Method: CNN Task: FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.34 Spearmanunitless · higher
Uncertainty: Not yet extracted
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Methods, coverage and source CNN on FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split
Trained on the split's training set with ten percent held out for validation, scored on its test set.
Aggregation: Not reported
flip primary benchmark evidence · Table 6, row(CNN), column(Mixed) Configuration: ESM-1b (mean) Task: FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.68 Spearmanunitless · higher
Uncertainty: Not yet extracted
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Methods, coverage and source ESM-1b (mean) on FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split
Trained on the split's training set with ten percent held out for validation, scored on its test set.
Aggregation: Not reported
flip primary benchmark evidence · Table 6, row(ESM-1b (mean)), column(Mixed) Configuration: ESM-1b (per AA) Task: FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.68 Spearmanunitless · higher
Uncertainty: Not yet extracted
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Methods, coverage and source ESM-1b (per AA) on FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split
Trained on the split's training set with ten percent held out for validation, scored on its test set.
Aggregation: Not reported
flip primary benchmark evidence · Table 6, row(ESM-1b (per AA)), column(Mixed) Configuration: ESM-1v (mean) Task: FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.67 Spearmanunitless · higher
Uncertainty: Not yet extracted
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Methods, coverage and source ESM-1v (mean) on FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split
Trained on the split's training set with ten percent held out for validation, scored on its test set.
Aggregation: Not reported
flip primary benchmark evidence · Table 6, row(ESM-1v (mean)), column(Mixed) Configuration: ESM-1v (per AA) Task: FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.65 Spearmanunitless · higher
Uncertainty: Not yet extracted
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Methods, coverage and source ESM-1v (per AA) on FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split
Trained on the split's training set with ten percent held out for validation, scored on its test set.
Aggregation: Not reported
flip primary benchmark evidence · Table 6, row(ESM-1v (per AA)), column(Mixed) Configuration: ESM-untrained (mean) Task: FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.36 Spearmanunitless · higher
Uncertainty: Not yet extracted
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Methods, coverage and source ESM-untrained (mean) on FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split
Trained on the split's training set with ten percent held out for validation, scored on its test set.
Aggregation: Not reported
flip primary benchmark evidence · Table 6, row(ESM-untrained (mean)), column(Mixed) Configuration: ESM-untrained (per AA) Task: FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.44 Spearmanunitless · higher
Uncertainty: Not yet extracted
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Methods, coverage and source ESM-untrained (per AA) on FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split
Trained on the split's training set with ten percent held out for validation, scored on its test set.
Aggregation: Not reported
flip primary benchmark evidence · Table 6, row(ESM-untrained (per AA)), column(Mixed) Method: Ridge Task: FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.17 Spearmanunitless · higher
Uncertainty: Not yet extracted
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Methods, coverage and source Ridge on FLIP THERMOSTABILITY-MIXED: Thermostability fitness prediction, Mixed split
Trained on the split's training set with ten percent held out for validation, scored on its test set.
Aggregation: Not reported
flip primary benchmark evidence · Table 6, row(Ridge), column(Mixed) Configuration: AMPLIFY-350M (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.585 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source AMPLIFY-350M on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 6, row 'Meltome mixed', column 'AMPLIFY-350M' (page 12 of 13) Configuration: Ankh-base (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.527 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source Ankh-base on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 7, row 'Meltome mixed', column 'Ankh-base' (page 13 of 13) Configuration: Ankh3-large (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.545 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source Ankh3-large on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 7, row 'Meltome mixed', column 'Ankh3-large' (page 13 of 13) Configuration: DPLM-650M (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.623 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source DPLM-650M on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 6, row 'Meltome mixed', column 'DPLM-650M' (page 12 of 13) Configuration: E1-600M (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.69 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source E1-600M on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 6, row 'Meltome mixed', column 'E1-600M' (page 12 of 13) Configuration: ESM-1b (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.593 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source ESM-1b on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 6, row 'Meltome mixed', column 'ESM-1b' (page 12 of 13) Configuration: ESM-2 650M (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.633 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source ESM-2 650M on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 6, row 'Meltome mixed', column 'ESM-2 650M' (page 12 of 13) Configuration: ESM C 600M (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.67 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source ESM C 600M on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 6, row 'Meltome mixed', column 'ESM C 600M' (page 12 of 13) Configuration: ESM3-open (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.573 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source ESM3-open on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 7, row 'Meltome mixed', column 'ESM3-open' (page 13 of 13) Configuration: HyPLM-Flash, BF16 evaluation (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.672 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source HyPLM-Flash on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 6, row 'Meltome mixed', column 'HyPLM-Flash' (page 12 of 13) Configuration: ProtBERT (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.621 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source ProtBERT on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 7, row 'Meltome mixed', column 'ProtBERT' (page 13 of 13) Configuration: ProtT5-XL (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.661 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source ProtT5-XL on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 7, row 'Meltome mixed', column 'ProtT5-XL' (page 13 of 13) Configuration: xTrimoPGLM-1B (HyPLM report) Protocol: Meltome mixed (HyPLM report) Dataset subset: FLIP Thermostability, Mixed split (FLIP split) 0.578 Spearman (Spearman correlation on the FLIP Meltome mixed split)unitless · higher
Uncertainty: Not reported by the source: Section 4 says intervals are paired bootstrap intervals; none is printed here.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The recent-PDB likelihood panel is described two ways. Section 4.3 and the Table 3 caption give 887 PDB chains released after 2 September 2026; Table 4 gives 887 RCSB PDB entries released 9 September to 7 October 2026. The start of the window and whether the unit is a chain or an entry both differ, so the panel cannot be rebuilt from the report.HyPLM Technical Report (Hyra Team, 2026-10-12) · Section 4.3; Table 3 caption; Table 4 row 'PDB time split'
Appendix A states that the authors could not reproduce xTrimoPGLM's published ProteinGym scores (their runs fall 0.07 to 0.09 below the leaderboard) and that its likelihood is implausibly high, pointing to an incompatibility between their adapter and its masked-language-model head. They withhold its zero-shot and likelihood results and keep its embedding-based results as unaffected, but those results come from the same adapter and that claim was not tested.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Deviations'
Appendix A states that for GB1 two-vs-rest, GB1 three-vs-rest and the cross-protein hydrolase split of FLIP2, validation data come from the training distribution while the test sets are strongly shifted, so probe selection is dominated by noise and numerically equivalent runs of the same model differ by up to 0.2. The authors print these cells but leave them out of every summary statistic.HyPLM Technical Report (Hyra Team, 2026-10-12) · Appendix A 'Statistics'
Methods, coverage and source xTrimoPGLM-1B on meltome-mixed (HyPLM report)
hyplm-flash-20261106-protocol-meltome-mixed
Aggregation: Not reported
HyPLM Technical Report (Hyra Team, 2026-10-12) · Table 7, row 'Meltome mixed', column 'xTrimoPGLM-1B' (page 13 of 13)