Configuration: TorchCraft on A100 with cuequivariance kernels Protocol: TorchCraft runtime per design at 100 residues Dataset: TorchCraft runtime benchmark design runs 3.92 Runtime (wall-clock minutes per design at 100 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 100 residues, GPU w/ cuequivariance
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "GPU w/ cuequivariance", column "100 aa" Configuration: TorchCraft on A100 with cuequivariance kernels Protocol: TorchCraft runtime per design at 200 residues Dataset: TorchCraft runtime benchmark design runs 4.27 Runtime (wall-clock minutes per design at 200 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 200 residues, GPU w/ cuequivariance
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "GPU w/ cuequivariance", column "200 aa" Configuration: TorchCraft on A100 with cuequivariance kernels Protocol: TorchCraft runtime per design at 300 residues Dataset: TorchCraft runtime benchmark design runs 4.54 Runtime (wall-clock minutes per design at 300 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 300 residues, GPU w/ cuequivariance
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "GPU w/ cuequivariance", column "300 aa" Configuration: TorchCraft on A100 with cuequivariance kernels Protocol: TorchCraft runtime per design at 400 residues Dataset: TorchCraft runtime benchmark design runs 5.85 Runtime (wall-clock minutes per design at 400 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 400 residues, GPU w/ cuequivariance
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "GPU w/ cuequivariance", column "400 aa" Configuration: TorchCraft on A100 with cuequivariance kernels Protocol: TorchCraft runtime per design at 500 residues Dataset: TorchCraft runtime benchmark design runs 7.59 Runtime (wall-clock minutes per design at 500 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 500 residues, GPU w/ cuequivariance
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "GPU w/ cuequivariance", column "500 aa" Configuration: TorchCraft on A100 without cuequivariance kernels Protocol: TorchCraft runtime per design at 100 residues Dataset: TorchCraft runtime benchmark design runs 4.76 Runtime (wall-clock minutes per design at 100 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 100 residues, GPU w/o cuequivariance
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "GPU w/o cuequivariance", column "100 aa" Configuration: TorchCraft on A100 without cuequivariance kernels Protocol: TorchCraft runtime per design at 200 residues Dataset: TorchCraft runtime benchmark design runs 7.96 Runtime (wall-clock minutes per design at 200 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 200 residues, GPU w/o cuequivariance
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "GPU w/o cuequivariance", column "200 aa" Configuration: TorchCraft on A100 without cuequivariance kernels Protocol: TorchCraft runtime per design at 300 residues Dataset: TorchCraft runtime benchmark design runs 13.1 Runtime (wall-clock minutes per design at 300 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 300 residues, GPU w/o cuequivariance
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "GPU w/o cuequivariance", column "300 aa" Configuration: TorchCraft on A100 without cuequivariance kernels Protocol: TorchCraft runtime per design at 400 residues Dataset: TorchCraft runtime benchmark design runs 21.6 Runtime (wall-clock minutes per design at 400 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 400 residues, GPU w/o cuequivariance
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "GPU w/o cuequivariance", column "400 aa" Configuration: TorchCraft on A100 without cuequivariance kernels Protocol: TorchCraft runtime per design at 500 residues Dataset: TorchCraft runtime benchmark design runs 33.3 Runtime (wall-clock minutes per design at 500 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 500 residues, GPU w/o cuequivariance
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "GPU w/o cuequivariance", column "500 aa" Configuration: TorchCraft on Ascend 910B with fused attention Protocol: TorchCraft runtime per design at 100 residues Dataset: TorchCraft runtime benchmark design runs 3.1 Runtime (wall-clock minutes per design at 100 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 100 residues, NPU w/ FA
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "NPU w/ FA", column "100 aa" Configuration: TorchCraft on Ascend 910B with fused attention Protocol: TorchCraft runtime per design at 200 residues Dataset: TorchCraft runtime benchmark design runs 3.75 Runtime (wall-clock minutes per design at 200 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 200 residues, NPU w/ FA
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "NPU w/ FA", column "200 aa" Configuration: TorchCraft on Ascend 910B with fused attention Protocol: TorchCraft runtime per design at 300 residues Dataset: TorchCraft runtime benchmark design runs 6.23 Runtime (wall-clock minutes per design at 300 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 300 residues, NPU w/ FA
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "NPU w/ FA", column "300 aa" Configuration: TorchCraft on Ascend 910B with fused attention Protocol: TorchCraft runtime per design at 400 residues Dataset: TorchCraft runtime benchmark design runs 11.3 Runtime (wall-clock minutes per design at 400 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 400 residues, NPU w/ FA
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "NPU w/ FA", column "400 aa" Configuration: TorchCraft on Ascend 910B with fused attention Protocol: TorchCraft runtime per design at 500 residues Dataset: TorchCraft runtime benchmark design runs 16.6 Runtime (wall-clock minutes per design at 500 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 500 residues, NPU w/ FA
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "NPU w/ FA", column "500 aa" Configuration: TorchCraft on Ascend 910B without fused attention Protocol: TorchCraft runtime per design at 100 residues Dataset: TorchCraft runtime benchmark design runs 3 Runtime (wall-clock minutes per design at 100 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 100 residues, NPU w/o FA
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "NPU w/o FA", column "100 aa" Configuration: TorchCraft on Ascend 910B without fused attention Protocol: TorchCraft runtime per design at 200 residues Dataset: TorchCraft runtime benchmark design runs 4.13 Runtime (wall-clock minutes per design at 200 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 200 residues, NPU w/o FA
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "NPU w/o FA", column "200 aa" Configuration: TorchCraft on Ascend 910B without fused attention Protocol: TorchCraft runtime per design at 300 residues Dataset: TorchCraft runtime benchmark design runs 8.63 Runtime (wall-clock minutes per design at 300 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 300 residues, NPU w/o FA
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "NPU w/o FA", column "300 aa" Configuration: TorchCraft on Ascend 910B without fused attention Protocol: TorchCraft runtime per design at 400 residues Dataset: TorchCraft runtime benchmark design runs 14.6 Runtime (wall-clock minutes per design at 400 residues, batch size 2 run divided by 2)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 400 residues, NPU w/o FA
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "NPU w/o FA", column "400 aa" Configuration: TorchCraft on Ascend 910B without fused attention Protocol: TorchCraft runtime per design at 500 residues Dataset: TorchCraft runtime benchmark design runs 33.5 Runtime (wall-clock minutes per design at 500 residues, measured at batch size 1 because batch size 2 exceeded memory, not the batch-size-2 run divided by 2 that every other cell of the table reports)minute · lower
Uncertainty: Not reported by the source: No spread is printed; GPU cells are stated to be the mean of three runs without a spread.
Coverage: Not reported scored / Not reported eligible
Author-reported evaluation · Source checked Evidence concern: excluded from comparisons The site's closing line says the GPU and NPU implementations, the TorchFold training code and the TorchFold-AbAg model weights are open under the MIT license. The repository's own LICENSE file and README at commit ebf9dc6676ecb1be505bc72f87de408995f0d8fd say Apache License 2.0, and the GitHub repository metadata reports apache-2.0. The gated Hugging Face dataset is tagged mit but conditions access on no redistribution, rehosting, selling or sharing without written permission. The MIT statement is therefore not relied on.TorchX project site (torchx-cpl.github.io) · Project site, closing line below the citations block
Methods, coverage and source TorchCraft runtime at 500 residues, NPU w/o FA
Not reported
Aggregation: Not reported
TorchCraft: Unified binder design by inverting an all-atom structure predictor (arXiv:2609.19770v1) ; TorchX project site (torchx-cpl.github.io) · Table S9, row "NPU w/o FA", column "500 aa"