FLIP2 IRED two-to-many; held-out test set · IRED
Source-specific evaluation. No equivalence to other releases, protocols or model families is inferred.
Published comparisons
Each figure keeps its source, protocol, dataset and metric together. Results from different protocols are shown separately, without a pooled ranking.
FLIP2 IRED two-to-many; held-out test set · IRED: spearman
spearman (correlation) · Higher values are better for this metric.
Complete selected source table is retained across source-order panels. These point estimates do not establish statistical significance or a universal ranking.
- Ridge (one-hot) · Configuration · Author-reported evaluation0.193
- Ridge (one-hot + likelihoods) · Configuration · Author-reported evaluation0.211
- Dayhoff likelihood · Configuration · Author-reported evaluation0.142
- ESM2-650M likelihood · Configuration · Author-reported evaluation0.141
- CARP-640M likelihood · Configuration · Author-reported evaluation0.144
- CARP-640M supervised · Configuration · Author-reported evaluation0.072 ± 0.046
- CARP-640M naive supervised · Configuration · Author-reported evaluation0.012 ± 0.017
- ESMC-300M supervised · Configuration · Author-reported evaluation0.160 ± 0.022
- ESMC-300M naive supervised · Configuration · Author-reported evaluation0.106 ± 0.005
Source order is preserved. Plotted marks show point estimates; uncertainty, where reported, is retained in the printed values and table. Differences do not establish statistical significance.
FLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row Ridge (one-hot); column spearman through Table A5; row ESMC-300M naive supervised; column spearmanValues, uncertainty and evidence
| Tested entity | Printed value | Uncertainty | Evidence |
|---|---|---|---|
| Ridge (one-hot) · Configuration | 0.193 correlation | Not reported | Author-reported evaluation · source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row Ridge (one-hot); column spearman |
| Ridge (one-hot + likelihoods) · Configuration | 0.211 correlation | Not reported | Author-reported evaluation · source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row Ridge (one-hot + likelihoods); column spearman |
| Dayhoff likelihood · Configuration | 0.142 correlation | Not reported | Author-reported evaluation · source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row Dayhoff likelihood; column spearman |
| ESM2-650M likelihood · Configuration | 0.141 correlation | Not reported | Author-reported evaluation · source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row ESM2-650M likelihood; column spearman |
| CARP-640M likelihood · Configuration | 0.144 correlation | Not reported | Author-reported evaluation · source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row CARP-640M likelihood; column spearman |
| CARP-640M supervised · Configuration | 0.072 ± 0.046 correlation | printed spread: 0.046; type: unreported; value: 0.046 | Author-reported evaluation · source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row CARP-640M supervised; column spearman |
| CARP-640M naive supervised · Configuration | 0.012 ± 0.017 correlation | printed spread: 0.017; type: unreported; value: 0.017 | Author-reported evaluation · source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row CARP-640M naive supervised; column spearman |
| ESMC-300M supervised · Configuration | 0.160 ± 0.022 correlation | printed spread: 0.022; type: unreported; value: 0.022 | Author-reported evaluation · source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row ESMC-300M supervised; column spearman |
| ESMC-300M naive supervised · Configuration | 0.106 ± 0.005 correlation | printed spread: 0.005; type: unreported; value: 0.005 | Author-reported evaluation · source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row ESMC-300M naive supervised; column spearman |
Scope and limitations
- Source-specific evaluation. No equivalence to other releases, protocols or model families is inferred.
- Exact source-defined evaluation scope; reported scores are not rewire reproductions.
- Missing source cells and quarantined conflicts are recorded in acquisition and audit tables. Per-result scoring denominators may be unreported.
Source transcription and grouping reviewed by automated source review on 2026-09-19. These experiments were not independently reproduced by rewire.
Tested entities and results
Release 2026-09-19-eb6149e5d766 · 9 evaluations · 18 metric rows. Different protocols are not a single leaderboard. Charts retain each source table’s protocol and dataset.
| Metric and finding | Coverage and uncertainty | Evidence |
|---|---|---|
| CARP-640M likelihood · FLIP2 IRED two-to-many; held-out test set · IRED Configuration: CARP-640M likelihoodProtocol: FLIP2 IRED two-to-many; held-out test set · IREDDataset subset: IRED FLIP2 IRED two-to-many; held-out test set Author-reported evaluation · Evaluation metadata: discovered | ||
| 0.144 spearman Unit: correlation · Direction: higher Aggregation: reported point estimate | Uncertainty: Not reported Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row CARP-640M likelihood; column spearman Source checking is not independent reproduction. |
| 0.956 ndcg Unit: unitless · Direction: higher Aggregation: reported point estimate | Uncertainty: Not reported Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row CARP-640M likelihood; column ndcg Source checking is not independent reproduction. |
| ESMC-300M naive supervised · FLIP2 IRED two-to-many; held-out test set · IRED Configuration: ESMC-300M naive supervisedProtocol: FLIP2 IRED two-to-many; held-out test set · IREDDataset subset: IRED FLIP2 IRED two-to-many; held-out test set Author-reported evaluation · Evaluation metadata: discovered | ||
| 0.106 ± 0.005 spearman Unit: correlation · Direction: higher Aggregation: mean of five random-seed fits (Section 4.3) | Uncertainty: printed spread: 0.005; type: unreported; value: 0.005 Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row ESMC-300M naive supervised; column spearman Source checking is not independent reproduction. |
| 0.949 ± 0.001 ndcg Unit: unitless · Direction: higher Aggregation: mean of five random-seed fits (Section 4.3) | Uncertainty: printed spread: 0.001; type: unreported; value: 0.001 Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row ESMC-300M naive supervised; column ndcg Source checking is not independent reproduction. |
| Dayhoff likelihood · FLIP2 IRED two-to-many; held-out test set · IRED Configuration: Dayhoff likelihoodProtocol: FLIP2 IRED two-to-many; held-out test set · IREDDataset subset: IRED FLIP2 IRED two-to-many; held-out test set Author-reported evaluation · Evaluation metadata: discovered | ||
| 0.955 ndcg Unit: unitless · Direction: higher Aggregation: reported point estimate | Uncertainty: Not reported Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row Dayhoff likelihood; column ndcg Source checking is not independent reproduction. |
| 0.142 spearman Unit: correlation · Direction: higher Aggregation: reported point estimate | Uncertainty: Not reported Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row Dayhoff likelihood; column spearman Source checking is not independent reproduction. |
| Ridge (one-hot + likelihoods) · FLIP2 IRED two-to-many; held-out test set · IRED Configuration: Ridge (one-hot + likelihoods)Protocol: FLIP2 IRED two-to-many; held-out test set · IREDDataset subset: IRED FLIP2 IRED two-to-many; held-out test set Author-reported evaluation · Evaluation metadata: discovered | ||
| 0.211 spearman Unit: correlation · Direction: higher Aggregation: reported point estimate | Uncertainty: Not reported Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row Ridge (one-hot + likelihoods); column spearman Source checking is not independent reproduction. |
| 0.964 ndcg Unit: unitless · Direction: higher Aggregation: reported point estimate | Uncertainty: Not reported Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row Ridge (one-hot + likelihoods); column ndcg Source checking is not independent reproduction. |
| ESM2-650M likelihood · FLIP2 IRED two-to-many; held-out test set · IRED Configuration: ESM2-650M likelihoodProtocol: FLIP2 IRED two-to-many; held-out test set · IREDDataset subset: IRED FLIP2 IRED two-to-many; held-out test set Author-reported evaluation · Evaluation metadata: discovered | ||
| 0.141 spearman Unit: correlation · Direction: higher Aggregation: reported point estimate | Uncertainty: Not reported Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row ESM2-650M likelihood; column spearman Source checking is not independent reproduction. |
| 0.953 ndcg Unit: unitless · Direction: higher Aggregation: reported point estimate | Uncertainty: Not reported Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row ESM2-650M likelihood; column ndcg Source checking is not independent reproduction. |
| CARP-640M naive supervised · FLIP2 IRED two-to-many; held-out test set · IRED Configuration: CARP-640M naive supervisedProtocol: FLIP2 IRED two-to-many; held-out test set · IREDDataset subset: IRED FLIP2 IRED two-to-many; held-out test set Author-reported evaluation · Evaluation metadata: discovered | ||
| 0.012 ± 0.017 spearman Unit: correlation · Direction: higher Aggregation: mean of five random-seed fits (Section 4.3) | Uncertainty: printed spread: 0.017; type: unreported; value: 0.017 Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row CARP-640M naive supervised; column spearman Source checking is not independent reproduction. |
| 0.944 ± 0.001 ndcg Unit: unitless · Direction: higher Aggregation: mean of five random-seed fits (Section 4.3) | Uncertainty: printed spread: 0.001; type: unreported; value: 0.001 Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row CARP-640M naive supervised; column ndcg Source checking is not independent reproduction. |
| CARP-640M supervised · FLIP2 IRED two-to-many; held-out test set · IRED Configuration: CARP-640M supervisedProtocol: FLIP2 IRED two-to-many; held-out test set · IREDDataset subset: IRED FLIP2 IRED two-to-many; held-out test set Author-reported evaluation · Evaluation metadata: discovered | ||
| 0.948 ± 0.004 ndcg Unit: unitless · Direction: higher Aggregation: mean of five random-seed fits (Section 4.3) | Uncertainty: printed spread: 0.004; type: unreported; value: 0.004 Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row CARP-640M supervised; column ndcg Source checking is not independent reproduction. |
| 0.072 ± 0.046 spearman Unit: correlation · Direction: higher Aggregation: mean of five random-seed fits (Section 4.3) | Uncertainty: printed spread: 0.046; type: unreported; value: 0.046 Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row CARP-640M supervised; column spearman Source checking is not independent reproduction. |
| Ridge (one-hot) · FLIP2 IRED two-to-many; held-out test set · IRED Configuration: Ridge (one-hot)Protocol: FLIP2 IRED two-to-many; held-out test set · IREDDataset subset: IRED FLIP2 IRED two-to-many; held-out test set Author-reported evaluation · Evaluation metadata: discovered | ||
| 0.193 spearman Unit: correlation · Direction: higher Aggregation: reported point estimate | Uncertainty: Not reported Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row Ridge (one-hot); column spearman Source checking is not independent reproduction. |
| 0.960 ndcg Unit: unitless · Direction: higher Aggregation: reported point estimate | Uncertainty: Not reported Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row Ridge (one-hot); column ndcg Source checking is not independent reproduction. |
| ESMC-300M supervised · FLIP2 IRED two-to-many; held-out test set · IRED Configuration: ESMC-300M supervisedProtocol: FLIP2 IRED two-to-many; held-out test set · IREDDataset subset: IRED FLIP2 IRED two-to-many; held-out test set Author-reported evaluation · Evaluation metadata: discovered | ||
| 0.954 ± 0.002 ndcg Unit: unitless · Direction: higher Aggregation: mean of five random-seed fits (Section 4.3) | Uncertainty: printed spread: 0.002; type: unreported; value: 0.002 Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row ESMC-300M supervised; column ndcg Source checking is not independent reproduction. |
| 0.160 ± 0.022 spearman Unit: correlation · Direction: higher Aggregation: mean of five random-seed fits (Section 4.3) | Uncertainty: printed spread: 0.022; type: unreported; value: 0.022 Scored: Not reported · Eligible: Not reported | source checkedFLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Table A5; row ESMC-300M supervised; column spearman Source checking is not independent reproduction. |
Overview
Source-specific evaluation. No equivalence to other releases, protocols or model families is inferred.
Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.
Evaluation design
Benchmarks bring together tasks and protocols. A task describes the biological question; a protocol defines a particular test.
Benchmarks
These source-backed links do not make different protocols or scores interchangeable.
Recorded evaluations
Each evaluation records what was tested and under which conditions.
- CARP-640M likelihood · FLIP2 IRED two-to-many; held-out test set · IRED
- Ridge (one-hot) · FLIP2 IRED two-to-many; held-out test set · IRED
- ESMC-300M supervised · FLIP2 IRED two-to-many; held-out test set · IRED
- ESMC-300M naive supervised · FLIP2 IRED two-to-many; held-out test set · IRED
- CARP-640M naive supervised · FLIP2 IRED two-to-many; held-out test set · IRED
- Ridge (one-hot + likelihoods) · FLIP2 IRED two-to-many; held-out test set · IRED
- ESM2-650M likelihood · FLIP2 IRED two-to-many; held-out test set · IRED
- Dayhoff likelihood · FLIP2 IRED two-to-many; held-out test set · IRED
- CARP-640M supervised · FLIP2 IRED two-to-many; held-out test set · IRED
Run instructions
No runnable recipe has been reviewed for this protocol. Dataset access, model requirements, licences and compute requirements must be checked against its sources before execution.
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
| Property and statement | Original source and location | Review and provenance |
|---|---|---|
| Relationship: part of discovery-benchmark-flip2 Individual claims | FLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications Table A5; row Ridge (one-hot); column spearman Version: Primary full-text snapshot retrieved 2026-09-17; exact bytes pinned by SHA-256 | source checked ai assisted source review · 2026-09-19 Audit detailsField: Claim: acquired-membership-41d62e8716093f949164 Source artifact SHA-256: Hash scope: Exact retrieved primary paper artifact bytes. |
Sources and history
View linked audit checks and correction history
Release 2026-09-19-eb6149e5d766 · Record review: discovered
1 source records and release history
- FLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications · Original source · Primary full-text snapshot retrieved 2026-09-17; exact bytes pinned by SHA-256
Technical metadata and extraction receipts
Stable ID: acquired-protocol-d0d5f30d1e07aa33d121
- source locator
- Table A5; row Ridge (one-hot); column spearman
- procedure
- FLIP2 IRED two-to-many; held-out test set
- comparison panels
- id: acquired-figure-d50b542e6ac928f343f2; title: FLIP2 IRED two-to-many; held-out test set · IRED: spearman; protocol id: acquired-protocol-d0d5f30d1e07aa33d121; dataset id: acquired-dataset-69828ada54ffa9b99628; metric: spearman; unit: correlation; direction: higher; result ids: acquired-result-65ce5f27fb8f68b25dce; acquired-result-4a8e21ac1fa10d3b644b; acquired-result-ab6342707a0b915fe424; acquired-result-52844b5ae6e2158a0c85; acquired-result-0266d7cc013aaa800b93; acquired-result-ea25d1feca38dc03a665; acquired-result-585bb988182cf2e96076; acquired-result-e203bae3f736b14d824e; acquired-result-13e7e4857d33c3728dca; source ids: evidence-expansion-flip2-d0e61ca2; source locator: Table A5; row Ridge (one-hot); column spearman through Table A5; row ESMC-300M naive supervised; column spearman; context: Complete selected source table is retained across source-order panels. These point estimates do not establish statistical significance or a universal ranking.; caveats: Source-specific evaluation. No equivalence to other releases, protocols or model families is inferred.; Exact source-defined evaluation scope; reported scores are not rewire reproductions.; Missing source cells and quarantined conflicts are recorded in acquisition and audit tables. Per-result scoring denominators may be unreported.; review: method: automated_source_review; date: 2026-09-19; id: acquired-figure-94a8fe620ad37b6a680e; title: FLIP2 IRED two-to-many; held-out test set · IRED: ndcg; protocol id: acquired-protocol-d0d5f30d1e07aa33d121; dataset id: acquired-dataset-69828ada54ffa9b99628; metric: ndcg; unit: unitless; direction: higher; result ids: acquired-result-6f94e0fc04e1346e9a9f; acquired-result-5f42feb3b46cf8cf0c70; acquired-result-1eb7a7b86baa71152c69; acquired-result-afdc7ddf4fc599e5d123; acquired-result-c5fadb81455dd1479a50; acquired-result-5d6eec3f9e8a96a2bf27; acquired-result-b9d9ebfa8f81ac42a6ab; acquired-result-9c4d0f5ee1c9a6e066a3; acquired-result-77387d2627c9fe69880b; source ids: evidence-expansion-flip2-d0e61ca2; source locator: Table A5; row Ridge (one-hot); column ndcg through Table A5; row ESMC-300M naive supervised; column ndcg; context: Complete selected source table is retained across source-order panels. These point estimates do not establish statistical significance or a universal ranking.; caveats: Source-specific evaluation. No equivalence to other releases, protocols or model families is inferred.; Exact source-defined evaluation scope; reported scores are not rewire reproductions.; Missing source cells and quarantined conflicts are recorded in acquisition and audit tables. Per-result scoring denominators may be unreported.; review: method: automated_source_review; date: 2026-09-19
- scope limitations
- Exact source-defined evaluation scope; reported scores are not rewire reproductions.
Related records
- part of: FLIP2
- benchmark: CARP-640M likelihood · FLIP2 IRED two-to-many; held-out test set · IRED
- benchmark: Ridge (one-hot) · FLIP2 IRED two-to-many; held-out test set · IRED
- benchmark: ESMC-300M supervised · FLIP2 IRED two-to-many; held-out test set · IRED
- benchmark: ESMC-300M naive supervised · FLIP2 IRED two-to-many; held-out test set · IRED
- benchmark: CARP-640M naive supervised · FLIP2 IRED two-to-many; held-out test set · IRED
- benchmark: Ridge (one-hot + likelihoods) · FLIP2 IRED two-to-many; held-out test set · IRED
- benchmark: ESM2-650M likelihood · FLIP2 IRED two-to-many; held-out test set · IRED
- benchmark: Dayhoff likelihood · FLIP2 IRED two-to-many; held-out test set · IRED
- benchmark: CARP-640M supervised · FLIP2 IRED two-to-many; held-out test set · IRED
- subject: FLIP2 IRED two-to-many; held-out test set: benchmark membership