rewire.itbenchmarks
Configuration

ESM-2 8M masked-marginal scoring (ProteinGym v1.3 AMFR substitution assay)

ESM-2 is a family of protein sequence encoders that produce representations for downstream protein analyses.

Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json

1 evaluation · 5 results

How it worksESM-2 workflow
ESM-2 workflow1. Protein sequence. Then: 2. Transformer layers. Then: 3. Residue embeddings. Then: 4. Specified downstream analysisESM-2 workflow1. Protein sequence. Then: 2. Transformer layers. Then: 3. Residue embeddings. Then: 4. Specified downstream analysisESM-2 workflow1. Protein sequence. Then: 2. Transformer layers. Then: 3. Residue embeddings. Then: 4. Specified downstream analysis

Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.

Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json

Overview

Model type

Masked-token protein transformer encoder

Inputs

Single amino-acid sequences.

Outputs

Residue embeddings, sequence representations and masked-token predictions.

Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json

limited source coverage · Automated source review, 2026-09-23. 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

Exact evaluated configurations and original reported results
Tested configurationProtocol and datasetFindingEvidence and details
Configuration: ESM-2 8M masked-marginal scoring (ProteinGym v1.3 AMFR substitution assay)Protocol: ProteinGym v1.3 AMFR substitution assay
Dataset subset: ProteinGym v1.3 AMFR substitution assay full assay
0.394 AUC
dimensionless · higher

Uncertainty: Not reported

Coverage: 2972/2972

Rewire evaluation · Source checked
Methods, coverage and source

ESM-2 8M masked-marginal scoring on ProteinGym v1.3 AMFR substitution assay

ProteinGym v1.3 AMFR substitution assay

Aggregation: Not reported

ESM-2 8M masked-marginal scoring: local execution report (20 September 2026); ESM-2 8M masked-marginal scoring: automated execution audit · protocol_results.per_assay.AMFR_HUMAN_Tsuboyama_2023_4G3O.metrics.AUC
Configuration: ESM-2 8M masked-marginal scoring (ProteinGym v1.3 AMFR substitution assay)Protocol: ProteinGym v1.3 AMFR substitution assay
Dataset subset: ProteinGym v1.3 AMFR substitution assay full assay
-0.139 MCC
dimensionless · higher

Uncertainty: Not reported

Coverage: 2972/2972

Rewire evaluation · Source checked
Methods, coverage and source

ESM-2 8M masked-marginal scoring on ProteinGym v1.3 AMFR substitution assay

ProteinGym v1.3 AMFR substitution assay

Aggregation: Not reported

ESM-2 8M masked-marginal scoring: local execution report (20 September 2026); ESM-2 8M masked-marginal scoring: automated execution audit · protocol_results.per_assay.AMFR_HUMAN_Tsuboyama_2023_4G3O.metrics.MCC
Configuration: ESM-2 8M masked-marginal scoring (ProteinGym v1.3 AMFR substitution assay)Protocol: ProteinGym v1.3 AMFR substitution assay
Dataset subset: ProteinGym v1.3 AMFR substitution assay full assay
0.44 NDCG
dimensionless · higher

Uncertainty: Not reported

Coverage: 2972/2972

Rewire evaluation · Source checked
Methods, coverage and source

ESM-2 8M masked-marginal scoring on ProteinGym v1.3 AMFR substitution assay

ProteinGym v1.3 AMFR substitution assay

Aggregation: Not reported

ESM-2 8M masked-marginal scoring: local execution report (20 September 2026); ESM-2 8M masked-marginal scoring: automated execution audit · protocol_results.per_assay.AMFR_HUMAN_Tsuboyama_2023_4G3O.metrics.NDCG
Configuration: ESM-2 8M masked-marginal scoring (ProteinGym v1.3 AMFR substitution assay)Protocol: ProteinGym v1.3 AMFR substitution assay
Dataset subset: ProteinGym v1.3 AMFR substitution assay full assay
-0.209 Spearman
dimensionless · higher

Uncertainty: Not reported

Coverage: 2972/2972

Rewire evaluation · Source checked
Methods, coverage and source

ESM-2 8M masked-marginal scoring on ProteinGym v1.3 AMFR substitution assay

ProteinGym v1.3 AMFR substitution assay

Aggregation: Not reported

ESM-2 8M masked-marginal scoring: local execution report (20 September 2026); ESM-2 8M masked-marginal scoring: automated execution audit · protocol_results.per_assay.AMFR_HUMAN_Tsuboyama_2023_4G3O.metrics.Spearman
Configuration: ESM-2 8M masked-marginal scoring (ProteinGym v1.3 AMFR substitution assay)Protocol: ProteinGym v1.3 AMFR substitution assay
Dataset subset: ProteinGym v1.3 AMFR substitution assay full assay
0.057 Top_recall
dimensionless · higher

Uncertainty: Not reported

Coverage: 2972/2972

Rewire evaluation · Source checked
Methods, coverage and source

ESM-2 8M masked-marginal scoring on ProteinGym v1.3 AMFR substitution assay

ProteinGym v1.3 AMFR substitution assay

Aggregation: Not reported

ESM-2 8M masked-marginal scoring: local execution report (20 September 2026); ESM-2 8M masked-marginal scoring: automated execution audit · protocol_results.per_assay.AMFR_HUMAN_Tsuboyama_2023_4G3O.metrics.Top_recall

Source checking is not independent reproduction. Release 2026-09-25-8af07e960e5f.

Use this model

How it works, versions and access

Related profile: ESM-2. This page retains the exact record and its evaluation context.

This configuration

unreported; UniRef50 pretraining may overlap benchmark proteins

record
ESM-2 8M masked-marginal scoring (ProteinGym v1.3 AMFR substitution assay)
configuration
Not reported
entity type
Configuration

How it works

How it works

ESM-2 tokenizes an amino-acid sequence and uses a transformer encoder trained to recover masked residues. Self-attention lets each residue representation depend on its sequence context. The released model returns token probabilities and embeddings; a specified pooling rule, task head or complete folding pipeline is needed for a particular biological prediction.

Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
Versions and reproducibility

ESM-2 checkpoint identifiers encode layer count, parameter scale and training-data tag. The checked esm2_t33_650M_UR50D configuration lists max_position_embeddings=1,026. This configuration field includes model positions and is not a claim of training or validated inference on 1,026 amino acids.

Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
Strengths, limitations and unresolved questions

Strengths and limitations

Profile review details

Follow-up review of Reference checkpoint, Context limits, Training data release, Training cutoff. Source locations and before/after decisions are recorded in the 23 September profile-evidence audit. Other explanatory content retains its earlier source scope. No human scientific review or independent reproduction is implied.

Stable record: discovery-model-esm-2

Specifications

Inputs, training, access and other details

Explanatory profile: limited source coverage · Automated source review, 2026-09-23. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.

Inputs, outputs and configuration
PropertyDescription and evidence
Model typeMasked-token protein transformer encoder
Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
ArchitectureMasked-token protein transformer encoder; the checked 650M checkpoint has 33 layers, hidden width 1,280, 20 attention heads and rotary positional encoding.
Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
InputsSingle amino-acid sequences.
Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
OutputsResidue embeddings, sequence representations and masked-token predictions.
Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
ParametersReleased scales: 8M, 35M, 150M, 650M, 3B and 15B.
Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
Known versionsESM-2 checkpoint identifiers encode layer count, parameter scale and training-data tag.
Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
Training dataUniRef50 clusters with UniRef90 sampling; the pretrained-model table labels UR50/D 2021_04.
Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
Training cutoffThe inspected official table establishes an April 2021 UniRef release, but no separate latest-deposited-sequence date. Keep the corpus release distinct from a temporal leakage guarantee. · Not reported in inspected sources
Sourcesfacebookresearch/esm: README.md · README: Pre-trained Models and Pre-training Dataset Split
Context limitsThe official extraction script truncates to 1,022 residues by default; --truncation_seq_length is configurable. This workflow default is not a universal architecture or validated biological-context limit.
Sourcesesm2 extract: primary artifact · create_parser: --truncation_seq_length; run: get_batch_converter
Weights licenceThe official facebook/esm2_t33_650M_UR50D model card declares MIT; this is the inspected checkpoint, not a licence inference from source code.
Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
AccessOfficial project documentation and implementation: https://github.com/facebookresearch/esm
Sources (3)facebookresearch/esm: README.md; facebook/esm2_t33_650M_UR50D: README.md; facebook/esm2_t33_650M_UR50D: config.json · ESM README: Pre-trained Models and Main models; official facebook/esm2_t33_650M_UR50D README licence metadata and config.json
Code licenceMIT
Sourcesfacebookresearch/esm: LICENSE · LICENSE: licence text
Reference checkpointExample release: facebook/esm2_t33_650M_UR50D at 08e4846e537177426273712802403f7ba8261b6c. Its model.safetensors has registry-reported SHA-256 a08adabb949fa67ad3c14b509d04fd60368b35007b0095e3358f81200c4f4db0. This identifies a downloadable 650M checkpoint, not every ESM-2 evaluation.
Sourcesesm2 release: primary artifact · sha; siblings[model.safetensors].lfs.sha256
Training data releaseThe official pretrained-model table labels ESM-2 training data UR50/D 2021_04.
Sourcesfacebookresearch/esm: README.md · README: Pre-trained Models table, ESM-2 rows

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.

2 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-09-25-8af07e960e5f
Property and statementOriginal source and locationReview and provenance
Relationship: family
catalog-model-esm-2
Individual claims
ESM-2 8M masked-marginal scoring: local execution report (20 September 2026)

Original source ↗

report.json: model.checkpoint and artifacts checkpoint_url/checkpoint_sha256; esm2_t6_8M_UR50D

Version: ca73fa47136d182f2d4ddb083d084712198fc0e2
Retrieved: 2026-09-20T20:54:03Z

source checked

automated source review · 2026-09-23

Audit details

Source review establishes this relationship only. Exact evaluated configurations and original numerical review status remain unchanged. Pinned run report explicitly identifies the 8M checkpoint and exact weight hash. Preserve masked-marginal ProteinGym assay protocol; this supports the size-qualified8M record without aliasing650M/15B variants.

Field: links:family:catalog-model-esm-2

Claim: model-evaluation-identity-e3db6ecc999ab32164a3

Source artifact SHA-256: 6a14b3866141d5779c76da8071d85c73b7cce0dc4e101c93122706873e847d26

Hash scope: Hash scope not separately documented; inspect source record

Inspected artifact

Relationship: family
discovery-model-esm-2
Individual claims
ESM-2 8M masked-marginal scoring: local execution report (20 September 2026)

Original source ↗

model; model_configuration; execution

Version: ca73fa47136d182f2d4ddb083d084712198fc0e2
Retrieved: 2026-09-20T20:54:03Z

source checked

automated execution evidence review · 2026-09-20T20:54:03Z

Audit details

Field: links:family:discovery-model-esm-2

Claim: rewire-local-20260920-configuration-proteingym-esm2-family-discovery-model-esm-2

Source artifact SHA-256: 6a14b3866141d5779c76da8071d85c73b7cce0dc4e101c93122706873e847d26

Hash scope: Hash scope not separately documented; inspect source record

Inspected artifact

Sources and history

View linked audit checks and correction history

Release 2026-09-25-8af07e960e5f · Record review: source checked

1 source records and release historyDownload this release
Technical metadata and extraction receipts

Stable ID: rewire-local-20260920-configuration-proteingym-esm2

areas
proteins-complexes
source locator
model; model_configuration; execution; environment
configuration
checkpoint: esm2_t6_8M_UR50D; device: cpu; seed: 0; strategy: masked_marginals; threads: 1
execution
adapter: ESM2Adapter; adapter provenance: checkpoint sha256: 46f002a9870c9bdecd0ea887acb1f9a38a6b561e8f8bf8a6990b679b9d31b928; checkpoint url: https://dl.fbaipublicfiles.com/fair-esm/models/esm2_t6_8M_UR50D.pt; device: cpu; fair esm version: 2.0.0; implementation sha256: 4ff38cfbb91d718b52a44da53aa5594fe73d9d1d4d0ce43724807597c3929889; maximum sequence length: 1022; model: esm2_t6_8M_UR50D; published result reproduction: false; strategy: masked_marginals_wild_type_context_sum_log_odds; training overlap: unreported; UniRef50 pretraining may overlap benchmark proteins; batch size: 32; fitting: none; inference and fit seconds: 0.18779424999956973; mode: local_adapter; prediction type: scalar
model
name: ESM-2 esm2_t6_8M_UR50D masked marginals; training overlap: unreported; UniRef50 pretraining may overlap benchmark proteins
environment
architecture: arm64; container digest: unreported; container runtime: unreported; platform: macOS-26.6.2-arm64-arm-64bit; python: 3.11.13; rewirebench: 0.4.0; sdk code sha256: fcc5ab65e6753cf29aabbdae231cbb5cdf72a3dd40ba8f109802c33223b33536; sif sha256: unreported
packages
None recorded
method role
public model
training overlap
unreported; UniRef50 pretraining may overlap benchmark proteins
Related records

Suggest a correction