OpenMOSS-Team/SWE-bench-Science
SWE-bench-Science: Harbor dataset on Hugging Face with 119 tasks. SWE-bench Science evaluates coding agents on software-engineering tasks drawn from scientific-computing repositories. The release contains 119 tasks across 20 scientific domains, with isolated environments and separate programmatic…
Tasks
- Repair an inconsistent transition-state rotor workflow
- Repair an inconsistent reduced validation workflow
- Repair extxyz stress and virial tensor-label preservation
- Repair split-read exon-overlap filtering
- Repair edited MRS subspectrum polarity handling
- Repair inconsistent genomic interval enrichment counts
- Repair unstable spectral registration for drifted MRS transients
- Establish first vacuum-plus-wall external-region capability in the TERPSICHORE-native path
- Enable projection-wall external-region support in the TERPSICHORE-native path
- Complete alternate perturbation-representation support
- Complete the Missing Response Capability
- Complete 3-D Cross-n FOURIN Spectral Coupling
- Establish boundary-regime operator and energy closure in the TERPSICHORE-native path
- Complete the native multi-equilibrium TERPSICHORE workflow
- Restore thermal-equilibrium invariants in a Liouville-space relaxation workflow
- Repair HMMRATAC peak refinement under sparse signal resolution
- Repair dark-matter line-of-sight maps in Gammapy
- Restore blacklist semantics in an HMMRATAC chromatin-accessibility workflow
- Complete robust topic-to-cistrome binarization for independent probability matrices
- Complete an edge-graph geometry and force/virial capability for atomistic models
- Complete a missing atomic-liquid thermodynamic-integration capability
- Repair an inconsistent volume-to-surface projection
- Restore continuous NPT periodic-boundary unwrapping
- Restore coherent inference in a centered factorial-effects model
- Repair inconsistent cross-protocol Osprey validation reports
- Restore isentropic interpolation for terrain-following model data
- Restore consistent NMR peak-set encoding under batch padding
- Repair control-referenced relative responses in a single-cell perturbation workflow
- Restore annotation equivalence after circular-record orientation normalization
- Repair format-conformant SHELX/AIRSS .res export in a crystallographic IO workflow
- Restore documented average-reference semantics for mixed-modality recordings
- Restore equivalent GROMACS nonbonded representations
- Repair inconsistent energies across equivalent periodic representations
- Repair an inconsistent stochastic trajectory calculation
- Repair an incomplete stiff-kinetics campaign
- Repair inconsistent sampling on a spherical curvilinear ocean grid
- Repair an inconsistent multiframe MR reconstruction
- Repair inconsistent crystal classification across equivalent representations
- Repair a river-network sediment transport calculation
- Restore consistent magnetic-dipole electromagnetic predictions
- Repair an inconsistent near-Earth vector conversion
- Repair an inconsistent solar-coordinate archive workflow
- Repair an inconsistent porous-volume characterization
- Repair inconsistent lithium-ion discharge predictions
- Repair inconsistent finite-volume operations on a connected multi-face grid
- Repair an inconsistent X-ray spectral-timing analysis
- Repair an inconsistent broadband microwave composite
- Repair an inconsistent alloy equilibrium assessment
- Restore consistent stochastic-field behavior on nonuniform grids
- Repair an inconsistent tensor-network state calculation
- Repair an inconsistent planetary gradient-tensor calculation
- Restore a nonuniform finite-volume phase-field workflow
- Repair a representation-dependent physical surface Dice workflow
- Repair an inconsistent tandem-mass spectral similarity workflow
- Restore a scientific template-comparison workflow
- Repair a physically inconsistent orbital history export
- Investigate a periodic trajectory-processing anomaly
- Repair a valid OpenMC fixed-source workflow
- Restore physical consistency in calibrated cell morphometry
- Restore a complete cross-band drift-correction recording
- Repair inconsistent celestial image representation in a mosaic
- Repair an inconsistent archived elasticity workflow
- Repair inconsistent population-history estimates after archive curation
- Repair inconsistent molecular-energy interpretation in a quantum-chemistry log parser
- Repair inconsistent particle-field visualization results
- Restore regional climate-model validation consistency
- Repair a seasonal fire-weather state transition
- Restore biologically consistent retrieval from a VCF archive
- Restore a complete Hi-C contact-decay analysis
- Restore chemically consistent proteoform ion annotations
- Repair an orientation-sensitive fiber-scattering reduction
- Repair an incomplete DICOM color image presentation workflow
- Repair inconsistent alternate-conformer structure connectivity
- Repair an ill-conditioned overlap band-postprocessing workflow
- Restore consistent compressed Deep Potential tabulation
- Repair a batched parametric tracking workflow
- Repair an inconsistent oriented-envelope fallback workflow
- Repair an inconsistent astronomical aperture workflow
- Repair molecular-string translation for catalogue workflows
- Repair inconsistent geospatial raster remapping
- Repair local study workflow
- Repair a workflow consistency bug
- Repair bond-order species reporting
- Repair ABACUS relaxation archive/reload semantics
- Inconsistent archived observables
- Repair a silent MatGL fine-tuning incompatibility
- Repair a silent diffusion-MRI denoising inconsistency
- Repair a boundary aggregation failure
- Repair inconsistent ORCA excited-state oscillator strengths
- Restore multidimensional NMR acquisition grids
- Repair a workflow regression
- Repair inconsistent Chemprop reaction observations
- Repair a patient-space round trip in MONAI
- Repair magnetic-space-group operations across archived settings
- Repair a repeated porous-electrode transport workflow
- Repair a wrapped tropical-cyclone hazard workflow
- Repair spin-model consistency in DeePMD-kit
- Repair inconsistent satellite parallax geolocation
- Repair an inconsistent GNSS point solution
- Repair calibrated framework embedding relaxation
- Repair an inconsistent battery-model reproduction
- Repair an inconsistent polymer coordinate workflow
- Repair symmetric composite laminate ABD calculations
- Repair inconsistent finite-volume advection edge states
- Restore a stereo-preserving tautomer workflow
- Reproduce and extend a version-pinned MAP4/MAPC retrieval protocol
- Repair a public symbolic workflow
- Restore stable Jacobi theta evaluations across difficult regimes
- Repair paired wrinkle FE analysis
- Restore consistent cross-section analysis results
- Restore consistent tapered-filament free response
- Repair an inconsistent hybrid-section plastic capacity workflow
- Repair an inconsistent solar-resource workflow
- Repair an implausible storage history at a planning-epoch boundary
- Restore physically meaningful structural stability checks
- Repair mode-dependent structural analysis observations
- Repair inconsistent lateral-frame member observations
- Repair inconsistent periodic molecular energy representations
- Repair a molecular curvature workflow for reaction-path analysis