Model type
Geometric encoder and inverse-folding transformer
Configuration
ESM-IF1
ESM-IF1 designs protein sequences conditioned on backbone coordinates.
Sources
facebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained ModelsHow it worksESM-IF1 workflow
Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.
Sources
facebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained ModelsOverview
Inputs
Protein backbone atom coordinates; the model supports missing backbone spans.
Outputs
Sampled protein sequences or conditional sequence likelihoods.
Access
Official project documentation and implementation: https://github.com/facebookresearch/esm
Sources
facebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Modelslimited source coverage · Automated source review, 2026-09-16. All specifications and missing details
Evaluations and results
1 evaluation · 5 results. Different protocols are not a single leaderboard.
Filter evaluations
Applied filters: All linked evaluations
| Tested configuration | Protocol and dataset | Finding | Evidence and details |
|---|---|---|---|
| Configuration: ESM-IF1 | Task: ProteinGym ZS-SUB-AUC: Zero-shot substitutions, AUC Dataset subset: ProteinGym substitution DMS assays (ProteinGym split) | 0.73 auc fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM-IF1 on ProteinGym ZS-SUB-AUC: Zero-shot substitutions, AUC Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes. Aggregation: Not reported ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(Inverse ESM-IF1), column(Zero-shot substitutions, AUC) |
| Configuration: ESM-IF1 | Task: ProteinGym ZS-SUB-MCC: Zero-shot substitutions, MCC Dataset subset: ProteinGym substitution DMS assays (ProteinGym split) | 0.331 mcc correlation · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM-IF1 on ProteinGym ZS-SUB-MCC: Zero-shot substitutions, MCC Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes. Aggregation: Not reported ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(Inverse ESM-IF1), column(Zero-shot substitutions, MCC) |
| Configuration: ESM-IF1 | Task: ProteinGym ZS-SUB-NDCG: Zero-shot substitutions, NDCG@10% Dataset subset: ProteinGym substitution DMS assays (ProteinGym split) | 0.748 ndcg fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM-IF1 on ProteinGym ZS-SUB-NDCG: Zero-shot substitutions, NDCG@10% Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes. Aggregation: Not reported ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(Inverse ESM-IF1), column(Zero-shot substitutions, NDCG@10%) |
| Configuration: ESM-IF1 | Task: ProteinGym ZS-SUB-RECALL: Zero-shot substitutions, top 10% recall Dataset subset: ProteinGym substitution DMS assays (ProteinGym split) | 0.223 recall fraction · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM-IF1 on ProteinGym ZS-SUB-RECALL: Zero-shot substitutions, top 10% recall Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes. Aggregation: Not reported ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(Inverse ESM-IF1), column(Zero-shot substitutions, top 10% recall) |
| Configuration: ESM-IF1 | Task: ProteinGym ZS-SUB-SPEARMAN: Zero-shot substitutions, Spearman Dataset subset: ProteinGym substitution DMS assays (ProteinGym split) | 0.422 spearman correlation · higher Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible | Author-reported evaluation · Source checkedMethods, coverage and sourceESM-IF1 on ProteinGym ZS-SUB-SPEARMAN: Zero-shot substitutions, Spearman Zero-shot scoring of substitution assays, averaged over assays with the correction the paper describes. Aggregation: Not reported ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Table 2, row(Inverse ESM-IF1), column(Zero-shot substitutions, Spearman) |
Source checking is not independent reproduction. Release 2026-09-29-06401fd5b220.
Use this model
How it works, versions and access
Related profile: ESM-IF1. This page retains the exact record and its evaluation context.
This configuration
Fitness prediction model evaluated by the ProteinGym authors under their harness.
- record
- ESM-IF1
- configuration
- Not reported
- entity type
- Configuration
How it works
How it works
ESM-IF1 designs protein sequences conditioned on backbone coordinates. Geometric-vector-perceptron input processing followed by a sequence-to-sequence transformer. The documented inputs are protein backbone atom coordinates; the model supports missing backbone spans. The output consists of sampled protein sequences or conditional sequence likelihoods.
Sources
facebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained ModelsVersions and reproducibility
esm_if1_gvp4_t16_142M_UR50. The applicable input limits require configuration-specific checking.
Sources
facebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained ModelsStrengths, limitations and unresolved questions
Strengths and limitations
Strengths and considerations
- Structure-conditioned design can use partially missing backbones because training included span masking.
Sources
facebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models
Limitations and conditions
- The public checkpoint name contains 142M but the README parameter table reports 124M. This discrepancy is retained rather than silently selecting a total.
Sources
facebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models
Profile review details
Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied.
Stable record: discovery-model-esm-if1Specifications
Inputs, training, access and other details
Explanatory profile: limited source coverage · Automated source review, 2026-09-16. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.
| Property | Description and evidence |
|---|---|
| Model type | Geometric encoder and inverse-folding transformerSourcesfacebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models |
| Architecture | Geometric-vector-perceptron input processing followed by a sequence-to-sequence transformer.Sourcesfacebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models |
| Inputs | Protein backbone atom coordinates; the model supports missing backbone spans.Sourcesfacebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models |
| Outputs | Sampled protein sequences or conditional sequence likelihoods.Sourcesfacebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models |
| Parameters | The official checkpoint identifier contains 142M but the same repository model table states 124M. Both values are retained as a source discrepancy, without choosing a total.Sourcesfacebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models |
| Known versions | esm_if1_gvp4_t16_142M_UR50.Sourcesfacebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models |
| Training data | CATH 4.3 and predicted UniRef50 structures; README reports 12M structures predicted by AlphaFold2.Sourcesfacebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models |
| Training cutoff | The official model table identifies CATH 4.3 and predicted UniRef50 structures, but does not state a common latest-structure or sequence date. · Not reported in inspected sourcesSourcesfacebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models |
| Context limits | The reviewed inverse-folding usage and model table do not establish a universal maximum backbone length; the structural graph and selected inference configuration determine resource use. · Not reported in inspected sourcesSourcesfacebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models |
| Weights licence | Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant. · Not reported in inspected sourcesSources (2)facebookresearch/esm: README.md; facebookresearch/esm: LICENSE · README.md: Inverse folding and Pre-trained Models; LICENSE: licence text |
| Access | Official project documentation and implementation: https://github.com/facebookresearch/esmSourcesfacebookresearch/esm: README.md · README.md: Inverse folding and Pre-trained Models |
| Code licence | MITSourcesfacebookresearch/esm: LICENSE · LICENSE: licence text |
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
| Property and statement | Original source and location | Review and provenance |
|---|---|---|
| Relationship: family discovery-model-esm-if1 Individual claims | ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design Section 4.1 Zero-shot benchmarks; Table 2 named model row Version: Primary full-text snapshot retrieved 2026-09-17; exact bytes pinned by SHA-256 | source checked automated source review · 2026-09-23 Audit detailsSource review establishes this relationship only. Exact evaluated configurations and original numerical review status remain unchanged. The primary table and baseline descriptions identify the evaluated family. ESM-1v remains an ensemble; ESM-2 remains 15B; inverse-folding conditions remain distinct. Field: Claim: model-evaluation-identity-41267a3a1f7abb612d10 Source artifact SHA-256: Hash scope: Exact retrieved primary paper artifact bytes. |
Sources and history
View linked audit checks and correction history
Release 2026-09-29-06401fd5b220 · Record review: source checked
1 source records and release history
- ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design · Original source · Primary full-text snapshot retrieved 2026-09-17; exact bytes pinned by SHA-256
Technical metadata and extraction receipts
Stable ID: proteingym-method-esm-if1
- areas
- proteins-complexes
- source locator
- Table 2, row(Inverse ESM-IF1)
- missing metadata
- checkpoint revision: unreported; parameters: unextracted
Related records
- family: ESM-IF1
- subject: ESM-IF1: family ESM-IF1
- model: ESM-IF1 on ProteinGym ZS-SUB-AUC: Zero-shot substitutions, AUC
- model: ESM-IF1 on ProteinGym ZS-SUB-MCC: Zero-shot substitutions, MCC
- model: ESM-IF1 on ProteinGym ZS-SUB-NDCG: Zero-shot substitutions, NDCG@10%
- model: ESM-IF1 on ProteinGym ZS-SUB-RECALL: Zero-shot substitutions, top 10% recall
- model: ESM-IF1 on ProteinGym ZS-SUB-SPEARMAN: Zero-shot substitutions, Spearman