rewirebio.iobenchmarks
Result

F1 (Intervention targets: straight-through top-k gate used in the decoder rollout (k = 3, Table 6); PerturbPFN in-prior synthetic episodes) of PerturbPFN step-3100 checkpoint, in-context forward pass on PerturbPFN synthetic prior, 1,024 held-out episodes

0.629 F1 (Intervention targets: straight-through top-k gate used in the decoder rollout (k = 3, Table 6); PerturbPFN in-prior synthetic episodes)

Methods

Tested configuration
PerturbPFN step-3100 checkpoint, in-context forward pass
Protocol
PerturbPFN in-prior latent recovery on held-out synthetic episodes (Table 1)
Dataset
PerturbPFN synthetic prior, 1,024 held-out episodes
Procedure
PerturbPFN in-prior latent recovery on held-out synthetic episodes (Table 1)
Evaluation
PerturbPFN on in-prior latent recovery (PerturbPFN Table 1)
Coverage
Not reported scored / Not reported eligible
Uncertainty
Not reported by the source: No spread is printed for this value.
Evidence
Author-reported evaluation · source checkedPerturbPFN: Probing the Limits of Synthetic Priors in Drug Perturbation Modelling (arXiv:2607.23447v1) · Section 4.1 'Results', sentence 'The straight-through top-k gate used by the decoder obtains a lower F1 of 0.6287'

A source-checked result verifies the numerical transcription, not every model or protocol detail. Evaluation metadata: source checked. Source checked does not mean independently reproduced.

Reproduction

Split
Held-out synthetic episodes
Adaptation
None: one forward pass, no parameter updates
Scoring implementation
Appendix C definitions

No execution recipe has been verified for this exact configuration and evaluation. A benchmark's general instructions may use different inputs, splits or model settings.

Reproducing this published result requires matching its model configuration, data, split and scorer. Source checking or a successful smoke test does not establish score reproduction.

Evidence

Source checking verifies the cited claim or transcription. It does not establish independent reproduction.

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

1 evidence row matching the loaded filters

Claims, original sources and review scope · Release 2026-11-21-c8bea25e5abb
Property and statementOriginal source and locationReview and provenance
Reported result
0.6287
Individual claims
PerturbPFN: Probing the Limits of Synthetic Priors in Drug Perturbation Modelling (arXiv:2607.23447v1)

Original source ↗

Section 4.1 'Results', sentence 'The straight-through top-k gate used by the decoder obtains a lower F1 of 0.6287'

Version: arXiv v1, submitted 2026-07-26
Retrieved: 2026-11-15T20:27:40Z

source checked

["independent-cell-check","deterministic-table-parse","pdf-text-parse","source-hash-verification"] · 2026-11-15T00:00:00Z

author reported

Audit details

Field: attributes.printed_value

Source artifact SHA-256: 42191efa04b45ca04b2be38b51ea374cd9bf891efe2399d331fec9eaeab200b1

Hash scope: Exact bytes of the PDF. Three downloads on 2026-11-15 (20:22Z to 20:27Z) gave identical bytes.

Inspected artifact

Extraction artifact SHA-256: 42191efa04b45ca04b2be38b51ea374cd9bf891efe2399d331fec9eaeab200b1

Extraction artifact

Sources and history

Release 2026-11-21-c8bea25e5abb · Record review: source checked

1 source record and release historyDownload this release (gzip)
Technical metadata and extraction receipts

Stable ID: perturbpfn-20261115-result-in-prior-target-topk-gate-f1

areas
biological-networks
contexts
research
metric
f1-score
metric direction
higher
metric qualifier
Intervention targets: straight-through top-k gate used in the decoder rollout (k = 3, Table 6); PerturbPFN in-prior synthetic episodes
missing metadata
uncertainty: note: No spread is printed for this value.; reason: unreported
numeric value
0.6287
printed value
0.6287
review
artifact sha256: 42191efa04b45ca04b2be38b51ea374cd9bf891efe2399d331fec9eaeab200b1; method: independent-cell-check; deterministic-table-parse; pdf-text-parse; source-hash-verification; method note: Independent review pass. Both pinned artifacts were re-downloaded and their SHA-256 checked against the source records, the mutable abstract page included, and both matched. A pdftotext -layout run (poppler 26.08.0) on the refetched PDF is byte for byte the contents of the archived gzipped text layer, so the archive is provably derived from the pinned bytes. npm run batch -- check-values reproduced the extractor's figures exactly (64 in_cell, 11 in_row, 6 locator_not_parsed, no wrong_column and no overclaimed_in_row) and wrote the same value-check.tsv. All 81 results were then re-read by review/recheck_tables.py, written for this pass and using neither extract/parse_tables.py nor extract/check_tables.py: it takes each table's column spans from the character offsets of that table's own printed header line, assigns every numeric token on a body line to the column whose header centre is nearest, and matches cells to records through each result's source_locator text rather than its ID, checking printed_value, numeric_value, the printed spread and the system the result's evaluation names. Two negative controls confirmed it fails on a swapped pair of values and on a cell shifted into the neighbouring column. A manual check of 16 cells covered column assignment in all three tables, including Table 2's unlabelled PerturbPFN score line, Table 1's third label and value pair on each of its three rows, and Table 3's middle columns. The printed overall score was re-derived as the equal-weight mean of each row's three normalised columns. The 54 descriptive records were read against Sections 2 to 5, Appendices A to E and Tables 4 to 8 of the pinned PDF.; retrieval url: https://arxiv.org/pdf/2607.23447v1; reviewed at: 2026-11-15T00:00:00Z; reviewer: claude; reviewer note: Claude review agent (independent of the extractor); no human review claimed. Transcription was checked against the pinned source; nothing was reproduced and no value was recomputed.
source locator
Section 4.1 'Results', sentence 'The straight-through top-k gate used by the decoder obtains a lower F1 of 0.6287'
unit
unitless
Related records

Suggest a correction