rewirebio.iobenchmarks
Result

Scib bio conservation score (Mean of seven scIB biological-conservation metrics (isolated-label score, NMI and ARI of k-means and of Leiden clusters against the cell types, cell-type silhouette and cLISI), per atlas or averaged over the seven atlases) of scGPT (HyCell report) on scIB human immune atlas, integer-count part (HyCell report)

0.629 Scib bio conservation score (Mean of seven scIB biological-conservation metrics (isolated-label score, NMI and ARI of k-means and of Leiden clusters against the cell types, cell-type silhouette and cLISI), per atlas or averaged over the seven atlases)

Methods

Tested configuration
scGPT (HyCell report)
Protocol
scIB biological conservation (HyCell report)
Dataset
scIB human immune atlas, integer-count part (HyCell report)
Procedure
scIB biological conservation (HyCell report)
Evaluation
scGPT on bio-conservation, immune (HyCell report)
Coverage
Not reported scored / Not reported eligible
Uncertainty
Not reported by the source: No uncertainty is printed anywhere in the report.
Evidence
Author-reported evaluation · source checkedHyCell: a Compact Embedding Model for Single-Cell Transcriptomes (Hyra Team, 2026-11-01) · Figure 1b, row 'Immune', column 'scGPT', printed value (page 2 of 9)

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
Whole atlas; batches held out in turn for annotation transfer
Adaptation
Frozen zero-shot embedding of each cell on its own, released weights and default inference settings, no batch, donor or cell-type labels at inference
Scoring implementation
One implementation built on scib-metrics, on the absolute scale without rescaling across methods.

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
0.629
Individual claims
HyCell: a Compact Embedding Model for Single-Cell Transcriptomes (Hyra Team, 2026-11-01)

Original source ↗

Figure 1b, row 'Immune', column 'scGPT', printed value (page 2 of 9)

Version: Dated 2026-11-01 on page 1; retrieved at commit 0dcc533cc9f8f76c90ecbfe944df816b6672328a (9 pages)
Retrieved: 2026-11-06T11:42:11Z

source checked

["independent-cell-check","deterministic-table-parse","pdf-text-parse","source-hash-verification"] · 2026-11-06T00:00:00Z

author reported

Audit details

Field: attributes.printed_value

Source artifact SHA-256: dc2cacf87ffae8525e81fede1b01f261d3a49cece9061465c36ad1fac6680a28

Hash scope: pdftotext -layout text layer, parsed by extract/parse.py with every row label and cell count asserted.

Inspected artifact

Extraction artifact SHA-256: dc2cacf87ffae8525e81fede1b01f261d3a49cece9061465c36ad1fac6680a28

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: hycell-20261106-result-bio-conservation-scgpt-immune

metric
scib-bio-conservation-score
metric direction
higher
metric qualifier
Mean of seven scIB biological-conservation metrics (isolated-label score, NMI and ARI of k-means and of Leiden clusters against the cell types, cell-type silhouette and cLISI), per atlas or averaged over the seven atlases
missing metadata
uncertainty: note: No uncertainty is printed anywhere in the report.; reason: unreported
numeric value
0.629
printed value
0.629
review
artifact sha256: dc2cacf87ffae8525e81fede1b01f261d3a49cece9061465c36ad1fac6680a28; method: independent-cell-check; deterministic-table-parse; pdf-text-parse; source-hash-verification; method note: Independent review pass by a different agent from the extractor. All nine pinned artifacts were re-downloaded at commit 0dcc533c and their SHA-256 checked against the source records, including the 64,439,816-byte weights, whose hash also equals the weights_sha256 the sidecar publishes; pdftotext -layout on the re-downloaded PDF reproduced the archived text layer byte for byte. npm run batch -- check-values was rerun over the whole batch against that text layer and reproduced the extractor's value-check.tsv exactly (60 in_cell, 96 in_row, 67 locator_not_parsed, no not_in_row, wrong_column or overclaimed_in_row). All 198 cells of Tables 3, 4 and 5 and of Figure 1b were then re-read by column position with a reviewer script written for this pass, which finds each table by its own printed header line and indexes the cell by column, so Table 4's Immune column cannot be read from Table 3; every printed value matched. Figure 1b's column identity was fixed two independent ways: its Mean row matches Table 3's Bio column for exactly one assignment of six models, and pdftotext -raw prints the rotated headers as plain words in that same order. The 36 arithmetic identities (S against its printed weights for all 12 rows, Table 3's S against the mean of Table 4 for all 12, and Figure 1b's column means and Mean row against Table 3's Bio for all six) were recomputed and all hold. The 25 prose values were checked by hand against the sentence each locator names, including the system each number belongs to. Each claim was re-read against the passage its locator names.; retrieval url: https://raw.githubusercontent.com/Tencent-Hunyuan/Hyra-results/0dcc533cc9f8f76c90ecbfe944df816b6672328a/AI4Science/hycell/paper/HyCell_Technical_Report.pdf; reviewed at: 2026-11-06T00:00:00Z; reviewer: claude; reviewer note: Claude research agent; no human review claimed. Reviewed by a different agent from the one that extracted the batch. Transcription checked against the pinned source; nothing was reproduced and no value was recomputed.
source locator
Figure 1b, row 'Immune', column 'scGPT', printed value (page 2 of 9)
unit
unitless
Related records

Suggest a correction