics journal, now graded like Nature Physics.
Effect on the index: Of the 37 entries from those three journals, 26 dropped by one point and 11 kept their score; none changed tier. New papers from them no longer outrank equally fresh papers from the five flagships on venue alone.