HyInv technical report inverse-folding comparison
Compares inverse-folding methods by the structural consistency, native recovery and diversity of the sequences they design for held-out protein backbones.
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. 19 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.
- Latency of a design request on one NVIDIA H20 (HyInv report) Protocol · part of this benchmark · 13 evaluations · 23 results
- Speed-up of a design request over ProteinMPNN on one NVIDIA H20 (HyInv report) Protocol · part of this benchmark · 4 evaluations · 4 results
- CA-lDDT of designed sequences under ESMFold2 (HyInv report) Protocol · part of this benchmark · 60 evaluations · 60 results
- Novel and foldable designs per de novo backbone at matched diversity (HyInv report) Protocol · part of this benchmark · 3 evaluations · 6 results
- pLDDT of ESMFold2 predictions of designed sequences (HyInv report) Protocol · part of this benchmark · 60 evaluations · 60 results
- Self-consistency RMSD of designed sequences under ESMFold2 (HyInv report) Protocol · part of this benchmark · 60 evaluations · 60 results
- scTM of designed sequences under Boltz2 (HyInv report) Protocol · part of this benchmark · 12 evaluations · 12 results
- Paired scTM difference from HyInv under Boltz2 (HyInv report) Protocol · part of this benchmark · 10 evaluations · 10 results
- Paired scTM difference from HyInv under ESMFold2 (HyInv report) Protocol · part of this benchmark · 50 evaluations · 50 results
- Paired scTM difference from HyInv under Protenix2 (HyInv report) Protocol · part of this benchmark · 10 evaluations · 10 results
- scTM of designed sequences under ESMFold2 (HyInv report) Protocol · part of this benchmark · 60 evaluations · 60 results
- BF16 path minus reference sampler in scTM at fixed diversity (HyInv report) Protocol · part of this benchmark · 4 evaluations · 4 results
- scTM at fixed sequence diversity on the development panel (HyInv report) Protocol · part of this benchmark · 2 evaluations · 2 results
- Gain in scTM at fixed diversity from reinforcement learning (HyInv report) Protocol · part of this benchmark · 1 evaluation · 1 result
- Gain in scTM at fixed diversity over the unmodified sampler (HyInv report) Protocol · part of this benchmark · 12 evaluations · 12 results
- HyInv above the ProteinMPNN temperature curve at HyInv's diversity (HyInv report) Protocol · part of this benchmark · 2 evaluations · 2 results
- scTM of designed sequences under Protenix2 (HyInv report) Protocol · part of this benchmark · 12 evaluations · 12 results
- Sequence diversity of a design set (HyInv report) Protocol · part of this benchmark · 55 evaluations · 55 results
- Native sequence recovery of designed sequences (HyInv report) Protocol · part of this benchmark · 55 evaluations · 55 results
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 38 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
- Latency of a design request on one NVIDIA H20 (HyInv report) · 0/2 roles measured
- Speed-up of a design request over ProteinMPNN on one NVIDIA H20 (HyInv report) · 0/2 roles measured
- CA-lDDT of designed sequences under ESMFold2 (HyInv report) · 0/2 roles measured
- Novel and foldable designs per de novo backbone at matched diversity (HyInv report) · 0/2 roles measured
- pLDDT of ESMFold2 predictions of designed sequences (HyInv report) · 0/2 roles measured
- Self-consistency RMSD of designed sequences under ESMFold2 (HyInv report) · 0/2 roles measured
- scTM of designed sequences under Boltz2 (HyInv report) · 0/2 roles measured
- Paired scTM difference from HyInv under Boltz2 (HyInv report) · 0/2 roles measured
- Paired scTM difference from HyInv under ESMFold2 (HyInv report) · 0/2 roles measured
- Paired scTM difference from HyInv under Protenix2 (HyInv report) · 0/2 roles measured
- scTM of designed sequences under ESMFold2 (HyInv report) · 0/2 roles measured
- BF16 path minus reference sampler in scTM at fixed diversity (HyInv report) · 0/2 roles measured
- scTM at fixed sequence diversity on the development panel (HyInv report) · 0/2 roles measured
- Gain in scTM at fixed diversity from reinforcement learning (HyInv report) · 0/2 roles measured
- Gain in scTM at fixed diversity over the unmodified sampler (HyInv report) · 0/2 roles measured
- HyInv above the ProteinMPNN temperature curve at HyInv's diversity (HyInv report) · 0/2 roles measured
- scTM of designed sequences under Protenix2 (HyInv report) · 0/2 roles measured
- Sequence diversity of a design set (HyInv report) · 0/2 roles measured
- Native sequence recovery of designed sequences (HyInv 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
- HyInv Technical Report (Hyra Team, 2026-10-14) · Original source · Dated 2026-10-14; retrieved at commit 0dcc533cc9f8f76c90ecbfe944df816b6672328a (21 pages)
Technical metadata and record identifiers
Technical metadata for this record is tracked under stable ID hyinv-20261106-benchmark-report-comparison.
Source locator: Sections 5 and 6; Table 3; Appendices A, D and G
Related records
- evaluates task: Protein design / inverse folding
- subject: HyInv report: application enzymes
- subject: HyInv report: application homology screen
- subject: HyInv report: homology screen of De novo backbones
- subject: HyInv report: homology screen of Loop and window, relaxed
- subject: HyInv report: homology screen of Loop and window, strict
- subject: HyInv report: homology screen of Named enzymes
- subject: HyInv report: homology screen of Pocket library
- subject: HyInv report: application homology screen
- subject: HyInv report: application loops
- subject: HyInv report: application pocket library
- subject: HyInv report: application statistics
- subject: HyInv report: baseline diversity mismatch
- subject: HyInv report: baseline reproduction check
- subject: HyInv report: reproduction check for CMLM
- subject: HyInv report: reproduction check for DPLM-2 3B
- subject: HyInv report: reproduction check for LigandMPNN
- subject: HyInv report: reproduction check for LM-Design 650M
- subject: HyInv report: reproduction check for PiFold
- subject: HyInv report: reproduction check for ProteinMPNN
- subject: HyInv report: reproduction check for ProteinMPNN grouped
- subject: HyInv report: reproduction check for ProteinMPNN (low-noise, grouped)
- subject: HyInv report: reproduction check for ProteinZero V2 (ESM1)
- subject: HyInv report: reproduction check for ProteinZero V2 (ESM2)
- subject: HyInv report: baseline reproduction check
- subject: HyInv report: checkpoint selection
- subject: HyInv report: panels
- subject: HyInv report: evaluation panel CATH4.2 test
- subject: HyInv report: evaluation panel CATH4.3 test
- subject: HyInv report: evaluation panel CATH4.3 validation
- subject: HyInv report: evaluation panel RecentPDB
- subject: HyInv report: panels
- subject: HyInv report: evaluation panel TS50
- subject: HyInv report: evaluation panel TS500
- subject: HyInv report: predictor independence
- subject: HyInv report: release location
- subject: HyInv report: reproducibility of report
- subject: HyInv report: run to run variation
- part of: Latency of a design request on one NVIDIA H20 (HyInv report)
- part of: Speed-up of a design request over ProteinMPNN on one NVIDIA H20 (HyInv report)
- part of: CA-lDDT of designed sequences under ESMFold2 (HyInv report)
- part of: Novel and foldable designs per de novo backbone at matched diversity (HyInv report)
- part of: pLDDT of ESMFold2 predictions of designed sequences (HyInv report)
- part of: Self-consistency RMSD of designed sequences under ESMFold2 (HyInv report)
- part of: scTM of designed sequences under Boltz2 (HyInv report)
- part of: Paired scTM difference from HyInv under Boltz2 (HyInv report)
- part of: Paired scTM difference from HyInv under ESMFold2 (HyInv report)
- part of: Paired scTM difference from HyInv under Protenix2 (HyInv report)
- part of: scTM of designed sequences under ESMFold2 (HyInv report)
- part of: BF16 path minus reference sampler in scTM at fixed diversity (HyInv report)
- part of: scTM at fixed sequence diversity on the development panel (HyInv report)
- part of: Gain in scTM at fixed diversity from reinforcement learning (HyInv report)
- part of: Gain in scTM at fixed diversity over the unmodified sampler (HyInv report)
- part of: HyInv above the ProteinMPNN temperature curve at HyInv's diversity (HyInv report)
- part of: scTM of designed sequences under Protenix2 (HyInv report)
- part of: Sequence diversity of a design set (HyInv report)
- part of: Native sequence recovery of designed sequences (HyInv report)