HyPLM technical report protein language model evaluation
Compares protein language models on zero-shot fitness, masked-residue likelihood and frozen-representation probes for structure, function, localisation and fitness.
Results are recorded on the linked protocols, tasks or configurations listed below. Their settings and evaluations are kept separate.
Overview
An explanatory profile has not yet been reviewed for this record.
Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.
Results on linked records
This benchmark has no results of its own in this release. 40 linked records have results. They are not assigned to the underlying benchmark: each record keeps its own settings and evaluations, and its scores are on its own page.
- AAV low-vs-high (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- AAV two-vs-many (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Amylase by mutation count (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Amylase one-to-many (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Arch retrieval (DGEB) (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Contacts, CASP14 (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Contacts, CASP15 (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Contacts, ProteinNet (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- EC classification (DGEB) (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- EC, DeepFRI split (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- EC, SaProt split (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Embedding throughput on one NVIDIA H20 (HyPLM report) Protocol · part of this benchmark · 1 evaluation · 1 result
- Fluorescence (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- GB1 low-vs-high (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- GB1 three-vs-rest (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- GB1 two-vs-rest (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- GO-BP (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- GO-CC (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- GO-MF (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- HumanPPI (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Hydrolase to P06241 (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Localization, 10-class (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Localization, binary (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Meltome human cell (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Meltome mixed (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Metal-ion binding (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- NLL, held-out UniRef panel (HyPLM report) Protocol · part of this benchmark · 12 evaluations · 12 results
- NLL, recent PDB chains (HyPLM report) Protocol · part of this benchmark · 12 evaluations · 12 results
- Masked-residue perplexity, held-out UniRef panel (HyPLM report) Protocol · part of this benchmark · 10 evaluations · 10 results
- Masked-residue perplexity, recent PDB chains (HyPLM report) Protocol · part of this benchmark · 10 evaluations · 10 results
- ProteinGym zero-shot, activity assays (HyPLM report) Protocol · part of this benchmark · 7 evaluations · 7 results
- ProteinGym zero-shot, binding assays (HyPLM report) Protocol · part of this benchmark · 7 evaluations · 7 results
- ProteinGym zero-shot, expression assays (HyPLM report) Protocol · part of this benchmark · 7 evaluations · 7 results
- ProteinGym zero-shot, organismal fitness assays (HyPLM report) Protocol · part of this benchmark · 7 evaluations · 7 results
- ProteinGym zero-shot paired difference from HyPLM-Flash (HyPLM report) Protocol · part of this benchmark · 4 evaluations · 4 results
- ProteinGym zero-shot, stability assays (HyPLM report) Protocol · part of this benchmark · 7 evaluations · 7 results
- ProteinGym zero-shot (HyPLM report) Protocol · part of this benchmark · 17 evaluations · 17 results
- Mean Spearman over ten regression splits (HyPLM report) Protocol · part of this benchmark · 5 evaluations · 5 results
- Remote homology (fold) (HyPLM report) Protocol · part of this benchmark · 13 evaluations · 13 results
- Secondary structure (CB513) (HyPLM report) Protocol · part of this benchmark · 14 evaluations · 14 results
2 further linked records are listed under Related records.
Methods and evaluation design
Procedure, tasks and evaluated configurations
Additional methodological details, tasks and evaluated configurations have not yet been extracted into a structured profile for this benchmark. Consult the linked sources and evaluation records below.
Baseline coverage
Reference methods help show what a model adds beyond simple controls. We track a null control and a conventional method for each protocol.
0 of 84 active baseline roles have published Rewire measurements in this release. Measurements on a selected protocol do not establish coverage of an entire suite.
Baseline status by linked protocol
- AAV low-vs-high (HyPLM report) · 0/2 roles measured
- AAV two-vs-many (HyPLM report) · 0/2 roles measured
- Amylase by mutation count (HyPLM report) · 0/2 roles measured
- Amylase one-to-many (HyPLM report) · 0/2 roles measured
- Arch retrieval (DGEB) (HyPLM report) · 0/2 roles measured
- Contacts, CASP14 (HyPLM report) · 0/2 roles measured
- Contacts, CASP15 (HyPLM report) · 0/2 roles measured
- Contacts, ProteinNet (HyPLM report) · 0/2 roles measured
- EC classification (DGEB) (HyPLM report) · 0/2 roles measured
- EC, DeepFRI split (HyPLM report) · 0/2 roles measured
- EC, SaProt split (HyPLM report) · 0/2 roles measured
- Embedding throughput on one NVIDIA H20 (HyPLM report) · 0/2 roles measured
- Fluorescence (HyPLM report) · 0/2 roles measured
- GB1 low-vs-high (HyPLM report) · 0/2 roles measured
- GB1 three-vs-rest (HyPLM report) · 0/2 roles measured
- GB1 two-vs-rest (HyPLM report) · 0/2 roles measured
- GO-BP (HyPLM report) · 0/2 roles measured
- GO-CC (HyPLM report) · 0/2 roles measured
- GO-MF (HyPLM report) · 0/2 roles measured
- HumanPPI (HyPLM report) · 0/2 roles measured
- Hydrolase to P06241 (HyPLM report) · 0/2 roles measured
- Localization, 10-class (HyPLM report) · 0/2 roles measured
- Localization, binary (HyPLM report) · 0/2 roles measured
- Meltome human cell (HyPLM report) · 0/2 roles measured
- Meltome mixed (HyPLM report) · 0/2 roles measured
- Metal-ion binding (HyPLM report) · 0/2 roles measured
- NLL, held-out UniRef panel (HyPLM report) · 0/2 roles measured
- NLL, recent PDB chains (HyPLM report) · 0/2 roles measured
- Masked-residue perplexity, held-out UniRef panel (HyPLM report) · 0/2 roles measured
- Masked-residue perplexity, recent PDB chains (HyPLM report) · 0/2 roles measured
- ProteinGym zero-shot, activity assays (HyPLM report) · 0/2 roles measured
- ProteinGym zero-shot, binding assays (HyPLM report) · 0/2 roles measured
- ProteinGym zero-shot, expression assays (HyPLM report) · 0/2 roles measured
- ProteinGym zero-shot, organismal fitness assays (HyPLM report) · 0/2 roles measured
- ProteinGym zero-shot paired difference from HyPLM-Flash (HyPLM report) · 0/2 roles measured
- ProteinGym zero-shot, stability assays (HyPLM report) · 0/2 roles measured
- ProteinGym zero-shot (HyPLM report) · 0/2 roles measured
- Mean Spearman over ten regression splits (HyPLM report) · 0/2 roles measured
- Remote homology (fold) (HyPLM report) · 0/2 roles measured
- Secondary structure (CB513) (HyPLM report) · 0/2 roles measured
- Stability (HyPLM report) · 0/2 roles measured
- Thermostability (HyPLM report) · 0/2 roles measured
Protocol coverage CSV (gzip) · Model evaluation matrix (gzip) · Source table (gzip) · Release and checksums (gzip)
Coverage is derived from release 2026-11-21-adc15377fab2. Source citations describe the original records; they do not validate an unreviewed baseline proposal. No results have been generated by this audit.
Run instructions
No runnable recipe has been reviewed for this benchmark. Dataset access, model requirements, licences and compute requirements must be checked against its sources before execution.
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.
0 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|
No evidence rows match these filters. Choose another scope or clear the search.
Sources and history
Release 2026-11-21-adc15377fab2 · Record review: source checked
1 source record and release history
- HyPLM Technical Report (Hyra Team, 2026-10-12) · Original source · Dated 2026-10-12 on page 1; retrieved at commit 0dcc533cc9f8f76c90ecbfe944df816b6672328a (13 pages)
Technical metadata and record identifiers
Technical metadata for this record is tracked under stable ID hyplm-flash-20261106-benchmark-report-evaluation.
Source locator: Section 4; Appendix A; Table 4
Related records
- evaluates task: ProteinGym mutation effects
- subject: HyPLM report: bootstrap
- subject: HyPLM report: figure1 split count
- subject: HyPLM report: figure1b dplm 650m lost
- subject: HyPLM report: figure1b dplm 650m won
- subject: HyPLM report: figure1b e1 600m lost
- subject: HyPLM report: figure1b e1 600m won
- subject: HyPLM report: figure1b esm 1b lost
- subject: HyPLM report: figure1b esm 1b won
- subject: HyPLM report: figure1b esm 2 3b lost
- subject: HyPLM report: figure1b esm 2 3b won
- subject: HyPLM report: figure1b esm 2 650m lost
- subject: HyPLM report: figure1b esm 2 650m won
- subject: HyPLM report: figure1b esm c 300m lost
- subject: HyPLM report: figure1b esm c 300m won
- subject: HyPLM report: figure1b esm c 600m lost
- subject: HyPLM report: figure1b esm c 600m won
- subject: HyPLM report: figure1b esm3 open lost
- subject: HyPLM report: figure1b esm3 open won
- subject: HyPLM report: leaderboard rank
- subject: HyPLM report: leaderboard reproduction
- subject: HyPLM report: model count
- subject: HyPLM report: noisy splits
- subject: HyPLM report: sequence only
- subject: HyPLM report: table5 provenance
- subject: HyPLM report: xtrimopglm gap
- part of: AAV low-vs-high (HyPLM report)
- part of: AAV two-vs-many (HyPLM report)
- part of: Amylase by mutation count (HyPLM report)
- part of: Amylase one-to-many (HyPLM report)
- part of: Arch retrieval (DGEB) (HyPLM report)
- part of: Contacts, CASP14 (HyPLM report)
- part of: Contacts, CASP15 (HyPLM report)
- part of: Contacts, ProteinNet (HyPLM report)
- part of: EC classification (DGEB) (HyPLM report)
- part of: EC, DeepFRI split (HyPLM report)
- part of: EC, SaProt split (HyPLM report)
- part of: Embedding throughput on one NVIDIA H20 (HyPLM report)
- part of: Fluorescence (HyPLM report)
- part of: GB1 low-vs-high (HyPLM report)
- part of: GB1 three-vs-rest (HyPLM report)
- part of: GB1 two-vs-rest (HyPLM report)
- part of: GO-BP (HyPLM report)
- part of: GO-CC (HyPLM report)
- part of: GO-MF (HyPLM report)
- part of: HumanPPI (HyPLM report)
- part of: Hydrolase to P06241 (HyPLM report)
- part of: Localization, 10-class (HyPLM report)
- part of: Localization, binary (HyPLM report)
- part of: Meltome human cell (HyPLM report)
- part of: Meltome mixed (HyPLM report)
- part of: Metal-ion binding (HyPLM report)
- part of: NLL, held-out UniRef panel (HyPLM report)
- part of: NLL, recent PDB chains (HyPLM report)
- part of: Masked-residue perplexity, held-out UniRef panel (HyPLM report)
- part of: Masked-residue perplexity, recent PDB chains (HyPLM report)
- part of: ProteinGym zero-shot, activity assays (HyPLM report)
- part of: ProteinGym zero-shot, binding assays (HyPLM report)
- part of: ProteinGym zero-shot, expression assays (HyPLM report)
- part of: ProteinGym zero-shot, organismal fitness assays (HyPLM report)
- part of: ProteinGym zero-shot paired difference from HyPLM-Flash (HyPLM report)
- part of: ProteinGym zero-shot, stability assays (HyPLM report)
- part of: ProteinGym zero-shot (HyPLM report)
- part of: Mean Spearman over ten regression splits (HyPLM report)
- part of: Remote homology (fold) (HyPLM report)
- part of: Secondary structure (CB513) (HyPLM report)
- part of: Stability (HyPLM report)
- part of: Thermostability (HyPLM report)