Vela

frontiers / frontier

AI-for-science benchmark state

constellation seal · derived from vfr_efc649fd772a1ff1
id
vfr_efc649fd772a1ff1
license
CC-BY-4.0
findings
12
accepted core
12
contested
0
links
0
sources
1
evidence
12
avg conf
0.30

used by 0 · replayed by 1 producer · second seat open

e24/24 · finding.noted · reviewer:will-blair · 2026-06-10 · 6c12→d02f

Evidence atom

back to sources

LEAKAGE HAZARD (ProteinGym). PLMs trained on UniProt may have seen sequences related to the assay proteins; 'zero-shot' is zero-shot on the LABELS, not necessarily on the SEQUENCES. VERIFICATION STATE: training-sequence overlap with assay proteins is under-audited. Open obligation: a sequence-similarity leakage audit between each model's training set and the assay proteins before banking any SOTA claim.

id
vea_874a2c26d4e672f2
frontier
AI-for-science benchmark state
source
vs_066123dd29a9c5b4
finding
vf_8212daf3d7034a93

evidence boundary

unknown

computational

finding binding

bound

computational

LEAKAGE HAZARD (ProteinGym). PLMs trained on UniProt may have seen sequences related to the assay proteins; 'zero-shot' is zero-shot on the LABELS, not necessarily on the SEQUENCES. VERIFICATION STATE: training-sequence overlap with assay proteins is under-audited. Open obligation: a sequence-similarity leakage audit between each model's training set and the assay proteins before banking any SOTA claim.

source binding

source-bound

manual finding

vs_066123dd29a9c5b4

review context

unverified

2 events

2 reviewable changes and 0 evaluation records target this atom or its bound objects.

statement

LEAKAGE HAZARD (ProteinGym). PLMs trained on UniProt may have seen sequences related to the assay proteins; 'zero-shot' is zero-shot on the LABELS, not necessarily on the SEQUENCES. VERIFICATION STATE: training-sequence overlap with assay proteins is under-audited. Open obligation: a sequence-similarity leakage audit between each model's training set and the assay proteins before banking any SOTA claim.

extraction method

manual_curation

support relation

unknown

condition refs

vcnd_f640e1e86c8fd161

caveats

  • missing evidence locator

Review, event, and evaluation records

4

events

  • vev_07e40e45981061adfinding.asserted

    Manual finding added to frontier state

    reviewer:will-blairreviewer:will-blair · 2026-06-10

  • vev_c4da9db8be63634efinding.noted

    HARDENING (benchmark-state): label_provenance=attested (records-not-reruns; ground truth is an answer key, not a frozen-verifier rederivation), valid_as_of=2026-06-10, model_cutoff=unknown. Under the trust ladder, attested label provenance caps this record below 'verified' until a deterministic rederivation exists.

    reviewer:will-blairreviewer:will-blair · 2026-06-10

reviewable changes

  • vpr_c555bef607043399finding.add

    Manual finding added to frontier state

    reviewer:will-blairapplied · reviewer:will-blair · 2026-06-10

  • vpr_fd307d5d15c2cef7finding.note

    HARDENING (benchmark-state): label_provenance=attested (records-not-reruns; ground truth is an answer key, not a frozen-verifier rederivation), valid_as_of=2026-06-10, model_cutoff=unknown. Under the trust ladder, attested label provenance caps this record below 'verified' until a deterministic rederivation exists.

    agent — machine actor, no signing keyapplied · agent:hardening-2026-06-10 · 2026-06-10

evaluations

No evaluation rows are attached.

statement.registered · agent:claude-proxy · 4 days

renders the record as of vev_e73c9b6c · 1,355 events · hub

Search Vela

Jump to a section, signal, campaign, document, primitive, work path, frontier, record index, atlas, constellation, agent, capability, or full-state search.