task_id stringlengths 17 84 | title stringlengths 5 62 ⌀ | summary stringlengths 76 372 | category stringclasses 14
values | subdomain stringlengths 6 53 | task_split stringclasses 3
values | task_prompt stringlengths 439 5.84k | agent_must_do listlengths 0 9 | software listlengths 0 9 | input_files listlengths 0 14 | taxonomy dict | source_repo_path stringlengths 23 90 |
|---|---|---|---|---|---|---|---|---|---|---|---|
life_sciences/amber_three_stage_mmgbsa_workflow_instance_1 | Amber Three-Stage MMGBSA Workflow Authoring | Author a Linux Amber workflow bundle for a three-chain protein complex, including the two SLURM driver scripts, the MMGBSA submission script, and the final MMGBSA result file based on a staged trajectory. | life_sciences | Biomolecular Structure & Design | near-term | You are preparing a Linux Amber workflow definition for a protein-only three-chain complex.
Task directory:
- `base`
Input directory:
- `base/input`
Software:
- Benchmark-owned AmberTools 23 CLI wrapper: `base/software/run_ambertools.sh`
- Invoke staged AmberTools commands as: `base/software/run_ambertools.sh <tool>... | [
"Inspect `complex_structure.pdb` and the two staged markdown specs under `input/`.",
"Treat chain `A` as the receptor and chains `B` plus `C` together as the ligand.",
"Write `submit_min.sh` that covers topology construction plus minimization and short equilibration.",
"Write `submit_prod.sh` for one implicit... | [
"AmberTools 23 via software/run_ambertools.sh",
"Amber 22 / SLURM / CUDA 11.6.2 (reference environment only)"
] | [
{
"name": "task_sop.md",
"format": "Markdown",
"path": "input/task_sop.md",
"description": "Agent-facing Amber workflow instructions and deliverable expectations."
},
{
"name": "input_environment_spec.md",
"format": "Markdown",
"path": "input/input_environment_spec.md",
"descript... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.1",
"subdomain_code": "biomolecular",
"subdomain_name": "Biomolecular Structure & Design"
} | tasks/life_sciences/amber_three_stage_mmgbsa_workflow_instance_1 |
life_sciences/cell_tracking_instance_1 | Cell Tracking Instance 1 | Segment and track cells across a 30-frame fluorescence microscopy sequence and export Cell Tracking Challenge style masks and lineage metadata. | life_sciences | Cell & Imaging Biology | near-term | You are performing cell tracking on a fluorescence microscopy time-lapse sequence.
Task directory:
- `base`
Visible inputs:
- 30 grayscale TIFF frames: `base/input/01/t000.tif` through `base/input/01/t029.tif`
- Optional Python dependency manifest: `base/input/runtime_env/pyproject.toml`
- Optional Python dependency ... | [
"Inspect the 30 staged fluorescence microscopy frames.",
"Segment individual cells in each frame.",
"Track cells over time with consistent positive integer labels.",
"Write exactly 30 labeled TIFF masks named `mask000.tif` through `mask029.tif`.",
"Write a valid CTC-style `res_track.txt` lineage table."
] | [
"Python",
"NumPy",
"Pillow",
"tifffile"
] | [
{
"name": "t000.tif ... t029.tif",
"format": "TIFF image sequence",
"path": "input/01/",
"description": "Thirty grayscale 1024 by 1024 fluorescence microscopy frames."
},
{
"name": "pyproject.toml",
"format": "Python project manifest",
"path": "input/runtime_env/pyproject.toml",
... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.3",
"subdomain_code": "cell_imaging",
"subdomain_name": "Cell & Imaging Biology"
} | tasks/life_sciences/cell_tracking_instance_1 |
life_sciences/cell_translocation_analysis | Cell Translocation Analysis | Analyze paired DNA and GFP microscopy images to quantify nuclear translocation across a drug dose series and report the minimum effective dose. | life_sciences | Cell & Imaging Biology | full-spectrum | You are analyzing two-channel cell microscopy images for GFP nuclear translocation.
Task directory:
- `base`
Visible inputs:
- Paired TIFF images: `base/input/images`
- Dose and control metadata: `base/input/images/Translocation_doses_and_controls.csv`
- Detailed task instructions: `base/input/task_instructions.md`
-... | [
"Pair DNA (`w1`) and GFP (`w2`) TIFF images by well.",
"Segment nuclei and derive cell/cytoplasm compartments.",
"Measure GFP intensity, channel-correlation, location, and ratio-style object features.",
"Classify positive translocation using the dose/control metadata.",
"Save `Cells.csv`, `Cytoplasm.csv`, `... | [
"CellProfiler",
"Python",
"numpy",
"pandas",
"scikit-image",
"tifffile"
] | [
{
"name": "BBBC013_<well>_s1_w1.tif / BBBC013_<well>_s1_w2.tif",
"format": "tiff",
"path": "input/images/",
"description": "Paired DNA and GFP microscopy channels for each staged well."
},
{
"name": "Translocation_doses_and_controls.csv",
"format": "csv",
"path": "input/images/Transl... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.3",
"subdomain_code": "cell_imaging",
"subdomain_name": "Cell & Imaging Biology"
} | tasks/life_sciences/cell_translocation_analysis |
life_sciences/gene_expression_differential_analysis_functional_enrichment_analysis_1 | BRCA Differential Expression And KEGG Enrichment Analysis | Analyze a staged BRCA count matrix and metadata on Ubuntu, run pydeseq2 differential expression for tumor versus normal, and generate paired KEGG enrichment outputs. | life_sciences | Genomics & Sequence Analysis | near-term | You are a bioinformatics analyst working on a Linux VM.
Task directory:
- `base`
Visible input files:
- Count matrix: `base/input/BRCA_selected_samples_counts.tsv`
- Sample metadata: `base/input/BRCA_selected_samples_metadata.tsv`
- Analysis spec: `base/input/analysis_spec.json`
- Output contract: `base/input/output_... | [
"Read the staged BRCA count matrix, metadata, analysis spec, and output contract.",
"Install the required Python packages into an agent-owned environment from the staged manifest.",
"Run pydeseq2 with the explicit design formula `~ batch + condition` to compare `tumor` against `normal`.",
"Populate the DEG `g... | [
"Python",
"pydeseq2",
"gseapy"
] | [
{
"name": "BRCA_selected_samples_counts.tsv",
"format": "TSV",
"path": "input/BRCA_selected_samples_counts.tsv",
"description": "BRCA count matrix with leading `gene_id` column."
},
{
"name": "BRCA_selected_samples_metadata.tsv",
"format": "TSV",
"path": "input/BRCA_selected_samples_... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.2",
"subdomain_code": "genomics",
"subdomain_name": "Genomics & Sequence Analysis"
} | tasks/life_sciences/gene_expression_differential_analysis_functional_enrichment_analysis_1 |
life_sciences/genomic_interval_processing_1 | ENCODE CTCF Union Peak Intervals | Process three ENCODE CTCF narrowPeak BED files into the requested non-overlapping union peak set, with a reproducible command log and summary counts. | life_sciences | Genomics & Sequence Analysis | full-spectrum | You are a bioinformatics analyst working on a Linux genomic interval processing task.
Task directory:
- `base`
Visible input files:
- `base/input/ENCFF483KVM.bed`
- `base/input/ENCFF511NNV.bed`
- `base/input/ENCFF758CQW.bed`
- Operation specification: `base/input/operation_specification.txt`
- Detailed task instructi... | [
"Read `operation_specification.txt` and apply it to all three ENCODE CTCF narrowPeak BED inputs.",
"Write a sorted, non-overlapping 3-column BED file named `union_peaks.bed`.",
"Record the interval-processing commands or script steps in `commands.sh`.",
"Write `summary.json` with input interval counts, total ... | [
"BEDTools",
"GNU sort",
"Python"
] | [
{
"name": "ENCFF483KVM.bed",
"format": "BED/narrowPeak",
"path": "input/ENCFF483KVM.bed",
"description": "ENCODE CTCF narrowPeak-style interval file."
},
{
"name": "ENCFF511NNV.bed",
"format": "BED/narrowPeak",
"path": "input/ENCFF511NNV.bed",
"description": "ENCODE CTCF narrowPe... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.2",
"subdomain_code": "genomics",
"subdomain_name": "Genomics & Sequence Analysis"
} | tasks/life_sciences/genomic_interval_processing_1 |
life_sciences/hg002_chr22_germline_variant_pipeline | HG002 Chr22 Germline Variant Pipeline | Repair a staged chr22 germline variant-calling workflow for HG002 and produce the required filtered variants, annotation outputs, and QC artifacts. | life_sciences | Genomics & Sequence Analysis | near-term | You are working on a Linux germline-variant-calling task.
Visible solve-time input root:
- `base/input/starter_project`
Your job is to repair the starter workflow and write a completed `submission/`
tree under:
- `base/output/submission`
Inside the visible starter project you have:
- paired-end reads under `fastq/`
... | [
"Read the staged HG002 chr22 starter workflow under `input/starter_project/`.",
"Repair the broken samplesheet, Nextflow config, known-sites naming, and `run_vep.sh` assumptions.",
"Rebuild the missing `bwa-mem2` index and run the chr22 calling workflow.",
"Write the required filtered VCF, annotated VCF, QC o... | [
"Nextflow",
"nf-core/sarek",
"BWA-MEM2",
"GATK",
"VEP",
"MultiQC"
] | [
{
"name": "starter_project/fastq/",
"format": "gzipped FASTQ pair",
"path": "input/starter_project/fastq/",
"description": "Paired-end HG002 chr22 reads for the germline calling run."
},
{
"name": "starter_project/reference/",
"format": "FASTA, BED, and bgzipped VCF inputs",
"path": ... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.2",
"subdomain_code": "genomics",
"subdomain_name": "Genomics & Sequence Analysis"
} | tasks/life_sciences/hg002_chr22_germline_variant_pipeline |
life_sciences/idp_ensemble_scoring | IDP Ensemble Scoring with Given Open-Source Tools | Rank 5 IDP ensemble generation models by how well their ensembles match experimental NMR data (chemical shifts, J-couplings, NOE/PRE) using UCBShift and X-EISD. | life_sciences | Biomolecular Structure & Design | last-exam | You are a computational structural biologist. Your task is to rank 5 IDP (intrinsically disordered protein) ensemble generation models by how well their ensembles match experimental NMR data.
## Your Task
Use the tools provided locally to back-calculate experimental observables from protein ensemble conformers, score... | [
"Set up a Python environment with correct dependencies (biopython==1.74, scikit-learn==0.22)",
"Back-calculate CS via UCBShift and JC/NOE/PRE via xeisd calculator",
"Score ensembles using xeisd.optimizer.XEISD.calc_scores()",
"Min-max normalize per observable and compute total ranking",
"Save Final_Output.c... | [
"UCBShift (CSpred)",
"X-EISD (xeisd)"
] | [
{
"name": "CSpred",
"format": "directory",
"path": "input/CSpred/",
"description": "UCBShift chemical shift predictor with models and binaries"
},
{
"name": "xeisd",
"format": "directory",
"path": "input/xeisd/",
"description": "X-EISD ensemble scoring module"
},
{
"name"... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.1",
"subdomain_code": "biomolecular",
"subdomain_name": "Biomolecular Structure & Design"
} | tasks/life_sciences/idp_ensemble_scoring |
life_sciences/merfish_image_decoding_segmentation_1 | Single-FOV MERFISH Image Decoding And Cell Segmentation | Run the full MERFISH image-analysis pipeline on one field of view of U2OS cells: decode transcripts against a 130-gene MHD4 codebook, segment nuclei from the DAPI tile, assign transcripts to cells, and emit a cell-by-gene count matrix plus quality metrics. | life_sciences | Cell & Imaging Biology | near-term | You are working on an Ubuntu VM to run the full MERFISH image-analysis pipeline
for one field of view of a human U2OS cell population stained against a
130-gene panel.
Task directory:
- `base`
Inputs (all under `base/input`; treat as read-only):
- `experiment.json` plus `primary_images.json`, `nuclei.json` and the pe... | [
"Invoke the canonical Python entry point `software/merfish_runtime.sh` (or materialize the env with `uv sync --frozen --project input/runtime_env/`).",
"Load the staged starfish experiment (16 primary TIFFs + 1 DAPI TIFF + manifests + codebook) and sanity-check alignment.",
"Preprocess images: high-pass / backg... | [
"Python",
"uv",
"starfish",
"Cellpose",
"PyTorch"
] | [
{
"name": "Starfish experiment manifest",
"format": "json",
"path": "input/experiment.json",
"description": "Top-level starfish Experiment v5.0.0 manifest with per-(round, channel) `scale_factors`."
},
{
"name": "MHD4 codebook",
"format": "json",
"path": "input/codebook.json",
"d... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.3",
"subdomain_code": "cell_imaging",
"subdomain_name": "Cell & Imaging Biology"
} | tasks/life_sciences/merfish_image_decoding_segmentation_1 |
life_sciences/protein_function_annotation_instance_1 | Protein Function Annotation With InterProScan | Annotate one staged yeast protein with InterProScan, export the required InterPro and GO TSV files, and write a short gamma-tubulin functional summary. | life_sciences | Genomics & Sequence Analysis | near-term | You are working on a Linux VM to annotate one staged protein with InterProScan.
## Visible Task Directory
- `base`
## Visible Inputs
- Protein FASTA: `base/input/protein_sequence.fasta`
- Organism name: `base/input/organism_name.txt`
- InterProScan install script: `base/software/install_software.sh`
- InterProScan wr... | [
"Run the install script to install InterProScan 5.77-108.0 before scanning.",
"Run the staged InterProScan wrapper on the visible FASTA input.",
"Write `interpro_domains.tsv` with the exact required columns, one retained row per non-empty InterPro accession, merged spans for repeated accessions, and 2-decimal `... | [
"InterProScan",
"Python",
"bash"
] | [
{
"name": "protein_sequence.fasta",
"format": "FASTA",
"path": "input/protein_sequence.fasta",
"description": "Single staged query protein sequence to annotate"
},
{
"name": "organism_name.txt",
"format": "TXT",
"path": "input/organism_name.txt",
"description": "Visible organism ... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.2",
"subdomain_code": "genomics",
"subdomain_name": "Genomics & Sequence Analysis"
} | tasks/life_sciences/protein_function_annotation_instance_1 |
life_sciences/pseudotime_de | Pseudotime Differential Expression (Palantir + tradeSeq) | Discover genes with different expression patterns along pseudotime in a bone-marrow scRNA-seq dataset using Palantir for pseudotime estimation and tradeSeq for lineage-specific differential expression testing. | life_sciences | Genomics & Sequence Analysis | last-exam | You are a computational biologist performing pseudotime differential expression analysis on a bone-marrow scRNA-seq dataset on Linux.
Task directory:
- `base`
Input files:
- scRNA-seq counts: `base/input/marrow_sample_scseq_counts.h5ad`
- Detailed task brief: `base/input/task_description.txt`
- Python dependency mani... | [
"Install Python analysis packages (scanpy, palantir, anndata, etc.) and R Bioconductor packages (SingleCellExperiment, tradeSeq).",
"Run Palantir pseudotime estimation with the specified start cell and terminal states.",
"Export Python outputs to disk and reassemble as a SingleCellExperiment in R.",
"Run trad... | [
"Python",
"R",
"scanpy",
"palantir",
"anndata",
"tradeSeq",
"SingleCellExperiment"
] | [
{
"name": "marrow_sample_scseq_counts.h5ad",
"format": "h5ad",
"path": "input/marrow_sample_scseq_counts.h5ad",
"description": "Bone-marrow scRNA-seq raw counts (4142 cells x 16106 genes)."
},
{
"name": "task_description.txt",
"format": "txt",
"path": "input/task_description.txt",
... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.2",
"subdomain_code": "genomics",
"subdomain_name": "Genomics & Sequence Analysis"
} | tasks/life_sciences/pseudotime_de |
life_sciences/rgi_mcr1_colistin_v2 | RGI Contig Resistance Annotation | Annotate one staged antimicrobial-resistance contig with CARD RGI and report the best-hit ARO gene name, percent identity, drug class, and resistance mechanism in a JSON output. | life_sciences | Systems & Microbial Biology | full-spectrum | You are working on a Linux VM to annotate one plasmid-mediated antimicrobial-resistance gene from a staged DNA contig with CARD RGI.
Visible task workspace:
- `base`
Visible inputs:
- Contig FASTA: `base/input/input_contig.fasta`
- CARD database JSON: `base/input/card.json`
- Runtime manifest for installing the offic... | [
"Create a task-local Python environment from `input/runtime_env/pyproject.toml` so the official GitHub `rgi` package is available.",
"Load the staged CARD JSON with `rgi load --local -i card.json`.",
"Run `rgi main` in contig mode against `input_contig.fasta` with `--local --clean -g PYRODIGAL`.",
"Parse the ... | [
"Python",
"uv",
"RGI",
"CARD",
"NCBI BLAST+"
] | [
{
"name": "input_contig.fasta",
"format": "FASTA",
"path": "input/input_contig.fasta",
"description": "Staged DNA contig for resistance annotation."
},
{
"name": "card.json",
"format": "JSON",
"path": "input/card.json",
"description": "CARD detection-model payload used by `rgi lo... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.4",
"subdomain_code": "systems_microbial",
"subdomain_name": "Systems & Microbial Biology"
} | tasks/life_sciences/rgi_mcr1_colistin_v2 |
life_sciences/spatial_transcriptomics_spatial_domain_identification | Spatial Transcriptomics Spatial Domain Identification | Cluster 12 human DLPFC 10x Visium slices into spatial domains and submit per-spot labels, summary metrics, manifest metadata, and one UMAP overlay. | life_sciences | Cell & Imaging Biology | near-term | You are given 12 human DLPFC 10x Visium slices and must identify spatial domains on every slice.
Task directory:
- `base`
Visible inputs:
- Slice data directories: `base/input/data/<slice_id>/`
- Per-slice target cluster counts: `base/input/data/slice_config.csv`
- Optional Python runtime manifest: `base/input/runtim... | [
"Load the staged matrix, features, barcodes, and spatial-coordinate files for all 12 Visium slices.",
"Use `input/data/slice_config.csv` as the exact target number of clusters per slice.",
"Run an unsupervised spatial-domain identification workflow.",
"Write one `per_slice/<slice_id>_labels.csv` file per slic... | [
"Python",
"uv",
"Scanpy",
"Squidpy",
"AnnData",
"scikit-learn"
] | [
{
"name": "matrix.mtx / features.tsv / barcodes.tsv",
"format": "MTX/TSV",
"path": "input/data/<slice_id>/",
"description": "Per-slice Visium count matrices and feature/barcode tables."
},
{
"name": "spatial",
"format": "CSV/JSON",
"path": "input/data/<slice_id>/spatial/",
"descr... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.3",
"subdomain_code": "cell_imaging",
"subdomain_name": "Cell & Imaging Biology"
} | tasks/life_sciences/spatial_transcriptomics_spatial_domain_identification |
life_sciences/tcga_brca_deg_analysis | TCGA BRCA Differential Expression Analysis | Analyze TCGA Breast Cancer expression and clinical matrices to identify genes differentially expressed between primary tumor and solid tissue normal samples. | life_sciences | Genomics & Sequence Analysis | near-term | You are a bioinformatics analyst performing a TCGA Breast Cancer differential gene expression analysis on Linux.
Task directory:
- `base`
Visible input files:
- Expression matrix: `base/input/expression_matrix.tsv.gz`
- Clinical annotations: `base/input/clinical_matrix.tsv`
- Benchmark gene list: `base/input/truth_br... | [
"Match expression-matrix sample columns to clinical metadata rows.",
"Compare `Primary Tumor` samples against `Solid Tissue Normal` samples.",
"Filter genes with more than 80% zero expression across matched samples.",
"Run Welch's t-test for each retained gene.",
"Compute `log2FC = mean(tumor) - mean(normal... | [
"Python",
"pandas",
"scipy",
"statsmodels",
"matplotlib",
"scikit-learn"
] | [
{
"name": "expression_matrix.tsv.gz",
"format": "TSV.GZ",
"path": "input/expression_matrix.tsv.gz",
"description": "Gene-by-sample TCGA BRCA expression matrix."
},
{
"name": "clinical_matrix.tsv",
"format": "TSV",
"path": "input/clinical_matrix.tsv",
"description": "Clinical samp... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.2",
"subdomain_code": "genomics",
"subdomain_name": "Genomics & Sequence Analysis"
} | tasks/life_sciences/tcga_brca_deg_analysis |
life_sciences/tms_marrow_cell_type_annotation_instance_1 | Marrow Cell Type Annotation | Annotate every cell in a staged Smart-seq2 mouse bone marrow AnnData object using a fixed 21-label ontology and return one predicted label per cell. | life_sciences | Cell & Imaging Biology | near-term | You are given an unlabeled Smart-seq2 mouse bone marrow single-cell dataset and must annotate every cell with one label from a fixed 21-class ontology.
Task directory:
- `base`
Visible inputs:
- AnnData matrix: `base/input/tms_marrow_unlabeled.h5ad`
- Allowed labels: `base/input/allowed_labels.txt`
- Agent-facing run... | [
"Load the staged Smart-seq2 marrow AnnData object on Linux and, if needed, install the staged Python runtime from `input/runtime_env/` with `uv`.",
"Run a biologically reasonable batch-aware single-cell annotation workflow across the fixed 21-label ontology.",
"Write exactly one UTF-8 CSV to `output/predictions... | [
"Python",
"scanpy",
"anndata",
"harmonypy"
] | [
{
"name": "tms_marrow_unlabeled.h5ad",
"format": "h5ad",
"path": "input/tms_marrow_unlabeled.h5ad",
"description": "Unlabeled Smart-seq2 mouse bone marrow AnnData object with 14,517 cells."
},
{
"name": "allowed_labels.txt",
"format": "txt",
"path": "input/allowed_labels.txt",
"d... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.3",
"subdomain_code": "cell_imaging",
"subdomain_name": "Cell & Imaging Biology"
} | tasks/life_sciences/tms_marrow_cell_type_annotation_instance_1 |
life_sciences/tp53_locus_variant_histone_browser_svg | K562 Genome Browser SVG Export | Create an hg19 genome-browser SVG for a K562 regulatory locus using a structural-variant VCF and an H3K27ac BigWig signal track. | life_sciences | Genomics & Sequence Analysis | near-term | You are a bioinformatics analyst working on a K562 regulatory genomics visualization task.
Task directory:
- `base`
Goal:
- Create a genome-browser SVG view for the hg19 TP53 locus: `chr17:7571651-7590910`.
- The view must include both the provided K562 structural variant VCF track and the K562 H3K27ac BigWig signal ... | [
"Open the specified hg19 locus for the assigned variant.",
"Load or display the provided VCF and H3K27ac BigWig tracks together.",
"Export or create a valid SVG with locus labels, both tracks, and browser/build provenance.",
"Save the final artifact as output/output.svg and use no alternate filename."
] | [
"UCSC Genome Browser",
"Python"
] | [
{
"name": "ENCFF960SSF.vcf.gz",
"format": "VCF.GZ",
"path": "input/ENCFF960SSF.vcf.gz",
"description": "K562 structural variant calls to show as the variant track."
},
{
"name": "wgEncodeBroadHistoneK562H3k27acStdSig.bigWig",
"format": "BigWig",
"path": "input/wgEncodeBroadHistoneK56... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.2",
"subdomain_code": "genomics",
"subdomain_name": "Genomics & Sequence Analysis"
} | tasks/life_sciences/tp53_locus_variant_histone_browser_svg |
life_sciences/yeast_colony_detection | Yeast Colony Detection | Detect red yeast colonies in a masked agar-plate image and export a colony count plus centroid measurement table. | life_sciences | Cell & Imaging Biology | full-spectrum | You are analyzing a yeast colony plate image on Linux.
## Your Task
Detect the red yeast colonies growing on the agar plate while excluding white dots and other visual noise.
## Visible Inputs
- Plate image: `base/input/6-1.jpg`
- Plate-region mask: `base/input/PlateTemplate.png`
- CellProfiler entry point, if provis... | [
"Use the plate image and mask to identify red yeast colonies inside the plate region.",
"Exclude white dots and unrelated visual noise from the colony detections.",
"Create a reproducible CellProfiler or image-analysis workflow on the Linux VM.",
"Write `answer.json` with an integer `colony_count`.",
"Write... | [
"CellProfiler 4.2.8",
"Python",
"NumPy"
] | [
{
"name": "6-1.jpg",
"format": "JPEG",
"path": "input/6-1.jpg",
"description": "Yeast agar plate image."
},
{
"name": "PlateTemplate.png",
"format": "PNG",
"path": "input/PlateTemplate.png",
"description": "Aligned plate-region mask."
}
] | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.3",
"subdomain_code": "cell_imaging",
"subdomain_name": "Cell & Imaging Biology"
} | tasks/life_sciences/yeast_colony_detection |
life_sciences/zdock_hiv_dimer_interface_scoring_v1 | ZDOCK HIV Dimer Interface Scoring | Evaluate ten precomputed HIV protease dimer docking poses by computing fixed interface overlap, Fnat, IRMSD, and final ranking metrics. | life_sciences | Biomolecular Structure & Design | near-term | You are evaluating precomputed HIV protease dimer docking predictions on a Linux VM.
## Input Files
- Native complex: `base/input/1HVR.pdb`
- Chain A structure: `base/input/1HVR_chainA.pdb`
- Chain B structure: `base/input/1HVR_chainB.pdb`
- Docking-pose archive: `base/input/top_preds.tar.gz`
## Your Task
1. Use `bas... | [
"Use protein `ATOM` records from `input/1HVR.pdb` to derive native Chain A / Chain B heavy-atom interface residues with a 5 Angstrom cutoff; exclude `HETATM` records.",
"Unpack `input/top_preds.tar.gz` and process `complex.1.pdb` through `complex.10.pdb`.",
"Compute overlap, Fnat, interface C-alpha IRMSD after ... | [
"Python",
"BioPython",
"NumPy"
] | [
{
"name": "1HVR.pdb",
"format": "pdb",
"path": "input/1HVR.pdb",
"description": "Native Chain A / Chain B complex used to derive the native interface and contacts."
},
{
"name": "1HVR_chainA.pdb",
"format": "pdb",
"path": "input/1HVR_chainA.pdb",
"description": "Submitted Chain A... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.1",
"subdomain_code": "biomolecular",
"subdomain_name": "Biomolecular Structure & Design"
} | tasks/life_sciences/zdock_hiv_dimer_interface_scoring_v1 |
other/aerobics_wc2026_portugal_trio_difficulty_scoring | Aerobics World Cup Trio Difficulty Scoring | Inspect the staged aerobics trio routine video and FIG rulebook, then produce the required chronological difficulty-element spreadsheet for the routine. | other | Sports | last-exam | You are working on Ubuntu.
## Your Task
Inspect the routine video and the FIG Aerobic Gymnastics Code of Points, then
produce a difficulty-element spreadsheet for the routine.
## Visible Inputs
- Video: `variant_1/input/trio.mov`
- Rulebook PDF: `variant_1/input/FIG Aerobic Gymnastics Code of Points (2025-2028).pdf`
... | [
"Open the staged routine video at `input/trio.mov`.",
"Read the staged FIG Code of Points PDF at `input/FIG Aerobic Gymnastics Code of Points (2025-2028).pdf`.",
"Identify each credited difficulty element in chronological order.",
"Write exactly one workbook to `output/difficulty_element_log.xlsx` with column... | [
"VLC",
"LibreOffice",
"Python",
"xdg-open"
] | [
{
"name": "trio.mov",
"format": "video",
"path": "input/trio.mov",
"description": "Staged competition routine video for the men's trio performance"
},
{
"name": "FIG Aerobic Gymnastics Code of Points (2025-2028).pdf",
"format": "PDF",
"path": "input/FIG Aerobic Gymnastics Code of Poi... | {
"domain_id": "14",
"domain_code": "other",
"subdomain_id": "14.1",
"subdomain_code": "sports",
"subdomain_name": "Sports"
} | tasks/other/aerobics_wc2026_portugal_trio_difficulty_scoring |
other/mota_exploration | Game Port Reference Capture: Magic Tower | Stand-in for the asset-capture step of a Flash-to-modern-engine port: drive the legacy Magic Tower (mota-24) build under Ruffle and collect per-floor reference screenshots that downstream engineers will use as ground-truth for tile layout, HUD placement, and level-script reconstruction. | other | 3D, Animation & Interactive Media | near-term | ## Context
A small game studio plans to re-implement the legacy Flash dungeon-crawler **Magic Tower (mota-24)** on a modern stack. Before any port code is written, the porting workflow requires a clean set of **per-floor reference screenshots** captured from the original build — these screenshots become the visual con... | [
"Focus the Ruffle window with mota-24.swf loaded (the task setup auto-launches it)",
"Wait for the title / loading screens and enter the actual game world",
"Play forward through floors 1 → 3 using normal movement and interaction (no cheats)",
"On arrival at each new floor, save a full-window screenshot as ou... | [
"Ruffle (Flash emulator)"
] | [
{
"name": "mota-24.swf",
"format": "SWF",
"path": "input/mota-24.swf",
"description": "Flash game to play"
}
] | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.1",
"subdomain_code": "animation_3d",
"subdomain_name": "3D, Animation & Interactive Media"
} | tasks/other/mota_exploration |
physical_sciences/adapt_vqe_molecular_energy | ADAPT-VQE Molecular Ground-State Energy | Use only NumPy and SciPy to solve three visible Jordan-Wigner molecular Hamiltonians with VQE / ADAPT-VQE style logic, then write a structured `results.json` with tier-wise energies and method metadata. | physical_sciences | Quantum Computing | near-term | You are solving a molecular ground-state energy benchmark using only NumPy and
SciPy. The visible inputs are a task statement and three Jordan-Wigner
Hamiltonians:
- `base/input/h2_hamiltonian.json`
- `base/input/lih_hamiltonian.json`
- `base/input/beh2_hamiltonian.json`
Read `base/input/problem_spec.md` first. It de... | [
"Read the staged prompt and three visible Hamiltonian JSON files under `input/`.",
"Prepare the staged NumPy / SciPy runtime manifest under `input/runtime_env/`.",
"Implement VQE / ADAPT-VQE style logic without any quantum-computing libraries.",
"Write `output/results.json` with tier-wise `molecule`, `energy_... | [
"Python",
"NumPy",
"SciPy"
] | [
{
"name": "Problem specification",
"format": "Markdown",
"path": "input/problem_spec.md",
"description": "Sanitized visible task statement and output schema."
},
{
"name": "Tier Hamiltonians",
"format": "JSON",
"path": "input/{h2_hamiltonian.json,lih_hamiltonian.json,beh2_hamiltonian... | {
"domain_id": "9",
"domain_code": "computing_math",
"subdomain_id": "9.7",
"subdomain_code": "quantum",
"subdomain_name": "Quantum Computing"
} | tasks/physical_sciences/adapt_vqe_molecular_energy |
physical_sciences/climate_prediction | CMIP6 Climate Emulation Pipeline | Build an end-to-end climate-emulation pipeline from a staged CMIP6-derived Zarr store and predict the held-out SSP245 test window for temperature and precipitation fields. | physical_sciences | Earth & Atmospheric Sciences | full-spectrum | You are a climate scientist building an end-to-end climate-emulation pipeline on Linux.
## Your Task
Build a climate-emulation pipeline from a staged CMIP6-derived Zarr store and predict the held-out SSP245 test window.
1. Bootstrap the Python runtime: `bash <software_dir>/bootstrap_runtime.sh`
2. Open the staged Za... | [
"1. Bootstrap the staged Python runtime via `software/bootstrap_runtime.sh`",
"2. Open the masked CMIP6-derived `data.zarr` with cftime-aware decoding",
"3. Build training tensors from `ssp126`, `ssp370`, `ssp585` and test inputs from last 120 months of `ssp245`",
"4. Use `member_id = 0` for targets; broadcas... | [
"Python",
"xarray",
"numpy",
"pandas",
"PyTorch"
] | [
{
"name": "data.zarr",
"format": "zarr",
"path": "input/data.zarr/",
"description": "Benchmark-safe CMIP6-derived Zarr store with held-out SSP245 test-window targets masked to NaN"
},
{
"name": "metadata.json",
"format": "json",
"path": "input/metadata.json",
"description": "Data... | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.4",
"subdomain_code": "earth_atmo",
"subdomain_name": "Earth & Atmospheric Sciences"
} | tasks/physical_sciences/climate_prediction |
physical_sciences/computational_materials_science | Silicon GW Band Gap | Compute the indirect GW quasiparticle band gap of bulk silicon using Quantum ESPRESSO and BerkeleyGW. | physical_sciences | Chemistry & Materials Computation | last-exam | You are a computational materials scientist working on a Linux VM.
## Your Task
Compute the indirect GW quasiparticle band gap of bulk silicon from the staged structure and pseudopotential.
## Input Files
- Silicon structure: `base/input/silicon/silicon.vasp`
- Silicon pseudopotential: `base/input/silicon/Si.UPF`
##... | [
"Read the staged silicon structure and pseudopotential under `input/silicon/`.",
"Create the missing QE and BerkeleyGW input decks from scratch for the SCF, NSCF/bands, pw2bgw, epsilon, sigma, and inteqp workflow.",
"Run the silicon GW workflow with approximately `5x5x5` wavefunction k-point sampling, a `10 Ry`... | [
"Python",
"Quantum ESPRESSO",
"BerkeleyGW"
] | [
{
"name": "Silicon structure",
"format": "vasp",
"path": "input/silicon/silicon.vasp",
"description": "Bulk silicon crystal structure"
},
{
"name": "Silicon pseudopotential",
"format": "UPF",
"path": "input/silicon/Si.UPF",
"description": "PZ/LDA norm-conserving silicon pseudopot... | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.2",
"subdomain_code": "chemistry_materials",
"subdomain_name": "Chemistry & Materials Computation"
} | tasks/physical_sciences/computational_materials_science |
physical_sciences/egt710_table1_smiles_extraction | Egt710 Table1 SMILES Extraction | Extract Table 1 from the staged paper, reconstruct the nine compounds, and deliver a CSV with validated SMILES and reported assay values. | physical_sciences | Chemistry & Materials Computation | full-spectrum | You are a medicinal chemistry researcher extracting a structure-activity table.
## Your Task
Use the uploaded paper PDF to reconstruct the nine compounds in **Table 1** and produce a CSV with validated SMILES strings.
## Input Files
- Paper PDF: `table1\input\source_paper.pdf`
## Software
- Launch the mandated SMILE... | [
"Open the manuscript PDF from `input/source_paper.pdf`.",
"Locate Table 1 in the EGT710 paper and inspect compounds `1` through `9`.",
"Reconstruct each full molecule and validate the structure with the staged ChemInfo launcher.",
"Save one CSV exactly to `output/submission.csv`."
] | [
"Chrome"
] | [
{
"name": "Source paper",
"format": "`.pdf`",
"path": "input/source_paper.pdf",
"description": "Full paper containing Table 1 and the compound drawings"
}
] | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.2",
"subdomain_code": "chemistry_materials",
"subdomain_name": "Chemistry & Materials Computation"
} | tasks/physical_sciences/egt710_table1_smiles_extraction |
physical_sciences/exact_diag_heisenberg_j1j2 | Exact Diagonalization of the J1-J2 Heisenberg Model | Implement an exact-diagonalization workflow for a frustrated 4x4 spin-1/2 J1-J2 Heisenberg antiferromagnet and produce ground-state, correlation, and dynamical-structure-factor outputs. | physical_sciences | Physics | full-spectrum | You are working on a Linux VM.
## Your Task
Implement an exact diagonalization workflow for the spin-1/2 `J1-J2` Heisenberg antiferromagnet on a `4x4` periodic square lattice in the conserved `S_z = 0` sector.
## Visible Input
- Problem specification: `base/input/problem_spec.md`
## Runtime
- Use the benchmark-provi... | [
"Read the staged scientific specification in `input/problem_spec.md`.",
"Compute Tier 1 energies, spin gap, and a normalized ground-state vector in the `S_z = 0` sector.",
"Compute the full 16x16 spin-spin correlation matrix and static structure factor.",
"Compute the dynamical structure factor on 16 q-points... | [
"Python 3.10",
"NumPy",
"SciPy"
] | [
{
"name": "Problem specification",
"format": "Markdown",
"path": "input/problem_spec.md",
"description": "Visible task statement, output contract, and scientific constraints for the 4x4 J1-J2 exact-diagonalization benchmark."
}
] | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.1",
"subdomain_code": "physics",
"subdomain_name": "Physics"
} | tasks/physical_sciences/exact_diag_heisenberg_j1j2 |
physical_sciences/gillespie_gene_regulatory_network | Gillespie SSA For A Tristable Gene Regulatory Network | Implement exact Gillespie stochastic simulation and tau-leaping in NumPy for a three-gene mutual-inhibition chemical kinetics model, then report validation, multistability, bifurcation, and acceleration results. | physical_sciences | Systems & Microbial Biology | near-term | You are implementing a NumPy-only stochastic simulation workflow for a
three-gene mutual-inhibition regulatory network. The full scientific
specification is in `base/input/problem_spec.md`.
## Your Task
1. Read `base/input/problem_spec.md` end to end. It defines:
- Tier 1: exact Gillespie SSA for a birth-death proc... | [
"Read `input/problem_spec.md` for the birth-death validation, three-gene mutual-inhibition model, bifurcation scan, tau-leaping requirements, seeds, event counts, and JSON schemas.",
"Provision the Python runtime via `uv sync --frozen --project input/runtime_env` and run with the resulting NumPy environment.",
... | [
"Python",
"NumPy"
] | [
{
"name": "Problem specification",
"format": "`.md`",
"path": "input/problem_spec.md",
"description": "Full task specification with models, parameters, seeds, required simulations, output schemas, banned libraries, and runtime budget."
},
{
"name": "Runtime manifest",
"format": "`.toml`"... | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.4",
"subdomain_code": "systems_microbial",
"subdomain_name": "Systems & Microbial Biology"
} | tasks/physical_sciences/gillespie_gene_regulatory_network |
physical_sciences/glm_lake_calibration | GLM Lake Calibration | Calibrate a staged GLM 3 Lake Mendota model by editing one namelist file so the simulated vertical temperature profiles match the staged field observations. | physical_sciences | Environmental Modeling, Engineering & Water Resources | full-spectrum | You are calibrating GLM (General Lake Model) 3 for Lake Mendota, Wisconsin.
## Your Task
Tune the staged GLM namelist so the simulated vertical water-temperature
profiles match the staged observations with RMSE below 1.5 C.
## Visible Inputs
- GLM namelist to edit: `base/input/glm3.nml`
- Observation file: `base/inpu... | [
"Work in the staged Linux benchmark tree under `/media/user/data/agenthle/physical_sciences/glm_lake_calibration/base/`.",
"Modify only `input/glm3.nml`.",
"Run the staged entrypoint `software/run_glm_from_input.sh`, which exposes the bundled GLM binary and libraries.",
"Produce `output/output.nc` covering th... | [
"Python"
] | [
{
"name": "glm3.nml",
"format": "GLM namelist",
"path": "input/glm3.nml",
"description": "Only benchmark file the agent may modify; it contains the GLM parameter settings the agent must tune"
},
{
"name": "field_temp_oxy.csv",
"format": "CSV",
"path": "input/field_temp_oxy.csv",
... | {
"domain_id": "12",
"domain_code": "agriculture_env",
"subdomain_id": "12.1",
"subdomain_code": "env_water",
"subdomain_name": "Environmental Modeling, Engineering & Water Resources"
} | tasks/physical_sciences/glm_lake_calibration |
physical_sciences/hst_acs_wfc_visit_reduction | HST ACS/WFC Visit Reduction | Implement a reproducible Python reduction workflow for synthetic HST ACS/WFC imaging visits, including masking, alignment, drizzling, source photometry, astrometry, and QC reporting. | physical_sciences | Astronomy & Astrophysics | full-spectrum | You are reducing synthetic HST ACS/WFC visit data on a Linux VM. Read the full
instructions in `base/input/TASK_PROMPT.md` and use the visible visit data under:
`base/input/acs_visit_f606w_lockman`.
## Goal
Implement a reusable ACS/WFC visit-reduction script at:
`base/output/reduce_visit.py`
Your script must accept:
... | [
"Read the sanitized task prompt, starter project, visible ACS/WFC visit, and runtime manifest under `input/`.",
"Implement `output/reduce_visit.py` as a reusable command-line reducer for any visit folder with the staged schema.",
"Mask cosmic-ray and hot-pixel DQ pixels, align the dithered exposures, and create... | [
"Python",
"NumPy",
"Astropy",
"Photutils"
] | [
{
"name": "Task prompt",
"format": "`.md`",
"path": "input/TASK_PROMPT.md",
"description": "Sanitized agent-facing instructions and required output contract."
},
{
"name": "Starter reducer",
"format": "`.py`",
"path": "input/starter_project/reduce_visit.py",
"description": "Incom... | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.3",
"subdomain_code": "astronomy",
"subdomain_name": "Astronomy & Astrophysics"
} | tasks/physical_sciences/hst_acs_wfc_visit_reduction |
physical_sciences/ketcher_smiles_reproduction | Ketcher SMILES Reproduction | Recreate the target molecule in Ketcher from a staged structure image and export the result as a SMILES file. | physical_sciences | Chemistry & Materials Computation | near-term | You are a chemistry researcher reproducing a molecular structure in Ketcher.
## Your Task
Recreate molecule **7YY** from the provided SVG image and export it as a SMILES file.
## Input Files
- Molecule structure image: `7yy\input\structure.svg`
## Software
- Launch Ketcher from: `7yy\software\Ketcher.lnk`
- The shor... | [
"Open the task-local Ketcher shortcut from `software/`.",
"Inspect the source molecule drawing in `input/structure.svg`.",
"Recreate the molecule in Ketcher while preserving atom identities, bond orders, and ring connectivity.",
"Export a single-entry SMILES file.",
"Save the exported file exactly to `outpu... | [
"Ketcher"
] | [
{
"name": "Molecule structure image",
"format": "`.svg`",
"path": "input/structure.svg",
"description": "Reference drawing for molecule 7YY"
}
] | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.2",
"subdomain_code": "chemistry_materials",
"subdomain_name": "Chemistry & Materials Computation"
} | tasks/physical_sciences/ketcher_smiles_reproduction |
physical_sciences/lenacapavir_sar_table2_extraction | Lenacapavir SAR Table2 Extraction | Extract every compound row from Table 2 of the staged manuscript, reconstruct the full molecules, and deliver a CSV with SMILES and potency values. | physical_sciences | Biomolecular Structure & Design | near-term | You are a medicinal chemistry researcher extracting one SAR table from a staged paper PDF.
## Your Task
Use the provided manuscript PDF to extract every compound in **Table 2** and reconstruct a full-molecule SMILES string for each row.
## Input Files
- Manuscript PDF: `table2\input\source_paper.pdf`
## Software
- P... | [
"Open the manuscript PDF from `input/source_paper.pdf`.",
"Locate Table 2, titled `SAR for R1 Analogs`, on page 8 of the paper and use the scaffold drawn immediately above that table together with the per-row R1 substituent drawings.",
"Reconstruct the full molecule for each listed ligand in Table 2 and write a... | [
"Microsoft Edge",
"Visual Studio Code",
"Python"
] | [
{
"name": "Source paper",
"format": "`.pdf`",
"path": "input/source_paper.pdf",
"description": "Full lenacapavir medicinal chemistry paper containing Table 2 and the scaffold figure"
}
] | {
"domain_id": "3",
"domain_code": "life_sciences",
"subdomain_id": "3.1",
"subdomain_code": "biomolecular",
"subdomain_name": "Biomolecular Structure & Design"
} | tasks/physical_sciences/lenacapavir_sar_table2_extraction |
physical_sciences/molecular_structure_plausibility | Molecular Structure Plausibility | Inspect a set of molecular XYZ files and identify the structures that are physically implausible under basic chemistry and geometry constraints. | physical_sciences | Chemistry & Materials Computation | near-term | You are filtering molecular structure files based on physical plausibility.
## Variant
`base`: Molecular structure plausibility filtering
## Input Files
- Structure directory: `base/input/xyz_files`
- Task brief: `base/input/task_brief.md`
- Optional Python dependency manifest: `base/input/runtime_env`
## Optional P... | [
"Read `input/task_brief.md` and inspect the 54 molecular `.xyz` files under `input/xyz_files/`.",
"Identify every file whose molecular structure is physically implausible under the brief, using chemistry and geometry sanity checks.",
"Write `output/problematic_structures.txt` with one exact `.xyz` filename per ... | [
"Python",
"RDKit",
"NumPy",
"SciPy"
] | [
{
"name": "task_brief.md",
"format": "Markdown",
"path": "input/task_brief.md",
"description": "Agent-visible instructions for the molecular plausibility filtering task."
},
{
"name": "xyz_files",
"format": "XYZ molecule directory",
"path": "input/xyz_files",
"description": "Dire... | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.2",
"subdomain_code": "chemistry_materials",
"subdomain_name": "Chemistry & Materials Computation"
} | tasks/physical_sciences/molecular_structure_plausibility |
physical_sciences/mose2_bse_absorption_soc | Mose2 Bse Absorption Soc | Your Task Compute the SOC enabled GW BSE optical response of monolayer MoSe2 from the staged structure and pseudopotentials. | physical_sciences | Chemistry & Materials Computation | last-exam | You are a computational materials scientist working on a Linux VM.
## Your Task
Compute the SOC-enabled GW-BSE optical response of monolayer MoSe2 from the staged structure and pseudopotentials.
## Input Files
- MoSe2 structure: `base/input/MoSe2.vasp`
- Mo pseudopotential: `base/input/Mo.soc.upf`
- Se pseudopotentia... | [
"Read the staged input files under `input/`.",
"Create the missing QE and BerkeleyGW input decks from scratch. At minimum this includes QE SCF and NSCF inputs with explicit SOC/noncollinear settings plus BerkeleyGW inputs for `epsilon`, `sigma`, `kernel`, and `absorption`.",
"Run the QE mean-field workflow for ... | [
"NumPy",
"Python"
] | [
{
"name": "MoSe2 structure",
"format": "`vasp`",
"path": "input/MoSe2.vasp",
"description": "Monolayer MoSe2 structure"
},
{
"name": "Mo pseudopotential",
"format": "`UPF`",
"path": "input/Mo.soc.upf",
"description": "SOC-enabled Mo pseudopotential"
},
{
"name": "Se pseud... | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.2",
"subdomain_code": "chemistry_materials",
"subdomain_name": "Chemistry & Materials Computation"
} | tasks/physical_sciences/mose2_bse_absorption_soc |
physical_sciences/phonon_dispersion_thermodynamics | 2D Hexagonal Lattice Phonon Dispersion And Thermodynamics | Construct the dynamical matrix for a two-atom 2D hexagonal lattice, then compute the staged 1D validation, high-symmetry phonon dispersion, phonon DOS, and thermodynamic observables. | physical_sciences | Physics | last-exam | You are working on a Linux VM.
## Variant
`base`: Phonon dispersion and thermodynamics for a 2D hexagonal lattice
## Your Task
Use the staged problem statement to compute the phonon dispersion relation, phonon density of states, and thermodynamic properties for the specified 2D hexagonal lattice.
## Input Files
- Pr... | [
"Read the staged problem specification at input/problem_spec.md.",
"Construct the lattice dynamical matrix without using phonon or materials-science frameworks.",
"Produce diatomic_1d.npz, dispersion_2d.npz, dos.npz, thermodynamics.npz, and results.json under output/.",
"Match the staged file formats, array k... | [
"Python",
"uv"
] | [
{
"name": "problem_spec.md",
"format": "Markdown",
"path": "input/problem_spec.md",
"description": "Public task statement with lattice parameters, required numerical workflow, and exact output contract."
},
{
"name": "pyproject.toml",
"format": "TOML",
"path": "input/runtime_env/pypr... | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.1",
"subdomain_code": "physics",
"subdomain_name": "Physics"
} | tasks/physical_sciences/phonon_dispersion_thermodynamics |
physical_sciences/qm9_mmff94_forcefield_survey_1 | QM9 MMFF94 Force-Field Failure Survey | Survey the QM9 quantum chemistry dataset to characterize systematic failures of the MMFF94 molecular mechanics force field relative to B3LYP/6-31G(2df,p) reference geometries, across a five-phase computational pipeline. | physical_sciences | Chemistry & Materials Computation | last-exam | You are surveying systematic failures of the MMFF94 molecular-mechanics force field relative to B3LYP/6-31G(2df,p) reference geometries across the QM9 dataset.
## Variant
`base`: Five-phase MMFF94 vs QM9 force-field failure survey on the full QM9 dataset.
## Input
- QM9 archive (bz2-compressed tar of 130,831 XYZ file... | [
"Phase 1: Scan all 130,831 QM9 molecules streamed from the bz2 archive; for those with exactly 2 heteroatoms (O/N/F), embed one ETKDG conformer (randomSeed=42), MMFF94-optimize, and save rows with `|Discrepancy_A| ≥ 1.0` to `force_field_failures.csv` with continuous flushing.",
"Phase 2: For every Phase 1 candida... | [
"Python",
"RDKit",
"NumPy",
"pandas",
"matplotlib"
] | [
{
"name": "dsgdb9nsd.xyz.tar.bz2",
"format": "bzip2-compressed tar of XYZ files",
"path": "input/dsgdb9nsd.xyz.tar.bz2",
"description": "QM9 dataset of 130,831 molecules with B3LYP/6-31G(2df,p) geometries; stream with tarfile, do not extract on disk."
}
] | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.2",
"subdomain_code": "chemistry_materials",
"subdomain_name": "Chemistry & Materials Computation"
} | tasks/physical_sciences/qm9_mmff94_forcefield_survey_1 |
physical_sciences/silicon_bse_absorption | Silicon Bse Absorption | Your Task Compute the GW BSE absorption spectrum of bulk silicon from the staged structure and pseudopotential. | physical_sciences | Chemistry & Materials Computation | last-exam | You are a computational materials scientist working on a Linux VM.
## Your Task
Compute the GW-BSE absorption spectrum of bulk silicon from the staged structure and pseudopotential.
## Input Files
- Silicon structure: `base/input/silicon.vasp`
- Silicon pseudopotential: `base/input/Si.UPF`
## Software
Use the task-l... | [
"Read the staged input files under `input/`.",
"Create the missing QE and BerkeleyGW input decks from scratch. At minimum this includes QE SCF and NSCF inputs plus BerkeleyGW inputs for `epsilon`, `sigma`, `kernel`, `absorption`, and `inteqp`.",
"Run the QE mean-field workflow for bulk silicon to generate the c... | [
"NumPy",
"Python"
] | [
{
"name": "Silicon structure",
"format": "`vasp`",
"path": "input/silicon.vasp",
"description": "Bulk Si structure for the optical task"
},
{
"name": "Silicon pseudopotential",
"format": "`UPF`",
"path": "input/Si.UPF",
"description": "Norm-conserving Si pseudopotential"
}
] | {
"domain_id": "2",
"domain_code": "physical_sciences",
"subdomain_id": "2.2",
"subdomain_code": "chemistry_materials",
"subdomain_name": "Chemistry & Materials Computation"
} | tasks/physical_sciences/silicon_bse_absorption |
psychology_neuro/celegans_neuron_tracking | C. elegans Neuron Tracking | Complete sparse whole-brain neuron trajectories in a staged HDF5 recording by preserving neuron identity across time and writing the finished track file for one of four variants. | psychology_neuro | Computational Neuroscience | full-spectrum | You are completing a C. elegans neuron-tracking task on Linux.
## Task Directory
`137`
## Visible Inputs
- Task file: `137/input/137.h5`
- Solve guide: `137/input/AGENT_README.md`
- Variant metadata: `137/input/variant_manifest.json`
- Staged runtime manifest: `137/input/runtime_env/pyproject.toml`
- Staged lockfile:... | [
"Read the staged solve guide and variant manifest for the selected recording.",
"Open the sparse task HDF5 through the POINTS GUI launcher or the staged Python wrapper.",
"Fill `points[t, 1..30, :]` across time while preserving neuron identity and leaving uncertain coordinates as `NaN`.",
"Write the completed... | [
"POINTS",
"Python",
"h5py",
"PyQt5"
] | [
{
"name": "<variant>.h5",
"format": "HDF5",
"path": "input/<variant>.h5",
"description": "Sparse whole-brain neuron-tracking task file with 3 seeded frames and volumetric image data."
},
{
"name": "AGENT_README.md",
"format": "Markdown",
"path": "input/AGENT_README.md",
"descript... | {
"domain_id": "5",
"domain_code": "psychology_neuro",
"subdomain_id": "5.1",
"subdomain_code": "comp_neuro",
"subdomain_name": "Computational Neuroscience"
} | tasks/psychology_neuro/celegans_neuron_tracking |
psychology_neuro/reddit_ai_post_codebook_boolean_coding | Reddit AI Post Codebook Boolean Coding | Read a psychology codebook PDF and complete a boolean-coded Reddit-post annotation workbook in LibreOffice Calc. | psychology_neuro | Economics & Quantitative Social Research | near-term | You are completing a psychology coding workbook on a Windows VM.
## Input Files
- Dataset workbook: `base\input\ai_addiction_dataset_n95.xlsx`
- Codebook PDF: `base\input\ai_addiction_codebook.pdf`
## Software
- Open the workbook with `base\software\open_dataset.bat`
- Open the codebook with `base\software\open_codeb... | [
"Open the staged Reddit-post workbook and the staged codebook PDF.",
"Read the codebook definitions and skip rules.",
"Fill the boolean coding columns F through AR for Excel rows 2 through 96.",
"Choose exactly one focus label in columns F through H for each post row.",
"Save the completed workbook to `outp... | [
"LibreOffice Calc",
"Microsoft Edge PDF Viewer"
] | [
{
"name": "ai_addiction_dataset_n95.xlsx",
"format": "Excel workbook",
"path": "input/ai_addiction_dataset_n95.xlsx",
"description": "Visible Reddit-post dataset workbook with post metadata plus blank coding columns"
},
{
"name": "ai_addiction_codebook.pdf",
"format": "PDF",
"path": ... | {
"domain_id": "13",
"domain_code": "social_sciences",
"subdomain_id": "13.1",
"subdomain_code": "economics",
"subdomain_name": "Economics & Quantitative Social Research"
} | tasks/psychology_neuro/reddit_ai_post_codebook_boolean_coding |
psychology_neuro/scene2_resample | Scene2 Resample | Resample a 1 mm ROI mask onto the 2 mm statistical-map grid and save the required aligned output. | psychology_neuro | Experimental Psychology & Neuroimaging | full-spectrum | You are a neuroimaging analyst completing a GUI workflow in 3D Slicer 5.0.3.
## Your Task
Workflow 2: resample a 1mm ROI mask to the 2mm statistical-map grid
Goal:
1. Open the staged ROI mask and statistical map in 3D Slicer.
2. Resample roi_mask_1mm.nii.gz onto statmap_z_2mm.nii.gz using nearest-neighbor interpolati... | [
"Launch 3D Slicer from `software/launch_gui.sh`.",
"Open the staged `roi_mask_1mm.nii.gz` and `statmap_z_2mm.nii.gz` volumes.",
"Resample the 1 mm ROI mask onto the 2 mm stat-map grid using nearest-neighbor interpolation.",
"Save the resampled mask to `output/roi_mask_2mm_nn.nii.gz` and export a readable scre... | [
"Python"
] | [
{
"name": "statmap_z_2mm.nii.gz",
"format": "NIfTI volume",
"path": "input/statmap_z_2mm.nii.gz",
"description": "Reference 2 mm grid that the output mask must match."
},
{
"name": "roi_mask_1mm.nii.gz",
"format": "NIfTI mask",
"path": "input/roi_mask_1mm.nii.gz",
"description": ... | {
"domain_id": "5",
"domain_code": "psychology_neuro",
"subdomain_id": "5.2",
"subdomain_code": "exp_psychology",
"subdomain_name": "Experimental Psychology & Neuroimaging"
} | tasks/psychology_neuro/scene2_resample |
social_sciences/atwood_2022_measles_vaccine_reproduction | Atwood 2022 Measles Vaccine Coefficient Reproduction | Reproduce the six Table 2 vaccination-effect coefficient estimates from Atwood (2022) using the published paper, appendix, and archived replication package. | social_sciences | Economics & Quantitative Social Research | full-spectrum | You are working on a Linux VM as an applied-economics replication analyst.
## Task Directory
`base`
## Visible Inputs
- Task prompt: `base/input/TASK_PROMPT.md`
- Published paper: `base/input/paper.pdf`
- Online appendix: `base/input/paper-online-appendix.pdf`
- Archived replication package: `base/input/replication_p... | [
"Read the paper, online appendix, task metadata, and archived replication package.",
"Extract the six Table 2 \"Vaccination effect\" point estimates.",
"Inspect or run the replication package enough to report corresponding code coefficients.",
"Write the six required structured output files under the runtime ... | [
"Python",
"PDF tools",
"unzip"
] | [
{
"name": "TASK_PROMPT.md",
"format": "md",
"path": "input/TASK_PROMPT.md",
"description": "Detailed solve-time instructions and output contract."
},
{
"name": "paper.pdf",
"format": "pdf",
"path": "input/paper.pdf",
"description": "Published Atwood (2022) paper."
},
{
"n... | {
"domain_id": "13",
"domain_code": "social_sciences",
"subdomain_id": "13.1",
"subdomain_code": "economics",
"subdomain_name": "Economics & Quantitative Social Research"
} | tasks/social_sciences/atwood_2022_measles_vaccine_reproduction |
transport_safety/abm_hangzhou_metro | Hangzhou Metro Passenger Simulation | Simulate one operating day of Hangzhou metro passenger trips from staged AFC demand, GIS, and network configuration inputs. | transport_safety | Urban & Spatial Planning | full-spectrum | You are working on a Linux VM to produce a one-day Hangzhou metro passenger simulation output bundle.\n\n## Variant\n`base`: Hangzhou metro daily AFC simulation bundle\n\n## Visible Input Files\n- task prompt: `base/input/task_prompt.md`\n- output contract: `base/input/output_contract.json`\n- AFC demand: `base/input/d... | [
"Read the staged task prompt and output contract from `input/`.",
"Build a one-day Hangzhou metro passenger simulation from the AFC, GIS, and network configuration inputs.",
"Write `passenger_records.csv` and `validation_report.txt` into `output/` with the required schema.",
"Use the staged runtime manifest o... | [
"Python",
"uv",
"geopandas",
"matplotlib",
"networkx",
"numpy",
"pandas"
] | [
{
"name": "afc_hangzhou.csv",
"format": "CSV",
"path": "input/data/afc_hangzhou.csv",
"description": "Visible AFC trip records for one operating day."
},
{
"name": "hangzhou_lines.json",
"format": "GeoJSON",
"path": "input/gis/hangzhou_lines.json",
"description": "Hangzhou metro ... | {
"domain_id": "1",
"domain_code": "engineering",
"subdomain_id": "1.10",
"subdomain_code": "urban_planning",
"subdomain_name": "Urban & Spatial Planning"
} | tasks/transport_safety/abm_hangzhou_metro |
transport_safety/capacitated_vehicle_routing_problems | Capacitated Vehicle Routing Problems | Generate VRPLIB-format near-best-known CVRP solutions for three selected benchmark instances. | transport_safety | Mathematical & Operations Research | near-term | You are working on a Linux VM.
## Your Task
Build a reproducible CVRP-solving workflow and generate VRPLIB-format solutions for exactly three selected instances.
## Visible Inputs
- Problem specification: `base/input/problem_spec.md`
- Instance directory: `base/input/instances/`
- Runtime manifest: `base/input/runtim... | [
"Inspect the three staged `.vrp` files and implement a CVRP-solving workflow.",
"Use the staged wrapper if needed to run Python with the pinned task runtime.",
"Write the three required `.sol` files under `output/solutions/`."
] | [
"Python",
"uv",
"PyVRP",
"VRPLIB"
] | [
{
"name": "Problem specification",
"format": "Markdown",
"path": "input/problem_spec.md",
"description": "Public benchmark instructions and output contract."
},
{
"name": "Selected CVRP instances",
"format": "VRPLIB",
"path": "input/instances/",
"description": "Three benchmark in... | {
"domain_id": "9",
"domain_code": "computing_math",
"subdomain_id": "9.4",
"subdomain_code": "math_ops_research",
"subdomain_name": "Mathematical & Operations Research"
} | tasks/transport_safety/capacitated_vehicle_routing_problems |
transport_safety/fds_single_compartment_detector_reconstruction | FDS Single-Compartment Detector Reconstruction | Complete a Linux Python CLI that reconstructs a synthetic single-compartment FDS/Smokeview detector-response case from visible fire-safety engineering inputs. | transport_safety | Fire Science & Public Safety | full-spectrum | You are acting as a senior fire protection engineer reconstructing a synthetic single-compartment detector-response incident on a Linux VM.
## Visible Starter Project
- Starter project directory: `base/input/project`
- Task prompt: `base/input/TASK_PROMPT.md`
- Visible scenario input: `base/input/project/input/visible... | [
"Inspect the visible starter project and task prompt under input/project.",
"Infer the HRR ramp from the supplied field_sensor_traces.csv rather than hard-coding case IDs or output values.",
"Complete reconstruct_fire_case.py so it works for any same-schema scenario directory.",
"Place the completed reconstru... | [
"Python 3.10",
"FDS 6.10.1",
"Smokeview 6.10.1"
] | [
{
"name": "Starter project",
"format": "Python project directory",
"path": "input/project/",
"description": "Visible starter CLI, visible scenario inputs, prompt, and reconstruction contract."
}
] | {
"domain_id": "10",
"domain_code": "transport_safety",
"subdomain_id": "10.3",
"subdomain_code": "fire_safety",
"subdomain_name": "Fire Science & Public Safety"
} | tasks/transport_safety/fds_single_compartment_detector_reconstruction |
visual_media/atlas_outpost_graybox_navigation | Atlas Outpost Graybox Navigation | Finish a Windows GPU Blender/Godot graybox level blockout so the exported Atlas Outpost layout supports the required visible navigation routes and review constraints. | visual_media | 3D, Animation & Interactive Media | last-exam | You are working on a Windows GPU VM.
## Variant
`base`: Atlas Outpost graybox navigation blockout
## Your Task
Finish the Atlas Outpost graybox level-design starter project. This is a level-blockout workflow: author route geometry, re-export the glTF, wire the Godot validation scene, and write the visible handoff met... | [
"Read the visible task prompt and starter project under `input/`.",
"Complete the Blender/Godot graybox project and exported glTF route geometry.",
"Update the Godot validation scene and visible handoff notes.",
"Save the completed project with the same relative paths under `output/`."
] | [
"Blender 4.5.1 LTS",
"Godot Engine 4.6.2",
"Python",
"Windows"
] | [
{
"name": "TASK_PROMPT.md",
"format": "Markdown",
"path": "input/TASK_PROMPT.md",
"description": "Agent-visible task brief with visible route, zone, cover, and LOS requirements."
},
{
"name": "atlas_outpost_blockout.blend",
"format": "Blender file",
"path": "input/blender/atlas_outpo... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.1",
"subdomain_code": "animation_3d",
"subdomain_name": "3D, Animation & Interactive Media"
} | tasks/visual_media/atlas_outpost_graybox_navigation |
visual_media/blender_character_reconstruction_from_multiview_01 | Blender Character Reconstruction From Multiview | Reconstruct a stylized full-body character in Blender from five fixed clay reference views plus scale/orientation metadata, then save the scene and export the required deliverables under the exact submission layout. | visual_media | 3D, Animation & Interactive Media | last-exam | You are a 3D artist using Blender on a Windows VM.
Reconstruct the provided stylized character from the staged multiview references.
Agent-visible input:
- Modeling brief: `base\input\modeling_brief.md`
- Scale/orientation guide: `base\input\scale_orientation_guide.json`
- Validation camera guide: `base\input\validat... | [
"Open Blender through the staged launcher and reconstruct the stylized full-body character from the five staged clay reference views.",
"Use `scale_orientation_guide.json` to keep the character upright, centered, and scaled to the intended world-space frame.",
"Use `validation_views.json` as the fixed camera/re... | [
"Blender"
] | [
{
"name": "modeling_brief.md",
"format": "Markdown",
"path": "input/modeling_brief.md",
"description": "High-level modeling objective, required components, allowed simplifications, and required output files."
},
{
"name": "scale_orientation_guide.json",
"format": "JSON",
"path": "inp... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.1",
"subdomain_code": "animation_3d",
"subdomain_name": "3D, Animation & Interactive Media"
} | tasks/visual_media/blender_character_reconstruction_from_multiview_01 |
visual_media/butterfly_flap_animation | butterfly_flap_animation | Animate a supplied butterfly image in Adobe After Effects with four wing-flap cycles and a smooth curved 2.5D flight path. | visual_media | 3D, Animation & Interactive Media | near-term | Create a butterfly wing-flap animation in Adobe After Effects.
Remote workspace:
- Task folder: base
Software:
- Launch Adobe After Effects using: base\software\launch_after_effects.bat
- The installed After Effects version may differ from the original submission; use the available AE version opened by this launcher.... | [
"Launch Adobe After Effects from `software/launch_after_effects.bat` (intended tool); if AE is unavailable in the environment, any other approach that produces the required output is acceptable.",
"Import `input/butterfly.jpeg` and use it as the source butterfly artwork.",
"Split or mask the butterfly into a ce... | [
"Adobe After Effects"
] | [
{
"name": "Butterfly artwork",
"format": "JPEG image",
"path": "input/butterfly.jpeg",
"description": "Agent-visible source artwork to animate"
},
{
"name": "Task specification",
"format": "Markdown",
"path": "input/task_spec.md",
"description": "Public animation and export requi... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.1",
"subdomain_code": "animation_3d",
"subdomain_name": "3D, Animation & Interactive Media"
} | tasks/visual_media/butterfly_flap_animation |
visual_media/chroma_key_from_reference | chroma_key_from_reference | Isolate the correct foreground subject from an existing scene and rebuild it as a green-screen plate in DaVinci Resolve. | visual_media | 3D, Animation & Interactive Media | near-term | Goal: Isolate the correct foreground subject from the existing scene in the source clip, create a new Resolve project yourself, place that subject over a uniform bright-green background as indicated by the first-frame target screenshot, and export the finished green-screen plate.
Remote workspace:
- Task folder: crane... | [
"Launch DaVinci Resolve from `software/DaVinci Resolve.lnk`",
"Create a new Resolve project for the task run",
"Isolate the target foreground subject from the existing scene in `input/input.mp4`",
"Replace the original scene background with a uniform bright-green background in Resolve",
"Use `input/input.pn... | [
"DaVinci Resolve"
] | [
{
"name": "Input clip",
"format": "`.mp4`",
"path": "input/input.mp4",
"description": "Source composite containing the target foreground subject in an existing scene"
},
{
"name": "Input target screenshot",
"format": "`.png`",
"path": "input/input.png",
"description": "First-fram... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.1",
"subdomain_code": "animation_3d",
"subdomain_name": "3D, Animation & Interactive Media"
} | tasks/visual_media/chroma_key_from_reference |
visual_media/compress_3dgs_scene_ply | Compress A 3D Gaussian Splatting Scene | Compress a pretrained 3D Gaussian Splatting scene while preserving rendering quality on hidden LLFF holdout views, then report the resulting quality metrics and compression ratio. | visual_media | 3D, Animation & Interactive Media | near-term | You are working on a Windows GPU neural-rendering benchmark.
Read these staged inputs first:
- `base\input\task_prompt.md`
- `base\input\baseline_results.json`
- `base\input\scene_manifest.json`
Core visible asset roots:
- `base\input\scene\point_cloud_30000.ply`
- `base\input\scene\sparse\0`
- `base\input\scene\imag... | [
"Read the staged prompt, baseline metrics, manifest, source PLY, sparse COLMAP bundle, and visible scene images.",
"Compress `input/scene/point_cloud_30000.ply` into `output/point_cloud_30000_compressed.ply`.",
"Render one holdout-view image for every filename listed in `input/scene_manifest.json` and save them... | [
"Gaussian Splatting",
"Python"
] | [
{
"name": "task_prompt.md",
"format": "Markdown",
"path": "input/task_prompt.md",
"description": "Agent-facing task contract and reminders."
},
{
"name": "baseline_results.json",
"format": "JSON",
"path": "input/baseline_results.json",
"description": "Baseline average metrics for... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.1",
"subdomain_code": "animation_3d",
"subdomain_name": "3D, Animation & Interactive Media"
} | tasks/visual_media/compress_3dgs_scene_ply |
visual_media/human_mesh_animation_reproduction | Human Mesh Animation Reproduction | Rig and animate a provided human mesh in Blender so its exported per-frame OBJ sequence matches the motion in a reference video. | visual_media | 3D, Animation & Interactive Media | last-exam | You are a 3D artist using Blender 4.3 on a Windows VM.
Rig and animate the provided human mesh so it matches the motion shown in the reference video.
Agent-visible inputs:
- Mesh: `base\input\character.obj`
- Material sidecar: `base\input\character.mtl`
- Reference motion video: `base\input\reference.mp4`
Required s... | [
"Rig and animate `input/character.obj` so the motion matches `input/reference.mp4`.",
"Work at 30 fps and export exactly 60 OBJ frames for frames 1 through 60 inclusive.",
"Save the authored Blender file as `output/submission/final.blend`.",
"Export `output/submission/mesh_seq/frame_0001.obj` through `frame_0... | [
"Blender"
] | [
{
"name": "character.obj",
"format": "OBJ mesh",
"path": "input/character.obj",
"description": "Unrigged human mesh that the agent must animate"
},
{
"name": "character.mtl",
"format": "MTL material",
"path": "input/character.mtl",
"description": "Material sidecar for the input m... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.1",
"subdomain_code": "animation_3d",
"subdomain_name": "3D, Animation & Interactive Media"
} | tasks/visual_media/human_mesh_animation_reproduction |
visual_media/inkscape_cultural_poster_design | Inkscape Cultural Poster Design | Create a single-page cultural exhibition poster in Inkscape on Windows using a staged design brief, a source photo, and a structured instance specification, then save the final poster as an SVG. | visual_media | Graphic, Visual & Product Design | near-term | You are using Inkscape on Windows to design a cultural exhibition poster.
Read:
- `base\input\design_brief.txt`
- `base\input\instance_spec.txt` (GBK/GB18030 encoded Chinese)
- `base\input\installation_photo_01.jpg`
Create the poster in Inkscape and save the final SVG exactly to:
- `base\output\poster.svg`
Notes:
- ... | [
"Read the staged design brief, instance specification, and installation photo from `input/`.",
"Use Inkscape on Windows to design a single-page cultural exhibition poster.",
"Preserve the source photo's aspect ratio when placing it into the poster.",
"Save the final SVG exactly as `output/poster.svg`."
] | [
"Inkscape"
] | [
{
"name": "design_brief.txt",
"format": "text",
"path": "input/design_brief.txt",
"description": "Natural-language poster brief describing the exhibition theme and design direction."
},
{
"name": "instance_spec.txt",
"format": "text",
"path": "input/instance_spec.txt",
"descripti... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.2",
"subdomain_code": "graphic_design",
"subdomain_name": "Graphic, Visual & Product Design"
} | tasks/visual_media/inkscape_cultural_poster_design |
visual_media/music_transcription | Music Transcription | Listen to a recorded orchestral piece and transcribe it into a professional music score, exporting PDF, MIDI with correct instrument program assignments, and a screenshot of the notation software. | visual_media | Audio, Music & Post-Production Media | full-spectrum | Goal: Transcribe a recorded piece into musical notation, export both a PDF score and a MIDI file with correct instrument assignments. You may use any music notation software available on the system.
Read the task specification from `task_brief.json` in the variant's `input/` directory. It contains:
- `title`: the titl... | [
"Read `input/task_brief.json` for title, composer, tempo_bpm, and the instruments list (with clef and GM program).",
"Listen to `input/reference_song.mp3` to identify all instrumental parts.",
"Open a music notation application (Dorico 6 is staged) and create a new project with matching title, composer, and pla... | [
"Dorico 6"
] | [
{
"name": "Task specification",
"format": "`.json`",
"path": "input/task_brief.json",
"description": "Piece title, composer, tempo_bpm, and per-instrument name / clef / GM program number."
},
{
"name": "Reference recording",
"format": "`.mp3`",
"path": "input/reference_song.mp3",
... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.4",
"subdomain_code": "audio_music",
"subdomain_name": "Audio, Music & Post-Production Media"
} | tasks/visual_media/music_transcription |
visual_media/project_migration | Cubase Project Migration | Repair a Cubase project that opens with missing plugins by substituting equivalent installed VSTs, then save the migrated session, export stems, and capture a verification screenshot. | visual_media | Audio, Music & Post-Production Media | last-exam | Goal: Migrate a Cubase project to this computer by replacing all missing/invalid VST plugins with equivalent ones from the available plugins, then export stems and a mixdown.
Software:
- Cubase: \software\Cubase.lnk
Input:
- Cubase project: \input\project.cpr
Before you begin:
1. Generate the list of VST3 plugins in... | [
"Open `input/project.cpr` in Cubase from `software/Cubase.lnk`.",
"Replace each missing or invalid VST with a functionally equivalent installed plugin from `input/available_vsts.txt`.",
"Verify the migrated project plays back with audible, artifact-free tracks.",
"Export one WAV stem per project-defined track... | [
"Cubase"
] | [
{
"name": "Cubase project",
"format": "CPR",
"path": "input/project.cpr",
"description": "The broken Cubase session that opens with missing plugins on the target VM."
},
{
"name": "Available VST list",
"format": "TXT",
"path": "input/available_vsts.txt",
"description": "Agent-vis... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.4",
"subdomain_code": "audio_music",
"subdomain_name": "Audio, Music & Post-Production Media"
} | tasks/visual_media/project_migration |
visual_media/skeletal_animation_reproduction | Skeletal Animation Reproduction | Rig and animate the staged character in Blender so it reproduces the target body motion from the reference video. | visual_media | 3D, Animation & Interactive Media | last-exam | You are a 3D artist using Blender.
Your task is to rig and animate the provided character so that it reproduces the body motion shown in the reference video.
Official input:
- Unrigged mesh: `skeletal_animation_reproduction_singing_anime_character\input\Singing.obj`
- Material sidecar: `skeletal_animation_reproductio... | [
"Unrigged mesh: `E:\\agenthle\\visual_media\\skeletal_animation_reproduction\\skeletal_animation_reproduction_singing_anime_character\\input\\Singing.obj`",
"Material sidecar: `E:\\agenthle\\visual_media\\skeletal_animation_reproduction\\skeletal_animation_reproduction_singing_anime_character\\input\\Singing.mtl`... | [
"Blender"
] | [
{
"name": "<character>.obj",
"format": "obj",
"path": "input/<character>.obj",
"description": "**Input:** input/<character>.obj, input/<character>.mtl, input/reference.mp4"
},
{
"name": "<character>.mtl",
"format": "mtl",
"path": "input/<character>.mtl",
"description": "**Input:*... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.1",
"subdomain_code": "animation_3d",
"subdomain_name": "3D, Animation & Interactive Media"
} | tasks/visual_media/skeletal_animation_reproduction |
visual_media/uv_reproduction | UV Reproduction | Recreate the required UV layout and material appearance of the staged 3D asset in Blender. | visual_media | 3D, Animation & Interactive Media | last-exam | You are a 3D artist using Blender.
Your task is to reproduce the UV unwrap and textured material appearance of the provided model.
Official input:
- Mesh: `uv_reproduction_anime_singing_girl\input\singing_raw.obj`
- Optional input material sidecar: `None`
- Agent-visible reference images: `uv_reproduction_anime_singi... | [
"inspect the input object and agent-visible reference images",
"create or preserve valid UVs when required",
"create or attach materials and textures when required",
"export submission files into the required output directory"
] | [
"Blender"
] | [
{
"name": "*.obj",
"format": "obj",
"path": "input/*.obj",
"description": "input/*.obj"
},
{
"name": "*.mtl",
"format": "mtl",
"path": "input/*.mtl",
"description": "optional input/*.mtl"
},
{
"name": "*.png",
"format": "png",
"path": "input/reference_images/*.png... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.1",
"subdomain_code": "animation_3d",
"subdomain_name": "3D, Animation & Interactive Media"
} | tasks/visual_media/uv_reproduction |
visual_media/video_storyboard_001 | Pre-Restoration Shot Log: Vintage Animation Remaster | Produce a temporal per-shot continuity log (DOCX) for a public-domain 1931 Yasuji Murata animation as the archival-documentation pass that precedes a restoration / remaster pipeline. The log must preserve the factual visual record the downstream QC team's fact-check brief depends on, without the archivist answering or ... | visual_media | Audio, Music & Post-Production Media | near-term | ## Context
A small restoration studio is about to begin a scan + remaster pass on a public-domain print of a vintage Yasuji Murata animation (*A Wolf is a Wolf*, 1931). Before any restoration work starts, the pipeline requires an **archival shot log / continuity report** of the source print: a temporal, per-shot factu... | [
"Inspect the full source print end-to-end.",
"Use the QC fact-check brief to identify the visual facts the shot log must preserve (without answering the brief).",
"Write a temporal shot log with segment IDs, in/out times, visual descriptions, actions, objects, people/animals, and relevant on-screen text or dial... | [
"VLC",
"Word processor (any DOCX-capable editor)",
"Python"
] | [
{
"name": "Yasuji_Murata-_A_Wolf_is_a_Wolf_(1931).ogv",
"format": "ogv",
"path": "input/Yasuji_Murata-_A_Wolf_is_a_Wolf_(1931).ogv",
"description": "Input short video to inspect."
},
{
"name": "video_storyboard_qa_set.docx",
"format": "docx",
"path": "input/video_storyboard_qa_set.do... | {
"domain_id": "8",
"domain_code": "visual_media",
"subdomain_id": "8.4",
"subdomain_code": "audio_music",
"subdomain_name": "Audio, Music & Post-Production Media"
} | tasks/visual_media/video_storyboard_001 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.