{ "schema_version": 2, "benchmark": "Terminal-Bench 2.1", "dataset_source": "terminal-bench/terminal-bench-2-1", "dataset_pin_manifest_sha256": "245aacd4d3275aae4ab3722ce90cb8b9d90cdd0889e6c8dba46c4e09cdb2d246", "dataset_task_listing_sha256": "949932f06303fc67e39b5cf64c56a27f0ab8029380d4edab85ae7ad0826f9caf", "preclassification_sha256": "ab7190e0e7913929c6ca52c91ac3fa748c37d7fc4b7d957f9b68a349bd32b28b", "seed": 20260908, "training_fraction_target": 0.6, "selected_clusters": 14, "method": { "features": "Public task instructions, tools, operations and constraints only", "classification": "Researcher-defined semantic preclasses before partition; not official benchmark categories", "split": "Exact rounded 60% total; proportional class quotas with both partitions represented; maximize multi-task fine families represented on both sides within each class; seeded SHA256 combination order breaks ties", "purpose": "Provide opportunities to retrieve related training experience; task-level holdout, not unseen-family evaluation", "forbidden_inputs": [ "historical scores", "trajectories", "solutions", "verifier code" ] }, "training_task_ids": [ "adaptive-rejection-sampler", "bn-fit-modify", "break-filter-js-from-html", "build-pmars", "caffe-cifar-10", "cancel-async-tasks", "circuit-fibsqrt", "cobol-modernization", "configure-git-webserver", "count-dataset-tokens", "crack-7z-hash", "custom-memory-heap-crash", "distribution-search", "dna-insert", "extract-elf", "feal-linear-cryptanalysis", "financial-document-processor", "fix-code-vulnerability", "gcode-to-text", "git-leak-recovery", "hf-model-inference", "install-windows-3.11", "kv-store-grpc", "largest-eigenval", "llm-inference-batching-scheduler", "log-summary-date-ranges", "make-doom-for-mips", "mcmc-sampling-stan", "merge-diff-arc-agi-task", "modernize-scientific-stack", "mteb-retrieve", "multi-source-data-merger", "nginx-request-logging", "openssl-selfsigned-cert", "overfull-hbox", "path-tracing-reverse", "polyglot-c-py", "protein-assembly", "pytorch-model-recovery", "qemu-startup", "query-optimize", "raman-fitting", "regex-chess", "regex-log", "sam-cell-seg", "sanitize-git-repo", "schemelike-metacircular-eval", "sqlite-db-truncate", "sqlite-with-gcov", "torch-tensor-parallelism", "video-processing", "vulnerable-secret", "write-compressor" ], "test_task_ids": [ "mteb-leaderboard", "path-tracing", "headless-terminal", "train-fasttext", "pytorch-model-cli", "mailman", "prove-plus-comm", "torch-pipeline-parallelism", "filter-js-from-html", "make-mips-interpreter", "gpt2-codegolf", "extract-moves-from-video", "pypi-server", "constraints-scheduling", "db-wal-recovery", "code-from-image", "winning-avg-corewars", "fix-ocaml-gc", "dna-assembly", "feal-differential-cryptanalysis", "large-scale-text-editing", "fix-git", "git-multibranch", "chess-best-move", "sparql-university", "portfolio-optimization", "build-pov-ray", "rstan-to-pystan", "password-recovery", "polyglot-rust-c", "compile-compcert", "model-extraction-relu-logits", "build-cython-ext", "reshard-c4-data", "qemu-alpine-ssh", "tune-mjcf" ], "clusters": [ { "cluster_id": "C01", "class_id": "C01", "label": "Source builds, compatibility and code repair", "definition": "Build or adapt existing source code against a specified compiler, library ABI or language/runtime version, preserving its required behavior.", "total": 9, "training": 5, "test": 4, "training_fraction": 0.5555555555555556, "repeated_fine_families": 3, "fine_families_on_both_sides": 3, "task_ids": [ "build-cython-ext", "build-pmars", "build-pov-ray", "cobol-modernization", "compile-compcert", "custom-memory-heap-crash", "fix-ocaml-gc", "modernize-scientific-stack", "sqlite-with-gcov" ], "training_task_ids": [ "build-pmars", "cobol-modernization", "custom-memory-heap-crash", "modernize-scientific-stack", "sqlite-with-gcov" ], "test_task_ids": [ "build-cython-ext", "build-pov-ray", "compile-compcert", "fix-ocaml-gc" ] }, { "cluster_id": "C02", "class_id": "C02", "label": "Git history, recovery, merging and deployment", "definition": "Inspect or change Git repositories, object/history state, branches and hooks while preserving required unrelated contents.", "total": 6, "training": 4, "test": 2, "training_fraction": 0.6666666666666666, "repeated_fine_families": 2, "fine_families_on_both_sides": 2, "task_ids": [ "configure-git-webserver", "fix-git", "git-leak-recovery", "git-multibranch", "merge-diff-arc-agi-task", "sanitize-git-repo" ], "training_task_ids": [ "configure-git-webserver", "git-leak-recovery", "merge-diff-arc-agi-task", "sanitize-git-repo" ], "test_task_ids": [ "fix-git", "git-multibranch" ] }, { "cluster_id": "C03", "class_id": "C03", "label": "Persistent network services and protocol configuration", "definition": "Create or configure a service endpoint and its protocol artifacts, then leave a verified usable service or credential configuration.", "total": 5, "training": 3, "test": 2, "training_fraction": 0.6, "repeated_fine_families": 2, "fine_families_on_both_sides": 2, "task_ids": [ "kv-store-grpc", "mailman", "nginx-request-logging", "openssl-selfsigned-cert", "pypi-server" ], "training_task_ids": [ "kv-store-grpc", "nginx-request-logging", "openssl-selfsigned-cert" ], "test_task_ids": [ "mailman", "pypi-server" ] }, { "cluster_id": "C04", "class_id": "C04", "label": "Runtime control, emulation and executable interfaces", "definition": "Implement or configure process control, virtual machines, instruction emulation or executable-format access through an exact interface.", "total": 8, "training": 5, "test": 3, "training_fraction": 0.625, "repeated_fine_families": 3, "fine_families_on_both_sides": 3, "task_ids": [ "cancel-async-tasks", "extract-elf", "headless-terminal", "install-windows-3.11", "make-doom-for-mips", "make-mips-interpreter", "qemu-alpine-ssh", "qemu-startup" ], "training_task_ids": [ "cancel-async-tasks", "extract-elf", "install-windows-3.11", "make-doom-for-mips", "qemu-startup" ], "test_task_ids": [ "headless-terminal", "make-mips-interpreter", "qemu-alpine-ssh" ] }, { "cluster_id": "C05", "class_id": "C05", "label": "Web input validation and HTML security", "definition": "Analyze, exploit or repair web input handling and HTML sanitization under behavioral security tests.", "total": 3, "training": 2, "test": 1, "training_fraction": 0.6666666666666666, "repeated_fine_families": 1, "fine_families_on_both_sides": 1, "task_ids": [ "break-filter-js-from-html", "filter-js-from-html", "fix-code-vulnerability" ], "training_task_ids": [ "break-filter-js-from-html", "fix-code-vulnerability" ], "test_task_ids": [ "filter-js-from-html" ] }, { "cluster_id": "C06", "class_id": "C06", "label": "Cryptanalysis and secret recovery", "definition": "Recover a hidden key, password, plaintext or secret through cryptographic analysis, constrained candidate search or executable/storage inspection.", "total": 5, "training": 3, "test": 2, "training_fraction": 0.6, "repeated_fine_families": 2, "fine_families_on_both_sides": 2, "task_ids": [ "crack-7z-hash", "feal-differential-cryptanalysis", "feal-linear-cryptanalysis", "password-recovery", "vulnerable-secret" ], "training_task_ids": [ "crack-7z-hash", "feal-linear-cryptanalysis", "vulnerable-secret" ], "test_task_ids": [ "feal-differential-cryptanalysis", "password-recovery" ] }, { "cluster_id": "C07", "class_id": "C07", "label": "Statistical modeling, numerical inference and optimization", "definition": "Implement or infer a quantitative model, sample or fit its parameters, or optimize a numerical computation/plan while meeting explicit mathematical constraints.", "total": 11, "training": 7, "test": 4, "training_fraction": 0.6363636363636364, "repeated_fine_families": 2, "fine_families_on_both_sides": 2, "task_ids": [ "adaptive-rejection-sampler", "bn-fit-modify", "distribution-search", "largest-eigenval", "llm-inference-batching-scheduler", "mcmc-sampling-stan", "model-extraction-relu-logits", "portfolio-optimization", "raman-fitting", "rstan-to-pystan", "tune-mjcf" ], "training_task_ids": [ "adaptive-rejection-sampler", "bn-fit-modify", "distribution-search", "largest-eigenval", "llm-inference-batching-scheduler", "mcmc-sampling-stan", "raman-fitting" ], "test_task_ids": [ "model-extraction-relu-logits", "portfolio-optimization", "rstan-to-pystan", "tune-mjcf" ] }, { "cluster_id": "C08", "class_id": "C08", "label": "Molecular sequence and construct design", "definition": "Read biological sequences or structures and construct a sequence-level design that satisfies explicit biological and numerical constraints.", "total": 3, "training": 2, "test": 1, "training_fraction": 0.6666666666666666, "repeated_fine_families": 1, "fine_families_on_both_sides": 1, "task_ids": [ "dna-assembly", "dna-insert", "protein-assembly" ], "training_task_ids": [ "dna-insert", "protein-assembly" ], "test_task_ids": [ "dna-assembly" ] }, { "cluster_id": "C09", "class_id": "C09", "label": "Structured data transformations, queries and text extraction", "definition": "Read structured or semi-structured records and produce exact transformed, queried or aggregated outputs under schema, format and completeness constraints.", "total": 9, "training": 5, "test": 4, "training_fraction": 0.5555555555555556, "repeated_fine_families": 4, "fine_families_on_both_sides": 4, "task_ids": [ "constraints-scheduling", "count-dataset-tokens", "large-scale-text-editing", "log-summary-date-ranges", "multi-source-data-merger", "query-optimize", "regex-log", "reshard-c4-data", "sparql-university" ], "training_task_ids": [ "count-dataset-tokens", "log-summary-date-ranges", "multi-source-data-merger", "query-optimize", "regex-log" ], "test_task_ids": [ "constraints-scheduling", "large-scale-text-editing", "reshard-c4-data", "sparql-university" ] }, { "cluster_id": "C10", "class_id": "C10", "label": "SQLite storage recovery", "definition": "Recover logical rows from damaged SQLite database or WAL files and export complete structured records.", "total": 2, "training": 1, "test": 1, "training_fraction": 0.5, "repeated_fine_families": 1, "fine_families_on_both_sides": 1, "task_ids": [ "db-wal-recovery", "sqlite-db-truncate" ], "training_task_ids": [ "sqlite-db-truncate" ], "test_task_ids": [ "db-wal-recovery" ] }, { "cluster_id": "C11", "class_id": "C11", "label": "Neural model, embedding and inference workflows", "definition": "Select, load, reconstruct, train, partition or expose a neural/embedding model with required parameter and inference/output contracts.", "total": 11, "training": 6, "test": 5, "training_fraction": 0.5454545454545454, "repeated_fine_families": 4, "fine_families_on_both_sides": 4, "task_ids": [ "caffe-cifar-10", "gpt2-codegolf", "hf-model-inference", "mteb-leaderboard", "mteb-retrieve", "pytorch-model-cli", "pytorch-model-recovery", "sam-cell-seg", "torch-pipeline-parallelism", "torch-tensor-parallelism", "train-fasttext" ], "training_task_ids": [ "caffe-cifar-10", "hf-model-inference", "mteb-retrieve", "pytorch-model-recovery", "sam-cell-seg", "torch-tensor-parallelism" ], "test_task_ids": [ "gpt2-codegolf", "mteb-leaderboard", "pytorch-model-cli", "torch-pipeline-parallelism", "train-fasttext" ] }, { "cluster_id": "C12", "class_id": "C12", "label": "Constrained language and program synthesis", "definition": "Construct a program, proof or rewrite/encoding artifact in a restricted language or execution model and validate exact semantics.", "total": 8, "training": 5, "test": 3, "training_fraction": 0.625, "repeated_fine_families": 1, "fine_families_on_both_sides": 1, "task_ids": [ "circuit-fibsqrt", "polyglot-c-py", "polyglot-rust-c", "prove-plus-comm", "regex-chess", "schemelike-metacircular-eval", "winning-avg-corewars", "write-compressor" ], "training_task_ids": [ "circuit-fibsqrt", "polyglot-c-py", "regex-chess", "schemelike-metacircular-eval", "write-compressor" ], "test_task_ids": [ "polyglot-rust-c", "prove-plus-comm", "winning-avg-corewars" ] }, { "cluster_id": "C13", "class_id": "C13", "label": "Document and media interpretation or transformation", "definition": "Extract or modify information represented in images, video, rendered paths or formatted documents, using format-aware evidence and validation.", "total": 7, "training": 4, "test": 3, "training_fraction": 0.5714285714285714, "repeated_fine_families": 2, "fine_families_on_both_sides": 2, "task_ids": [ "chess-best-move", "code-from-image", "extract-moves-from-video", "financial-document-processor", "gcode-to-text", "overfull-hbox", "video-processing" ], "training_task_ids": [ "financial-document-processor", "gcode-to-text", "overfull-hbox", "video-processing" ], "test_task_ids": [ "chess-best-move", "code-from-image", "extract-moves-from-video" ] }, { "cluster_id": "C14", "class_id": "C14", "label": "Compact native program reconstruction", "definition": "Reconstruct a compact independent native program from visual or executable reference evidence under static-build and source-size restrictions.", "total": 2, "training": 1, "test": 1, "training_fraction": 0.5, "repeated_fine_families": 1, "fine_families_on_both_sides": 1, "task_ids": [ "path-tracing", "path-tracing-reverse" ], "training_task_ids": [ "path-tracing-reverse" ], "test_task_ids": [ "path-tracing" ] } ], "tasks": [ { "task_id": "adaptive-rejection-sampler", "task_version_id": "59306a3e-89ef-4868-bd01-a173e8827f2b", "task_revision": 3, "source_path": "prepared/tasks/adaptive-rejection-sampler", "instruction_sha256": "e6d28c28acd76bef61817427c74c2a29e3e8d75f535cbd708afcbfd02c54013f", "task_config_sha256": "6ed7b37f74beb46e9d47b0f5fad53025667c9548baff4e96d467b56d429fab1a", "feature_text_sha256": "89abdefba7eabdb7a50fbdd3c8a55fe60fc58e9f86a9157c297e420b8132bc28", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/adaptive-rejection-sampler:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "log_concave_rejection_sampling", "classification_reason": "Implements an R rejection sampler with density/shape checks and statistical tests against known distributions; transferable work is sampling and validating stochastic outputs.", "transfer_tags": [ "R", "log-concavity", "distribution-validation" ], "split": "training" }, { "task_id": "bn-fit-modify", "task_version_id": "f367f9eb-c5d0-4262-9110-1b53bf1c4e8b", "task_revision": 4, "source_path": "prepared/tasks/bn-fit-modify", "instruction_sha256": "5f5207ce23c375270c26719fb66d89843bc88cf1d40671ed87c6b7bab4d394e5", "task_config_sha256": "6493f7162da424a4fd95da721bd3fd1ce7e5dd26e792132cb9c46db536b67090", "feature_text_sha256": "f53c3f67a820c2a1f4479c5a4e97bf3c0d40240fad0a98dde28716913d3288da", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/bn-fit-modify:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "probabilistic_graphical_models", "classification_reason": "Infers a DAG from tabular samples, fits a Bayesian network, intervenes on one variable and resamples while preserving the schema.", "transfer_tags": [ "Bayesian-network", "DAG", "causal-intervention", "sampling" ], "split": "training" }, { "task_id": "break-filter-js-from-html", "task_version_id": "b0e89bbe-c0b0-48a2-91b7-645f32de8b3a", "task_revision": 4, "source_path": "prepared/tasks/break-filter-js-from-html", "instruction_sha256": "ad42558111b2a8e8534bdf52b584b804348e1a8d68d756f812a31d5147e65be5", "task_config_sha256": "f820eb1b47e4ccc548b87d9a418fef519137dcf975ad73d0fb5246df00a9669b", "feature_text_sha256": "a97e1c6b248d8d7b3510d6b141b808760a2b3f14f82f898c9846143ac3f896d4", "official_agent_timeout_seconds": 1200.0, "official_verifier_timeout_seconds": 1200.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/break-filter-js-from-html:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C05", "class_id": "C05", "class_label": "Web input validation and HTML security", "fine_family": "html_sanitization", "classification_reason": "Constructs HTML that survives an existing sanitizer and automatically executes JavaScript; directly shares parser and payload behavior with the sanitizer task.", "transfer_tags": [ "HTML", "XSS", "browser", "adversarial-input" ], "split": "training" }, { "task_id": "build-cython-ext", "task_version_id": "647bec6c-2733-4adc-af79-f3827d6f4c75", "task_revision": 3, "source_path": "prepared/tasks/build-cython-ext", "instruction_sha256": "7e52874e45f88505bd21540229719de6639a401477557d814489981d472f0e34", "task_config_sha256": "c472f0e7522c128fb7f03e00b50c8cdf2e5e6474daa1e094e7441f4b54e770b2", "feature_text_sha256": "35d9e4460f913d5b768fe1e09a740f803c78e2d4851ca055d563ecb806802403", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/build-cython-ext:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C01", "class_id": "C01", "class_label": "Source builds, compatibility and code repair", "fine_family": "python_dependency_compatibility", "classification_reason": "Repairs and compiles existing Cython extensions for NumPy 2.3.0, retaining package layout and verifying imports and existing functionality.", "transfer_tags": [ "Cython", "NumPy", "ABI", "source-build" ], "split": "test" }, { "task_id": "build-pmars", "task_version_id": "de60d8e0-2749-44b3-b297-193a0b502b1e", "task_revision": 3, "source_path": "prepared/tasks/build-pmars", "instruction_sha256": "dd4a3aa762e6b3c09e67bd7f61d9b6afbe379dcd7297c2c0ebe2e9bd0082215e", "task_config_sha256": "7ee2eac27b99fd462e0760552a430454db693f86608af5bb570b020270e4c9cd", "feature_text_sha256": "af87061109bc68e65415549121e748d318644aff9580fc0ce99b8cd10524fd8e", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/build-pmars:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C01", "class_id": "C01", "class_label": "Source builds, compatibility and code repair", "fine_family": "native_source_build", "classification_reason": "Builds a Debian-sourced native simulator without X11 and checks its installed binary and debugger behavior.", "transfer_tags": [ "C", "Debian", "headless-build", "linker", "pMARS" ], "split": "training" }, { "task_id": "build-pov-ray", "task_version_id": "4060d3ad-9318-4f34-9b75-f30996ce4aa2", "task_revision": 3, "source_path": "prepared/tasks/build-pov-ray", "instruction_sha256": "56c8ecdd0084a07f8c97986c43af5c37ed03dd5aa994da6735cf59806162d1b2", "task_config_sha256": "b90a47ec26c1191d3c523cc094ea96b170e42fbfe176ef1f814c57d05390c097", "feature_text_sha256": "1e36fbcf79548d4b060fa065df4e24f042f3dabc5ba62e8b0d3ea3815cba9a85", "official_agent_timeout_seconds": 12000.0, "official_verifier_timeout_seconds": 12000.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/build-pov-ray:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C01", "class_id": "C01", "class_label": "Source builds, compatibility and code repair", "fine_family": "native_source_build", "classification_reason": "Obtains and builds a legacy native renderer, installs its executable and validates it with an unchanged reference scene.", "transfer_tags": [ "C", "legacy-source", "compiler", "renderer" ], "split": "test" }, { "task_id": "caffe-cifar-10", "task_version_id": "36d1e2da-13b5-4961-a24d-7025385623a9", "task_revision": 4, "source_path": "prepared/tasks/caffe-cifar-10", "instruction_sha256": "46907b0ee048f74269c29a15e0e086546b2b1220149597017da8b5ccc05a0cea", "task_config_sha256": "2831165ad9d9e8bbc646e6d85c0a2290640dcc4f13749547a2937c7675012a8b", "feature_text_sha256": "819163a182b813f1a9693d6139e1c6ff47a2409239fd2a44b5e1838d649e54ed", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 1200.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/caffe-cifar-10:20260403", "cpus": 4, "memory_mb": 8192, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "supervised_model_training", "classification_reason": "Builds CPU Caffe and runs prescribed CIFAR-10 training, producing a named model artifact and checking train/test accuracy behavior.", "transfer_tags": [ "Caffe", "CPU", "CNN", "training-artifact", "native-build" ], "split": "training" }, { "task_id": "cancel-async-tasks", "task_version_id": "989e7aa0-bf1b-457b-8762-a36157c4ff5b", "task_revision": 4, "source_path": "prepared/tasks/cancel-async-tasks", "instruction_sha256": "23d22f7039b50682941652c716f96dac314b8585c32c133614cac77dc067f47e", "task_config_sha256": "14db0675e04408f2133b0067e8dbddf66a1d6e7baf6b8d38bef8ecccfc4e58a8", "feature_text_sha256": "28053b31055ae569416bd728f2512704a3076b31b81a2b81d54c98e9f0160733", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/cancel-async-tasks:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C04", "class_id": "C04", "class_label": "Runtime control, emulation and executable interfaces", "fine_family": "process_lifecycle_and_cleanup", "classification_reason": "Implements a bounded asynchronous task runner whose cleanup still executes on interruption; the central issue is lifecycle and cancellation control.", "transfer_tags": [ "Python", "asyncio", "cancellation", "concurrency", "cleanup" ], "split": "training" }, { "task_id": "chess-best-move", "task_version_id": "b5dfe64f-483a-4679-b114-37d3ce0ec13c", "task_revision": 3, "source_path": "prepared/tasks/chess-best-move", "instruction_sha256": "6ba25be4e291a4ad04ae42b2dddefcdaef67ca45ca9d5dce661a2e517b19ce7a", "task_config_sha256": "d8f01d3c84ed5836edbb49c6c3b3cc7b5b96d951cf01a0987467c9206fb22f9e", "feature_text_sha256": "7e58fcfa8229671cd722cfae6cad422a214442cdf86a341b338a3d71a25bb42e", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/chess-best-move:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C13", "class_id": "C13", "class_label": "Document and media interpretation or transformation", "fine_family": "visual_symbol_to_structure", "classification_reason": "Reads a chess position from an image and converts a selected move to coordinate notation; image-to-structured-state extraction precedes chess analysis.", "transfer_tags": [ "image-decoding", "board-coordinates", "chess-engine", "structured-transcription" ], "split": "test" }, { "task_id": "circuit-fibsqrt", "task_version_id": "3b71d43a-80ac-4463-8fbc-f7ff0da7bcd3", "task_revision": 4, "source_path": "prepared/tasks/circuit-fibsqrt", "instruction_sha256": "a7f500266a878d04ac9de39fba76220cfad216d8f6b8063bba93bebb3d4ab9ea", "task_config_sha256": "9f5aa1f7ebbbcec90fd109121e9bd509af3cfff6b4cec701bd91947988f40972", "feature_text_sha256": "de7ee9e7502aaad2f898ccd9809c94873ff5e9e90a4ecb30710f41d74fa16b51", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/circuit-fibsqrt:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C12", "class_id": "C12", "class_label": "Constrained language and program synthesis", "fine_family": "logic_gate_arithmetic_synthesis", "classification_reason": "Synthesizes a bounded logic-gate program computing an exact integer function in the supplied simulator's restricted instruction model.", "transfer_tags": [ "logic-gates", "bit-arithmetic", "circuit-generation", "resource-bound" ], "split": "training" }, { "task_id": "cobol-modernization", "task_version_id": "40f60d0e-5ed0-4f0c-be4e-8acdb6474b7a", "task_revision": 4, "source_path": "prepared/tasks/cobol-modernization", "instruction_sha256": "a87074f1141b7345394cd88053d0debb0dc8f7de95308882c6538ee9a3b359da", "task_config_sha256": "4b620211dc3d6b1b989f0210c5ecd2f606974581c8bce27bd78659d48cf0cfbd", "feature_text_sha256": "14d034b111c2fc00aab44164829f4a3937ade1c1112f19228893f66097c2b210", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/cobol-modernization:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C01", "class_id": "C01", "class_label": "Source builds, compatibility and code repair", "fine_family": "behavior_preserving_language_migration", "classification_reason": "Ports existing COBOL file-update semantics to Python and checks byte-identical data files from identical inputs.", "transfer_tags": [ "COBOL", "Python", "fixed-record-files", "differential-checking" ], "split": "training" }, { "task_id": "code-from-image", "task_version_id": "d10aeb38-b44f-403f-8fb3-736b6ec589af", "task_revision": 3, "source_path": "prepared/tasks/code-from-image", "instruction_sha256": "8c3af6483e409defef9786da51f318ae2ca71afe2327b94471d2f0e2e367442c", "task_config_sha256": "3a13c1e252bb8759f51215be0b2b2cb1b4c697f0dd570f2d7702cc28b28f08de", "feature_text_sha256": "573b72433c6beece03f36df038b59a9ffe6e01275f6b229cd9438d408fc6cfa1", "official_agent_timeout_seconds": 1200.0, "official_verifier_timeout_seconds": 1200.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/code-from-image:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C13", "class_id": "C13", "class_label": "Document and media interpretation or transformation", "fine_family": "visual_symbol_to_structure", "classification_reason": "Transcribes pseudocode from an image, implements its intended semantics and emits its final value; visual symbol recovery is the initial interface.", "transfer_tags": [ "image-decoding", "pseudocode", "transcription", "execution" ], "split": "test" }, { "task_id": "compile-compcert", "task_version_id": "6d51e06e-4c80-40a8-bded-759c6711e226", "task_revision": 3, "source_path": "prepared/tasks/compile-compcert", "instruction_sha256": "d6efce07c47484fcd13c7918f68bd4af6314392de43751c4cad1e62530c2ad06", "task_config_sha256": "1d9618d7178cfa15918a293da126b867f3a7db160e604d4fec369949884dac39", "feature_text_sha256": "2464d68f59e0fd7412626874c48b20255da972126e2ccee643fc7d6467f87ca0", "official_agent_timeout_seconds": 2400.0, "official_verifier_timeout_seconds": 2400.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/compile-compcert:20260403", "cpus": 2, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C01", "class_id": "C01", "class_label": "Source builds, compatibility and code repair", "fine_family": "native_source_build", "classification_reason": "Configures and builds an existing verified C compiler for the host architecture, requiring a functioning freshly built executable.", "transfer_tags": [ "Coq", "C", "compiler", "architecture", "source-build" ], "split": "test" }, { "task_id": "configure-git-webserver", "task_version_id": "61b7c71d-b6d9-41a0-8fb9-15c0c98b35fe", "task_revision": 3, "source_path": "prepared/tasks/configure-git-webserver", "instruction_sha256": "62151b7371bc364277a3c118bba755334c640fb0cefc7a5af106e6e763cf1d97", "task_config_sha256": "ebb40d72f689512cbd90d865d88ef966b6665ab215544f9247b898b7fa94751b", "feature_text_sha256": "9876bb5a98e604d868f009b3b0ca4505b3709b1d4d36af85789ed876df1b9a5f", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/configure-git-webserver:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C02", "class_id": "C02", "class_label": "Git history, recovery, merging and deployment", "fine_family": "git_receive_hook_deployment", "classification_reason": "Connects pushes to a Git repository with serving the pushed content over HTTP; repository hooks and deployment correctness are the core operations.", "transfer_tags": [ "Git", "bare-repository", "post-receive", "HTTP", "deployment" ], "split": "training" }, { "task_id": "constraints-scheduling", "task_version_id": "f485a409-500a-46a8-86c4-cbe065d960da", "task_revision": 3, "source_path": "prepared/tasks/constraints-scheduling", "instruction_sha256": "342def78177f9bc41bea4d11af679bc385a9dda2ac8a1e425be384a5b7f937b6", "task_config_sha256": "7b9842de16d74e8faf8c7c0aa03b0b81145941cb12dd322d81ae23fc48607250", "feature_text_sha256": "f4863db75b970c76487da3ab96bf7d97e767f5846ad9b96918df2463d980bc54", "official_agent_timeout_seconds": 1200.0, "official_verifier_timeout_seconds": 1200.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/constraints-scheduling:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C09", "class_id": "C09", "class_label": "Structured data transformations, queries and text extraction", "fine_family": "date_record_constraints", "classification_reason": "Parses ICS events and availability rules, selects a valid earliest meeting interval and emits a schema-correct calendar without changing inputs.", "transfer_tags": [ "ICS", "date-time", "intervals", "constraint-optimization" ], "split": "test" }, { "task_id": "count-dataset-tokens", "task_version_id": "b259c87f-dd28-4b66-a7c4-64a33b6ec31d", "task_revision": 3, "source_path": "prepared/tasks/count-dataset-tokens", "instruction_sha256": "baa843c8da9804d68ced71b2e4dea91afb19dd33de2c867b4a15542563a888f9", "task_config_sha256": "8220a1b8e6e08ecb95556920de358b6aa74b112d8a79625db6915f6b316ccd18", "feature_text_sha256": "925d47e443747dec2f1f2bcf420f566ac05c610fa8e4b34fea2a3203778f9533", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/count-dataset-tokens:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C09", "class_id": "C09", "class_label": "Structured data transformations, queries and text extraction", "fine_family": "nlp_corpus_processing", "classification_reason": "Loads a specified dataset, selects a domain and aggregates token counts using a mandated tokenizer and dataset usage instructions.", "transfer_tags": [ "Hugging-Face", "dataset-filtering", "tokenizer", "aggregation" ], "split": "training" }, { "task_id": "crack-7z-hash", "task_version_id": "c26ecbfa-4f30-457f-89f7-77bc3ab8ab0a", "task_revision": 5, "source_path": "prepared/tasks/crack-7z-hash", "instruction_sha256": "f5f89400c549eefb2c245caa7770356737b7a12098e6879787e846d9cb5eb822", "task_config_sha256": "8e192bcbe00e817419bde7ff9b562a5fa0c1e24ec4f3384cba4a31eda0a7b013", "feature_text_sha256": "af767f8634fcb7dd760a7a7bf6286d17d11535b4f8f36999a6a54bcdbf9107ad", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/crack-7z-hash:20251031", "cpus": 1, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C06", "class_id": "C06", "class_label": "Cryptanalysis and secret recovery", "fine_family": "password_and_artifact_recovery", "classification_reason": "Recovers access to a protected archive and extracts the required word from its contained file.", "transfer_tags": [ "7z", "password-search", "archive", "plaintext-validation" ], "split": "training" }, { "task_id": "custom-memory-heap-crash", "task_version_id": "d0f2d72a-1354-40ca-b432-a31309a32418", "task_revision": 4, "source_path": "prepared/tasks/custom-memory-heap-crash", "instruction_sha256": "53c68ae14a3c3af9a97c40bfd9bdc30a5d98b37eaa2f69f763475745f70f108d", "task_config_sha256": "ed474be1983e212913aa8866ebb66216acbe36aef2339ac375d232aa46617481", "feature_text_sha256": "890884bb65daef8ab1ad1ad80b2d4909c0a0aa1b52a333fe1eda05af56d7184a", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/custom-memory-heap-crash:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C01", "class_id": "C01", "class_label": "Source builds, compatibility and code repair", "fine_family": "native_runtime_memory_debugging", "classification_reason": "Diagnoses a release-only crash against custom standard-library builds, fixes only the allowed source file and verifies no memory leaks.", "transfer_tags": [ "C++", "heap", "Valgrind", "debug-release", "restricted-edit" ], "split": "training" }, { "task_id": "db-wal-recovery", "task_version_id": "a10991ec-f974-4bfa-8cc9-8b01b08e71d0", "task_revision": 4, "source_path": "prepared/tasks/db-wal-recovery", "instruction_sha256": "5a20375162921b8911c67035d1905df7fd335d1bed03044271e608b2c3c93b26", "task_config_sha256": "1cdc2c615e259bc10265efbe71b518332070c1c002e226566399e9788f18d0aa", "feature_text_sha256": "f996bfb658623147e4e3ec8b937dbf01874cb03604ffef4971ddac9605df13af", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/db-wal-recovery:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C10", "class_id": "C10", "class_label": "SQLite storage recovery", "fine_family": "sqlite_file_recovery", "classification_reason": "Repairs or interprets a damaged SQLite WAL so both base rows and pending WAL rows can be recovered into ordered JSON.", "transfer_tags": [ "SQLite", "WAL", "binary-recovery", "JSON" ], "split": "test" }, { "task_id": "distribution-search", "task_version_id": "de2c61de-0f75-46f1-96ec-f80a082af856", "task_revision": 3, "source_path": "prepared/tasks/distribution-search", "instruction_sha256": "aba2e1835b2f827686402c3d3ff9b309aab8c9cbac26e23fc46aaeb34e958915", "task_config_sha256": "70c8e73cafd0a75306375bec2c2e1e4d82e0cb969fd7e38a63e7c344de43b681", "feature_text_sha256": "5ebf0e1849c4229c5ec210fa6e23b563a5ecbfcdcc36ae4c635f795b215075ef", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/distribution-search:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "kl_constrained_probability_construction", "classification_reason": "Constructs a valid high-dimensional probability vector meeting two precise KL constraints, with numerical tolerance checks.", "transfer_tags": [ "NumPy", "SciPy", "KL-divergence", "probability-vector", "root-finding" ], "split": "training" }, { "task_id": "dna-assembly", "task_version_id": "18930920-f6a1-4258-9750-16d7334a741a", "task_revision": 3, "source_path": "prepared/tasks/dna-assembly", "instruction_sha256": "88998c98beabdd966ab46c04ede822bbc74796696aa1bf87c567a9131210533c", "task_config_sha256": "7b571544b64d9be1c91526837e5d805c2181afe0e6fd57a7f0676bba36c868d7", "feature_text_sha256": "4777797944653d9cc3e5658cded1c2b5dffa66f660ccde8d9ad552523ffcb9e1", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/dna-assembly:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C08", "class_id": "C08", "class_label": "Molecular sequence and construct design", "fine_family": "dna_primer_design", "classification_reason": "Designs minimal PCR primer pairs for circular-plasmid assembly while satisfying enzyme-site, annealing-length and melting-temperature constraints.", "transfer_tags": [ "FASTA", "primer3", "Golden-Gate", "melting-temperature", "sequence-design" ], "split": "test" }, { "task_id": "dna-insert", "task_version_id": "742d9c38-b66a-4760-99c8-62c3a1b74d92", "task_revision": 3, "source_path": "prepared/tasks/dna-insert", "instruction_sha256": "49fce0ee8b73d7ccdf4b8476c6bb87cf362689fd1d77b21a0d8737d16019a473", "task_config_sha256": "1a9efd5c826d64d31c4874b395f1f9c488b2f26581b0c752fc4619ff177f25eb", "feature_text_sha256": "565009dba68d2baa4577eee26f90c781256620ee2e125e08045879c2b5eda014", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/dna-insert:20251031", "cpus": 1, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C08", "class_id": "C08", "class_label": "Molecular sequence and construct design", "fine_family": "dna_primer_design", "classification_reason": "Designs minimal mutagenesis primer pairs transforming one circular plasmid into another, sharing sequence and primer-temperature validation with assembly.", "transfer_tags": [ "FASTA", "primer3", "mutagenesis", "melting-temperature", "sequence-design" ], "split": "training" }, { "task_id": "extract-elf", "task_version_id": "5efe7b1c-51a2-43cb-a235-b4ff16af6639", "task_revision": 3, "source_path": "prepared/tasks/extract-elf", "instruction_sha256": "f1dd8c8fd9f01101d2ecb65665bdedc33adc868abd853b3adf556f6aa4982073", "task_config_sha256": "14e2c28fd7cc1c9c7b64ab0db5565899fed538e38c55074824647cacc3f7a172", "feature_text_sha256": "1499029e9dc8f4cbe047918bd3db99be8628fc03cad55e8d184b2576ac26e53a", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/extract-elf:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C04", "class_id": "C04", "class_label": "Runtime control, emulation and executable interfaces", "fine_family": "mips_elf_execution_interface", "classification_reason": "Implements a JavaScript reader mapping executable addresses to integer memory values, sharing executable layout concepts with MIPS loading/emulation.", "transfer_tags": [ "ELF", "JavaScript", "memory-layout", "binary-parsing" ], "split": "training" }, { "task_id": "extract-moves-from-video", "task_version_id": "b4873679-3f72-4a0b-88e9-b3815961d193", "task_revision": 3, "source_path": "prepared/tasks/extract-moves-from-video", "instruction_sha256": "6697dca6199359a7409962db4609ad5909a8fe2b35e606a61ea63f1d25e44858", "task_config_sha256": "2c9e1b466871da3c3703ec3c6d9835f154b8f1cc34af1f1ba45d64fec56f3605", "feature_text_sha256": "00165fb7ed2955c9b8c66e3f6c2460fab839c85dc59a35d439e882b379725056", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/extract-moves-from-video:20260403", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C13", "class_id": "C13", "class_label": "Document and media interpretation or transformation", "fine_family": "video_frame_interpretation", "classification_reason": "Recovers ordered text commands from a recorded terminal game, requiring frame/text interpretation and temporal deduplication.", "transfer_tags": [ "video", "frame-extraction", "OCR", "temporal-transcription" ], "split": "test" }, { "task_id": "feal-differential-cryptanalysis", "task_version_id": "8738fe4a-7fa5-4aa0-ba42-4123bc5f2a86", "task_revision": 3, "source_path": "prepared/tasks/feal-differential-cryptanalysis", "instruction_sha256": "b4093780b283e5c9838997769fb7bca090e3f4cbbd2d7cb275ba60cdfd722572", "task_config_sha256": "dcb9f2f15047eee0be15ca18c39c357f89187a94aae6baeff276221e2383eecb", "feature_text_sha256": "b7353a6af7c402d67f51da03fe939704b2a8724df745dd4d6eba2220bc2c82a6", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/feal-differential-cryptanalysis:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C06", "class_id": "C06", "class_label": "Cryptanalysis and secret recovery", "fine_family": "feal_key_recovery", "classification_reason": "Recovers a FEAL-like round key from chosen-plaintext queries under a runtime bound and a reduced seed space.", "transfer_tags": [ "FEAL", "chosen-plaintext", "differential-attack", "uint32", "search-bound" ], "split": "test" }, { "task_id": "feal-linear-cryptanalysis", "task_version_id": "c145afba-733d-42ee-aa94-1dbdd63dc539", "task_revision": 3, "source_path": "prepared/tasks/feal-linear-cryptanalysis", "instruction_sha256": "a2f1669d4067af6c7b8ac05ff7d5849ca5df68397c94701645c4b9a3b6f71b03", "task_config_sha256": "a39ec2c427af534c96c719dc87f70eb536956b487ce909d04b6365160e7c9fc4", "feature_text_sha256": "0876cc3cc2185ca8f3d06862e0d3cc9c4df2218f1934094e0c12230d8032b53f", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/feal-linear-cryptanalysis:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C06", "class_id": "C06", "class_label": "Cryptanalysis and secret recovery", "fine_family": "feal_key_recovery", "classification_reason": "Recovers FEAL-like round keys from known plaintext/ciphertext pairs and verifies recovery by decrypting new ciphertexts.", "transfer_tags": [ "FEAL", "known-plaintext", "linear-attack", "key-search", "decryption" ], "split": "training" }, { "task_id": "filter-js-from-html", "task_version_id": "4c67d1e9-f2a8-4c67-8f46-5b05ebc6a597", "task_revision": 4, "source_path": "prepared/tasks/filter-js-from-html", "instruction_sha256": "48d6ac95e8ccb552195ce9fbdea86ba8d5d7ccf54e159c31590797802c5cf171", "task_config_sha256": "f7f888951c97b8120371139bb78d39499c2d3cd7aa3836d57fd6d8b9e4199344", "feature_text_sha256": "97ab248004203f46a4b656fcc6a2d107c34c19058b20f966814079ebb46e3d15", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/filter-js-from-html:20251031", "cpus": 1, "memory_mb": 8192, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C05", "class_id": "C05", "class_label": "Web input validation and HTML security", "fine_family": "html_sanitization", "classification_reason": "Implements an in-place HTML sanitizer that removes script execution while preserving legitimate structure and formatting.", "transfer_tags": [ "HTML", "XSS", "parser", "escaping", "content-preservation" ], "split": "test" }, { "task_id": "financial-document-processor", "task_version_id": "9b071acc-84db-4e17-99e5-0a9ca4849b81", "task_revision": 4, "source_path": "prepared/tasks/financial-document-processor", "instruction_sha256": "b21cf2af1891922732327bf586d14b250d36d96de7867479c15bd1c58ef0db58", "task_config_sha256": "163ee252895c05d36ca37137dba4d6c24b46be28d56ce84ef8bb35bb5b894bda", "feature_text_sha256": "c798f13ef2a4f92255bbcf1ee3070ef6be5e9e65ad6cf2b48f3763b8e5c02b62", "official_agent_timeout_seconds": 1200.0, "official_verifier_timeout_seconds": 1200.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/financial-document-processor:20251031", "cpus": 1, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C13", "class_id": "C13", "class_label": "Document and media interpretation or transformation", "fine_family": "document_text_and_layout", "classification_reason": "Classifies mixed image/PDF documents and extracts invoice totals and tax into a reconciled CSV while moving every source document.", "transfer_tags": [ "PDF", "JPG", "OCR", "document-classification", "amount-extraction" ], "split": "training" }, { "task_id": "fix-code-vulnerability", "task_version_id": "ff8760cb-5d3e-454e-a1d7-5bbc989faa8e", "task_revision": 3, "source_path": "prepared/tasks/fix-code-vulnerability", "instruction_sha256": "89b1abdf0af19399f720233b5cdf6b2e8e0ff7b6e3fe3e2a3af0d4d298a8641a", "task_config_sha256": "9aa2bd14ce9408289d199d0608bc3fa65b8b5e23374c9641020de62197dbf1e2", "feature_text_sha256": "2d6428543a624fbd66fa7021b8d336a256e359305429a20e7878c55b2eb470b0", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/fix-code-vulnerability:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C05", "class_id": "C05", "class_label": "Web input validation and HTML security", "fine_family": "web_input_validation_repair", "classification_reason": "Audits and repairs a web framework's input-handling vulnerability, reports its CWE and verifies precise error behavior with regression tests.", "transfer_tags": [ "Python", "WSGI", "CWE", "input-validation", "exceptions", "pytest" ], "split": "training" }, { "task_id": "fix-git", "task_version_id": "d4611085-331c-469a-a25f-8250df9640d9", "task_revision": 3, "source_path": "prepared/tasks/fix-git", "instruction_sha256": "f98174e47ba5ce040f6fad22ef3cbca4dd27b3192868ea5d082077222c3ec4e8", "task_config_sha256": "f59d711003b974e4ef8ed63556614e2ea9ea7d0e394af13b2071225e3fda7805", "feature_text_sha256": "c92ca55e2f9f173de5d38037664713fc5ee8547edd8438e8fa0fd78b7de08e63", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/fix-git:20260403", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C02", "class_id": "C02", "class_label": "Git history, recovery, merging and deployment", "fine_family": "git_history_recovery_and_hygiene", "classification_reason": "Finds lost repository changes after a branch checkout and incorporates them into the requested branch.", "transfer_tags": [ "Git", "reflog", "object-recovery", "merge" ], "split": "test" }, { "task_id": "fix-ocaml-gc", "task_version_id": "713b4cd8-681d-484d-95f9-eb45c7d07e6c", "task_revision": 3, "source_path": "prepared/tasks/fix-ocaml-gc", "instruction_sha256": "a8867e4f8a537b3b1622ca2b80cbb278eeba772ce95602ed735c2581d23b2dec", "task_config_sha256": "423aab8ce6f44ce9349dcde647ca3cbd09afd5ac023910c25699ed687a3c71b2", "feature_text_sha256": "d11b1be4cf64d6dbf0ccf241413383a6b356cbf5f55b58833a7fd65b183414b7", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/fix-ocaml-gc:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C01", "class_id": "C01", "class_label": "Source builds, compatibility and code repair", "fine_family": "native_runtime_memory_debugging", "classification_reason": "Diagnoses a modified garbage collector causing compiler bootstrap crashes and verifies the repaired native runtime with its basic testsuite.", "transfer_tags": [ "OCaml", "garbage-collector", "heap", "bootstrap", "regression-tests" ], "split": "test" }, { "task_id": "gcode-to-text", "task_version_id": "7371e199-8a3b-48de-9fae-ff99d5c1fda1", "task_revision": 4, "source_path": "prepared/tasks/gcode-to-text", "instruction_sha256": "ee524a8f846b35f1784a7c9fb01c17f077f6b3b61c8b61d0d2f2a87c0be3166b", "task_config_sha256": "3aa7a504fe5b3c6eaab16881b0e6dbbe8ac6d4f59084fa49b1e98ad9c71ea00c", "feature_text_sha256": "531cf11852cb8e57dd3438e11126a374d4931d1d2a67e6d10be9ccaacf624f56", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/gcode-to-text:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C13", "class_id": "C13", "class_label": "Document and media interpretation or transformation", "fine_family": "visual_symbol_to_structure", "classification_reason": "Interprets printer motion paths as rendered text and returns the text that those paths would print.", "transfer_tags": [ "G-code", "toolpaths", "geometry", "text-recognition" ], "split": "training" }, { "task_id": "git-leak-recovery", "task_version_id": "dc5cb523-c8c5-446f-9241-ff7abbb18309", "task_revision": 3, "source_path": "prepared/tasks/git-leak-recovery", "instruction_sha256": "54ac693fd26a4c003f610d62c4b2671f1d1d04ac5b1286f780605e385a9a5861", "task_config_sha256": "d7f23fda4a6ba3fff902d2a475db8ff74feb5d328980696a77068c02c57d3044", "feature_text_sha256": "43838fd9ccc57dc1697a8c84d75663917788deb9692bb5ad4bccd7affeded50b", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/git-leak-recovery:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C02", "class_id": "C02", "class_label": "Git history, recovery, merging and deployment", "fine_family": "git_history_recovery_and_hygiene", "classification_reason": "Recovers a secret from rewritten repository history, then removes its recoverable traces while preserving unrelated files and messages.", "transfer_tags": [ "Git", "objects", "history-rewrite", "secret-removal" ], "split": "training" }, { "task_id": "git-multibranch", "task_version_id": "bb4aafba-05f9-4d2e-b61b-d32907b7eda7", "task_revision": 3, "source_path": "prepared/tasks/git-multibranch", "instruction_sha256": "54014a38d7526d65faeb6c11d9b7ed3d4a5cb5d973d68d3da3b68b8c0289e125", "task_config_sha256": "4b9981604578a31205257cca36a28d0c89c9d8d1058541330acd11d3f8d65f82", "feature_text_sha256": "dc0ddf106555776f44e67dd0aaef4241c776dcf662ac94bca8d7252248895a8a", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/git-multibranch:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C02", "class_id": "C02", "class_label": "Git history, recovery, merging and deployment", "fine_family": "git_receive_hook_deployment", "classification_reason": "Configures authenticated Git pushes to deploy two branches through receive hooks to distinct HTTPS paths within a latency bound.", "transfer_tags": [ "Git", "SSH", "post-receive", "Nginx", "TLS", "branches" ], "split": "test" }, { "task_id": "gpt2-codegolf", "task_version_id": "49e98f14-36d2-428c-b505-d0e98e7ae028", "task_revision": 3, "source_path": "prepared/tasks/gpt2-codegolf", "instruction_sha256": "1c8de5f0365407d57fbec968bbeb76cba2648a02fbb2045d323d2798ff5dbccc", "task_config_sha256": "a3fa186d438a42af5909384b8dc0dfdc2ab7a3a68af9159e9d00464988895af0", "feature_text_sha256": "f1626083aa6da9cb6a2cec875d487e69d9dce2fc76dd39f4584959ca5bc25e96", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/gpt2-codegolf:20251031", "cpus": 1, "memory_mb": 8192, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "compact_gpt2_inference", "classification_reason": "Loads GPT-2 parameters and tokenization data into a very small dependency-free C inference program with deterministic next-token behavior.", "transfer_tags": [ "GPT-2", "C", "TensorFlow-checkpoint", "BPE", "dependency-free", "size-bound" ], "split": "test" }, { "task_id": "headless-terminal", "task_version_id": "ee251c1a-e512-4001-a7a0-fb9fb6137226", "task_revision": 3, "source_path": "prepared/tasks/headless-terminal", "instruction_sha256": "004ac3efa98dd4c564f3888fe199962772d23de405c96ec9232c1eab1a2089e8", "task_config_sha256": "abe0de1ac3adff1fbf0e22e1765efbc061a412891fd255926ed92a5ccacf69e6", "feature_text_sha256": "f26b085af0625ae353c995181411eb00edf86a1511fb9afe3db169fcd0d99baa", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/headless-terminal:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C04", "class_id": "C04", "class_label": "Runtime control, emulation and executable interfaces", "fine_family": "process_lifecycle_and_cleanup", "classification_reason": "Implements an interactive terminal interface supporting startup files, interactive programs and control keys, requiring process and signal handling.", "transfer_tags": [ "Python", "PTY", "interactive-shell", "signals", "terminal-I-O" ], "split": "test" }, { "task_id": "hf-model-inference", "task_version_id": "8502e6ff-d454-47b7-a356-e8fcc567dddb", "task_revision": 4, "source_path": "prepared/tasks/hf-model-inference", "instruction_sha256": "29151c5b2306f522f38fa905de3d34566865829355e5ca0895f4fca04d51dcc1", "task_config_sha256": "951ae476f4d078a25b7a487430148971e3a2222d93c4e0a932442440807eb3b7", "feature_text_sha256": "3e964c35bfa5f5b548044233ac603ca330e384a254dbf1d2ee402d47c6f73a96", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/hf-model-inference:20260430", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "http_sentiment_inference_service", "classification_reason": "Downloads and stores a pretrained sentiment model and exposes its inference with a precise JSON/error API as a persistent local service.", "transfer_tags": [ "Hugging-Face", "transformers", "Flask", "inference-service", "JSON" ], "split": "training" }, { "task_id": "install-windows-3.11", "task_version_id": "9e18f965-85a3-415e-aab6-8678ee81dc42", "task_revision": 2, "source_path": "prepared/tasks/install-windows-3.11", "instruction_sha256": "6757fff2157378201478ef17ff3da4d4992aa586fa577dd518c63b98bb1945c5", "task_config_sha256": "0485184bd16cc74883396115d93520bd25b36d23d61e248da88c0dceeab2eeac", "feature_text_sha256": "124fb1033cb26105391e5c06c405f5225c12589ab4312dce13dc189cf868632d", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/install-windows-3.11:20251031", "cpus": 2, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C04", "class_id": "C04", "class_label": "Runtime control, emulation and executable interfaces", "fine_family": "qemu_guest_boot_and_access", "classification_reason": "Boots a legacy guest to its desktop while preserving the disk image and exposing VNC and programmatic input interfaces.", "transfer_tags": [ "QEMU", "guest-boot", "VNC", "monitor-socket", "snapshot" ], "split": "training" }, { "task_id": "kv-store-grpc", "task_version_id": "1395047b-7b93-4f48-af43-95ebd878cb25", "task_revision": 4, "source_path": "prepared/tasks/kv-store-grpc", "instruction_sha256": "f4b3facc199d5b1c78db6e603ffa29ec1b93a77ff7d291e1a92b42e6bf582bed", "task_config_sha256": "6ff0f30a526bfff86a67b0b53439b9965229c0f368167dc1a5d05f34d82299a3", "feature_text_sha256": "9bf3ca9fc6ab2eb2275b760fc247bc8d2738e81d6e6dec38ebd5f24420b29137", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/kv-store-grpc:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C03", "class_id": "C03", "class_label": "Persistent network services and protocol configuration", "fine_family": "application_protocol_service", "classification_reason": "Defines protobuf messages, generates bindings and implements a persistent typed key/value RPC service on the specified port.", "transfer_tags": [ "gRPC", "protobuf", "Python", "request-response", "background-service" ], "split": "training" }, { "task_id": "large-scale-text-editing", "task_version_id": "ad6a61e4-f8cb-4201-923d-7c6e5c22e27e", "task_revision": 3, "source_path": "prepared/tasks/large-scale-text-editing", "instruction_sha256": "2746a16369b28b2e368ca168bc85222811cbe570f3b315f1b1c432afad07ddc8", "task_config_sha256": "5c9488e80c1186f72efb8193e6be13c74eadf334b3846db2e794d24513a88f5f", "feature_text_sha256": "954cdc7d8eacb9b484b1e0be482dc3acccdac8da5ac7c367a538fa03ed91a710", "official_agent_timeout_seconds": 1200.0, "official_verifier_timeout_seconds": 1200.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/large-scale-text-editing:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C09", "class_id": "C09", "class_label": "Structured data transformations, queries and text extraction", "fine_family": "constrained_text_transformation", "classification_reason": "Transforms a million-row CSV through strictly limited Vim macros, requiring exact output and compliance with permitted editing operations.", "transfer_tags": [ "CSV", "Vim", "macros", "regex", "keystroke-bound" ], "split": "test" }, { "task_id": "largest-eigenval", "task_version_id": "aa1290ac-52f7-4916-9a3b-276c5e835f50", "task_revision": 3, "source_path": "prepared/tasks/largest-eigenval", "instruction_sha256": "6fa938389f99648508e5a9772e0d5e95d1102738c68d1ff61bd65469d76c52ab", "task_config_sha256": "f7b9b7dda446f969a3ce88f61039d216f4d58b5fc5c10ff9faa24e2403bbd830", "feature_text_sha256": "bef432b898b2d64761513bf22bda6f422958a10b1d610463268eeaea54856fcd", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/largest-eigenval:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "numerical_kernel_optimization", "classification_reason": "Optimizes a small-matrix dominant-eigenpair function while preserving residual correctness for nonsymmetric and complex-valued cases.", "transfer_tags": [ "NumPy", "eigenpairs", "complex-values", "correctness", "benchmarking" ], "split": "training" }, { "task_id": "llm-inference-batching-scheduler", "task_version_id": "3a3827d6-8706-4bcf-a2cb-0339e7351f18", "task_revision": 3, "source_path": "prepared/tasks/llm-inference-batching-scheduler", "instruction_sha256": "0817374767533fbf51bb93345ee1bba266263c216c6cc55c096e71b4a1d37dfc", "task_config_sha256": "52aaac942f917e6a91ff658e0272195bd1cd842d41d71e302414c52603cc588b", "feature_text_sha256": "ffabde304d47a507c71bdd4645127c96d6c5e9a0788fff204ec619ec6de727e5", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/llm-inference-batching-scheduler:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "shape_aware_batch_optimization", "classification_reason": "Constructs request batches and aligned shapes against a supplied cost model while satisfying coverage, shape-count and performance constraints.", "transfer_tags": [ "batch-packing", "tensor-shapes", "cost-model", "latency", "constraint-optimization" ], "split": "training" }, { "task_id": "log-summary-date-ranges", "task_version_id": "1ec0328c-10b7-4225-9aa8-842c1303f990", "task_revision": 3, "source_path": "prepared/tasks/log-summary-date-ranges", "instruction_sha256": "c1d4516ad4ec2238209740eaf01291df9ab48b0028a3fc5df383c6f831453484", "task_config_sha256": "9a9593673d6805ef3a301ea734e9c4dbaf80da1b37c0fe9d1f2ceea2a8c57e18", "feature_text_sha256": "792ee9a0766625a8ddf15d180744d51a8ed19230d1a5dcb9478718c815244f32", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/log-summary-date-ranges:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C09", "class_id": "C09", "class_label": "Structured data transformations, queries and text extraction", "fine_family": "date_record_constraints", "classification_reason": "Extracts severity events from dated log files and computes several overlapping time-window counts in an exact CSV schema.", "transfer_tags": [ "logs", "date-time", "aggregation", "CSV" ], "split": "training" }, { "task_id": "mailman", "task_version_id": "ee00de2a-df1b-43aa-a5a1-fa95def7d666", "task_revision": 4, "source_path": "prepared/tasks/mailman", "instruction_sha256": "240d03a73e38087d40c55460d48a098c70143e911a279cffe2468f642322cdee", "task_config_sha256": "a04f4d0e9be25056d2cd79d6a018bc875b6e6fb51d429a9241f282272cc660d1", "feature_text_sha256": "d7c386197779b496d0f73fdecce148e94f388f42fecb4e11d7a38e40d600a8d3", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/mailman:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C03", "class_id": "C03", "class_label": "Persistent network services and protocol configuration", "fine_family": "daemon_configuration_and_routing", "classification_reason": "Configures interacting mail daemons and subscription policies, then validates local delivery and list lifecycle behavior.", "transfer_tags": [ "Postfix", "Mailman", "mail-routing", "subscriptions", "background-service" ], "split": "test" }, { "task_id": "make-doom-for-mips", "task_version_id": "60926372-2024-4a1e-b98c-eba04e39b429", "task_revision": 3, "source_path": "prepared/tasks/make-doom-for-mips", "instruction_sha256": "1903b660e4c64966ff257f828c932adff2b289de589bd6aee22b933bb4fef429", "task_config_sha256": "a7d7a498f0d3a992b384e5086ba2b411767c1c1a013d1b41f6a25cdaf6d05acc", "feature_text_sha256": "4e834ec93e6e309419945990bac02c597c162b56be3603956f3114084d478893", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/make-doom-for-mips:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C04", "class_id": "C04", "class_label": "Runtime control, emulation and executable interfaces", "fine_family": "mips_elf_execution_interface", "classification_reason": "Builds native C game sources into the executable format expected by a supplied JavaScript MIPS runtime and verifies frame/file output.", "transfer_tags": [ "MIPS", "ELF", "cross-compilation", "C", "JavaScript", "system-calls" ], "split": "training" }, { "task_id": "make-mips-interpreter", "task_version_id": "885f84a4-86b4-4585-b48b-fb541197f59b", "task_revision": 4, "source_path": "prepared/tasks/make-mips-interpreter", "instruction_sha256": "d25c0ed6d4697c2b2e6fe047b6cd7317f640dc84d7457ccc9d3d1b4cf8021ffd", "task_config_sha256": "3fda43066642fd38516d07b2679f9813bdb3f1a42bc80b0c06151f60b9210e82", "feature_text_sha256": "e5e8aabfe50a7764720b4f33da196409208a1fcc675b0cd9b45a69513bd56e93", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/make-mips-interpreter:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C04", "class_id": "C04", "class_label": "Runtime control, emulation and executable interfaces", "fine_family": "mips_elf_execution_interface", "classification_reason": "Implements a JavaScript MIPS interpreter and system-call handling sufficient to run a supplied ELF and produce expected frame files.", "transfer_tags": [ "MIPS", "ELF", "JavaScript", "emulation", "system-calls", "file-I-O" ], "split": "test" }, { "task_id": "mcmc-sampling-stan", "task_version_id": "49e5d4f7-0729-4874-b753-c9d0bdcbda2f", "task_revision": 3, "source_path": "prepared/tasks/mcmc-sampling-stan", "instruction_sha256": "5554ddc6d7b069e965eda4936cd02fce45ed54f51cf38045e6c51aeb39bed943", "task_config_sha256": "c1c0351670499e15b3818f7394894e2789f74fc2a61ea76eccab826c1a951a94", "feature_text_sha256": "08e86ef73c6fec66b001653a22480226e85823f3a9e2a0205685e104c7615f5b", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/mcmc-sampling-stan:20251031", "cpus": 4, "memory_mb": 8192, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "stan_posterior_workflows", "classification_reason": "Specifies and samples a hierarchical Bayesian model, preserves requested sampling settings and exports posterior parameter means.", "transfer_tags": [ "RStan", "Stan", "Bayesian-model", "MCMC", "posterior-validation" ], "split": "training" }, { "task_id": "merge-diff-arc-agi-task", "task_version_id": "3d1fc776-448c-4598-ad75-ae845c905232", "task_revision": 3, "source_path": "prepared/tasks/merge-diff-arc-agi-task", "instruction_sha256": "2d826479873e5ac1c94c1f802763bf85c22bb04624384685ebdf6b153fe0278e", "task_config_sha256": "6ced40e06b43b169857509765eb7f07791db558f9437cc7367d2536771c2ca5b", "feature_text_sha256": "5e828fb20f9ec704350ccf96fe471209557b90124f38de4c176fd473845666f5", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/merge-diff-arc-agi-task:20251031", "cpus": 1, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C02", "class_id": "C02", "class_label": "Git history, recovery, merging and deployment", "fine_family": "git_branch_merge_and_repair", "classification_reason": "Restores branches from bundles, resolves their merge and repairs the merged mapping code to satisfy public examples.", "transfer_tags": [ "Git", "bundles", "merge-conflicts", "Python", "example-validation" ], "split": "training" }, { "task_id": "model-extraction-relu-logits", "task_version_id": "1fe5bc51-82fa-4069-bfef-4afca46beaa6", "task_revision": 3, "source_path": "prepared/tasks/model-extraction-relu-logits", "instruction_sha256": "c9e179dd099065a34c362a865b0eb9fab5f7d96be582cf117e7519580c036ebe", "task_config_sha256": "eb64dfaf8ae88868911ae28855cdc4f767de464001e2704919015b595be6dc29", "feature_text_sha256": "800a080ceb3b587f436aa3f87a43b159204af3485512b7a5855dd52768aa8e05", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/model-extraction-relu-logits:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "piecewise_linear_parameter_recovery", "classification_reason": "Recovers first-layer parameters from numerical queries to a piecewise-linear model and validates an equivalent matrix up to permitted symmetries.", "transfer_tags": [ "ReLU", "black-box-queries", "linear-algebra", "parameter-identification" ], "split": "test" }, { "task_id": "modernize-scientific-stack", "task_version_id": "d7dc0872-879c-4a30-a8c7-a195f3be8dda", "task_revision": 4, "source_path": "prepared/tasks/modernize-scientific-stack", "instruction_sha256": "666b8ac370ae81b0e3780a36e722c91b0f986fccbdc6ed36ab3bd17e6943da0f", "task_config_sha256": "a3029f3d4caaa1f08bd1b45488247bf8cb586bf1ce69717acf1114d20f51eb48", "feature_text_sha256": "5a729f66d750945a553f0ab08666bb5e42f10a4acf89d677b2212df71fc0bc6e", "official_agent_timeout_seconds": 600.0, "official_verifier_timeout_seconds": 600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/modernize-scientific-stack:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C01", "class_id": "C01", "class_label": "Source builds, compatibility and code repair", "fine_family": "python_dependency_compatibility", "classification_reason": "Updates legacy Python scientific code and dependencies to current APIs while preserving specified data processing and output behavior.", "transfer_tags": [ "Python-migration", "pandas", "NumPy", "pathlib", "dependency-pinning" ], "split": "training" }, { "task_id": "mteb-leaderboard", "task_version_id": "39bb08fd-5f95-4017-830b-2b0ecd66ce42", "task_revision": 4, "source_path": "prepared/tasks/mteb-leaderboard", "instruction_sha256": "06ea44106eca0405c25bab00f5b3cfeb611c872093f6c0f82ed057839d1d13de", "task_config_sha256": "86574bd1c6c4fefd2e64eee9c2caa53d87c4bcf2e51547912440a0ddbd0badb8", "feature_text_sha256": "3080cedb107f9ccf36d6c0fd0b2140139dbf39a7d4cc900137c6ab66ffbca93a", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/mteb-leaderboard:20260430", "cpus": 1, "memory_mb": 8192, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "embedding_benchmark_workflows", "classification_reason": "Selects an embedding model from dated benchmark evidence with complete task coverage; related embedding tasks share model identity and benchmark-library context.", "transfer_tags": [ "MTEB", "embedding-model", "model-metadata", "benchmark-coverage" ], "split": "test" }, { "task_id": "mteb-retrieve", "task_version_id": "54400505-4f5d-4089-bfbc-a2460d6467fb", "task_revision": 4, "source_path": "prepared/tasks/mteb-retrieve", "instruction_sha256": "071305482e6989fe84634227be9a19386a31b3c2d1fc9fb0d3e604490b3f1dca", "task_config_sha256": "a2a4ded9824d296ce99bfba6aa2c9341dcc58f10e66de2e8f95dc8e07379c3aa", "feature_text_sha256": "16631a3f980c38fc99f127c470925d823bb3031787ae88b66a5d02549142e8be", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/mteb-retrieve:20260430", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "embedding_benchmark_workflows", "classification_reason": "Loads a revision-pinned embedding model through MTEB, encodes texts and returns an exact ranked similarity result.", "transfer_tags": [ "MTEB", "embedding-model", "revision-pinning", "cosine-similarity", "ranking" ], "split": "training" }, { "task_id": "multi-source-data-merger", "task_version_id": "bc4232a1-f6fd-486c-8cec-f23bcf2f92ed", "task_revision": 3, "source_path": "prepared/tasks/multi-source-data-merger", "instruction_sha256": "d1d28bb6e37f7a255a1b7f78aae8591f35c0494f33e4913392424ad954aab944", "task_config_sha256": "ee3964c0ee040f0a0cb121cefefbc15443d127684acb003fa5b61dd42f31c694", "feature_text_sha256": "8af5a106a9cc998d078c2a3015d4b43441ed968f7ec7c4077bd2fccfbb58c5ee", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/multi-source-data-merger:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C09", "class_id": "C09", "class_label": "Structured data transformations, queries and text extraction", "fine_family": "schema_merge_and_aggregation", "classification_reason": "Normalizes schemas across three record formats, merges by key with fixed precedence and emits reconciled data plus a conflict ledger.", "transfer_tags": [ "JSON", "CSV", "Parquet", "schema-mapping", "priority-merge" ], "split": "training" }, { "task_id": "nginx-request-logging", "task_version_id": "5f79157e-598a-430b-9443-e0da84e04500", "task_revision": 3, "source_path": "prepared/tasks/nginx-request-logging", "instruction_sha256": "26a7ac98aced6107f147206790ada77bbf8ed3c25fb360cf323d5fc6889edf99", "task_config_sha256": "b7a78d00be13656f75069197100b7b8c4fd268decf6f78929c853b3222a2729d", "feature_text_sha256": "40a1b5c2ab5c16975b5deb3c7a8f01b0b32aa8ed7ab14650f83be832e0657aa6", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/nginx-request-logging:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C03", "class_id": "C03", "class_label": "Persistent network services and protocol configuration", "fine_family": "daemon_configuration_and_routing", "classification_reason": "Configures an HTTP daemon with exact routes, structured logs, errors and rate limits, checking syntax and live endpoint behavior.", "transfer_tags": [ "Nginx", "HTTP", "logging", "rate-limits", "configuration-validation" ], "split": "training" }, { "task_id": "openssl-selfsigned-cert", "task_version_id": "1db45b51-7683-4593-9dd7-28714ebc3ab7", "task_revision": 3, "source_path": "prepared/tasks/openssl-selfsigned-cert", "instruction_sha256": "38054a3371f7c2c853478a2f9af0bff858437d5b28a9a07483238f06e566921f", "task_config_sha256": "75c8913f0fa15890786c11999cde920b8bfc10d4ce8a14edf1e76985a2ac5800", "feature_text_sha256": "bfc58841c194c415cb648910e2ba31b11348f8adde00789d4a495eb0ec18c669", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/openssl-selfsigned-cert:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C03", "class_id": "C03", "class_label": "Persistent network services and protocol configuration", "fine_family": "tls_service_artifacts", "classification_reason": "Creates and verifies the key/certificate artifacts used by TLS services, including required formats, metadata and file permissions.", "transfer_tags": [ "OpenSSL", "TLS", "PEM", "permissions", "certificate-validation" ], "split": "training" }, { "task_id": "overfull-hbox", "task_version_id": "76061d58-0df0-478c-860e-8c1c04d5de99", "task_revision": 3, "source_path": "prepared/tasks/overfull-hbox", "instruction_sha256": "91381d7490afbbf11255e22471d3ed76317353d8e1da372fbc91887e65c57f47", "task_config_sha256": "64680925bd47043216bffb2c44eb3620f1730e111a89ffbcc5550e5ffc89e72b", "feature_text_sha256": "ee81bfce621a10c7bf99a3e52e04098804c8196a6837a2aa339c000486d430db", "official_agent_timeout_seconds": 750.0, "official_verifier_timeout_seconds": 360.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/overfull-hbox:20260403", "cpus": 2, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C13", "class_id": "C13", "class_label": "Document and media interpretation or transformation", "fine_family": "tex_layout_diagnostics", "classification_reason": "Repairs document layout by restricted word substitutions and repeated TeX diagnostics; it is document-content/layout work rather than arbitrary compiler repair.", "transfer_tags": [ "LaTeX", "text-layout", "allowed-substitutions", "compiler-diagnostics" ], "split": "training" }, { "task_id": "password-recovery", "task_version_id": "8e22f580-ca02-4301-9354-d7e66816847e", "task_revision": 3, "source_path": "prepared/tasks/password-recovery", "instruction_sha256": "bc4f501148805bbccdb94a5c5b745f1ce9564f04ef0fe36bc37c3ca20b7810a5", "task_config_sha256": "cfd0362d201b957796f9bf1c02cc75646250cb2bd107ec88c768bfcfd091b24b", "feature_text_sha256": "13e408bd264c78ffb30ead6b2c15e9744290e84b9bf251e86ff286dfce2a14f3", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/password-recovery:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C06", "class_id": "C06", "class_label": "Cryptanalysis and secret recovery", "fine_family": "password_and_artifact_recovery", "classification_reason": "Recovers a password from deleted-file evidence using its stated structural constraints to validate candidates.", "transfer_tags": [ "deleted-file", "forensics", "candidate-filtering", "password-pattern" ], "split": "test" }, { "task_id": "path-tracing", "task_version_id": "2357c219-c288-4fb4-a77c-0ef8a5351ab3", "task_revision": 3, "source_path": "prepared/tasks/path-tracing", "instruction_sha256": "9f36e02d636f5dc0946619bc518f2e51c24fa75251da9d1f0ca9eca612cb9572", "task_config_sha256": "cf47013f69bf1b43242da984f231d4daf364fdf44d3845f2eea2ca2cb9308c46", "feature_text_sha256": "b6a544c140cb5d7bb189fee1a84f5a04771eaa5296274c7be1f481df5f4798d0", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/path-tracing:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C14", "class_id": "C14", "class_label": "Compact native program reconstruction", "fine_family": "compact_native_reconstruction", "classification_reason": "Infers a compact independent renderer from a reference image and validates visual agreement under a compressed-source size limit.", "transfer_tags": [ "C", "PPM", "rendering", "inverse-graphics", "static-binary", "size-bound" ], "split": "test" }, { "task_id": "path-tracing-reverse", "task_version_id": "b51cad68-469b-4ead-897c-01968ff2d284", "task_revision": 3, "source_path": "prepared/tasks/path-tracing-reverse", "instruction_sha256": "03e4ef347c1b6bde7ad020b297571cf22c74cfb0434b399b267f3d87462e1176", "task_config_sha256": "f609310c99bd5f379eea2711f4af8eca79b5bcd1178a034b2285b1171c031b5b", "feature_text_sha256": "5cca3e9faf7735db288728a2138dbacbff94653bc69efe09fd79b6f160d5c726", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/path-tracing-reverse:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C14", "class_id": "C14", "class_label": "Compact native program reconstruction", "fine_family": "compact_native_reconstruction", "classification_reason": "Reconstructs an unknown executable's behavior in a small independent static C program using behavioral or binary reference evidence; its public instructions do not reveal the operation.", "transfer_tags": [ "C", "static-binary", "reverse-engineering", "behavioral-probing", "size-bound" ], "split": "training" }, { "task_id": "polyglot-c-py", "task_version_id": "a0134323-ad2b-4c55-b9d8-87b90594ddcf", "task_revision": 3, "source_path": "prepared/tasks/polyglot-c-py", "instruction_sha256": "a57c0c0e1232113113cd29a154ea7e0ac9ea45e4aedc16958fbc6bc753a4ef96", "task_config_sha256": "d31b9f60c47eed0d79e10d3be8fd56a23e50b3986606e3f2546863d44ac9a9ea", "feature_text_sha256": "3f8fc6cfc00578322bec22212b520b97f7a7345e700e27f2e967c329fd55631e", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/polyglot-c-py:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C12", "class_id": "C12", "class_label": "Constrained language and program synthesis", "fine_family": "polyglot_source_construction", "classification_reason": "Constructs one source file valid in both C and Python with exact command-line behavior, exploiting overlapping parsing rules.", "transfer_tags": [ "C", "Python", "parser", "preprocessor", "Fibonacci" ], "split": "training" }, { "task_id": "polyglot-rust-c", "task_version_id": "bf750a9b-299d-4da5-87a7-ca6df659a0e5", "task_revision": 3, "source_path": "prepared/tasks/polyglot-rust-c", "instruction_sha256": "b0fbcd99a97fc886e4e198b6189b977409dada47cae0a87ac7dc185b440b2355", "task_config_sha256": "3c41e83437068c62a3fdff74641899146e34907aefd9e47346213f47caeb82de", "feature_text_sha256": "80d0fa024eab2af1a06ed8de6d062c166484bf43ba2ac0d7b7cc11492fc2b41e", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/polyglot-rust-c:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C12", "class_id": "C12", "class_label": "Constrained language and program synthesis", "fine_family": "polyglot_source_construction", "classification_reason": "Constructs one source file accepted by both Rust and C++ toolchains, sharing cross-parser design while respecting its own output convention.", "transfer_tags": [ "Rust", "C++", "parser", "preprocessor", "Fibonacci" ], "split": "test" }, { "task_id": "portfolio-optimization", "task_version_id": "ffd84731-10d7-4d1a-a2e3-182652ebffdc", "task_revision": 4, "source_path": "prepared/tasks/portfolio-optimization", "instruction_sha256": "ffaadcbf9f576b5d5e087d2a06a4b525dd52c6b84d43eb9d85107f0a9aebcb5d", "task_config_sha256": "4918687e4e261a9852e39e5a554156733c071c17dc89c5f544f15f2ef744ce0a", "feature_text_sha256": "8d767faf96f6de8cbd0062ac2d858b39f3eaf0ae19081634f943d5256ddc2648", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/portfolio-optimization:20251031", "cpus": 1, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "numerical_kernel_optimization", "classification_reason": "Implements fast portfolio return/risk kernels through a C/Python interface while matching reference numerical outputs and speed constraints.", "transfer_tags": [ "C-extension", "Python", "linear-algebra", "benchmarking", "numerical-equivalence" ], "split": "test" }, { "task_id": "protein-assembly", "task_version_id": "cbd36098-10d4-469b-ac96-56916d15fb7e", "task_revision": 3, "source_path": "prepared/tasks/protein-assembly", "instruction_sha256": "1bb2af1d85e9a823a8ae7273a87df8a3f67e2c93b06cddf34f0051a8b003f7c5", "task_config_sha256": "9b91f32925035fba7b5a1db16c7d4ed650a117af145d3d3f8e514cb4b1e970ab", "feature_text_sha256": "023e9adb6316b62c4ff767a5b10710faf0ecfb81b330c47589b0c0beb8a51cda", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/protein-assembly:20260403", "cpus": 1, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C08", "class_id": "C08", "class_label": "Molecular sequence and construct design", "fine_family": "protein_sequence_construct_design", "classification_reason": "Identifies protein components, assembles a requested fusion and designs a coding sequence under length and sliding-window GC constraints.", "transfer_tags": [ "protein-sequence", "PDB", "FPbase", "reverse-translation", "GC-constraint" ], "split": "training" }, { "task_id": "prove-plus-comm", "task_version_id": "b8d783c8-573e-4a6b-890f-d20a14d74efd", "task_revision": 3, "source_path": "prepared/tasks/prove-plus-comm", "instruction_sha256": "7618ac930d942f27376189fe6535d0e69150fa3eba2db24759ae86b2675d21fe", "task_config_sha256": "240f3ac0076e3362685c8598c4e56658e3882eb25a14ae6cc7e9ab3a4f6b09ea", "feature_text_sha256": "1693433408e2b8dabb64bf490123c15ace60b6b28af580825f4307f14af334c5", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/prove-plus-comm:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C12", "class_id": "C12", "class_label": "Constrained language and program synthesis", "fine_family": "formal_proof_construction", "classification_reason": "Completes an inductive arithmetic proof and checks it with the Coq compiler; proof-language semantics determine correctness.", "transfer_tags": [ "Coq", "induction", "tactics", "proof-checking" ], "split": "test" }, { "task_id": "pypi-server", "task_version_id": "1799e27b-fa53-4531-a1d7-840ec9b6f546", "task_revision": 3, "source_path": "prepared/tasks/pypi-server", "instruction_sha256": "1c537e4a311a8e71155e121f57bfc03f742fbfb99d05314cd03e433996a28a31", "task_config_sha256": "07c39d7a733c89adadf74bb63a3e906c3694feb6c490d5fc85ac187c003fd447", "feature_text_sha256": "70bdbc2e43c21fa8522dd505dffac53be01fed37d4a2062c159d67df21152fd7", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/pypi-server:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C03", "class_id": "C03", "class_label": "Persistent network services and protocol configuration", "fine_family": "application_protocol_service", "classification_reason": "Builds a Python distribution and exposes a pip-compatible package index, checking installation and import through the service protocol.", "transfer_tags": [ "Python-package", "build-artifact", "simple-index", "HTTP", "pip" ], "split": "test" }, { "task_id": "pytorch-model-cli", "task_version_id": "a045cc71-cadb-4524-b0ed-b5c20146f1d9", "task_revision": 3, "source_path": "prepared/tasks/pytorch-model-cli", "instruction_sha256": "ce36ee1f6c589f81ee6a6be4967e25c76a2ed697006928023f42dfb9f28a6750", "task_config_sha256": "fc1c018efbd5d0537ac0741d53d1ebad3886faa51d14cbde8ec953856411bc0d", "feature_text_sha256": "e8bb96055754290736d894d62c2f9cb74481fb1542c6adf782f9ab5af3765a1a", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/pytorch-model-cli:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "pytorch_parameter_reconstruction_and_inference", "classification_reason": "Exports or loads model parameters and packages image inference behind a strict command-line/output-file contract.", "transfer_tags": [ "PyTorch", "JSON-weights", "image-preprocessing", "CLI", "inference" ], "split": "test" }, { "task_id": "pytorch-model-recovery", "task_version_id": "3a7507ce-0733-4b82-9083-fad027d5cdb0", "task_revision": 4, "source_path": "prepared/tasks/pytorch-model-recovery", "instruction_sha256": "2bef07dc28be3ab2c35ae389584034c2e4dde92c2c31b3eb6bd829628c8c3f4a", "task_config_sha256": "851a51204f53482a31f47cdf8833128cf421113b0331011286940fc857c37dfd", "feature_text_sha256": "adc36c7c62264b823ec924c37d71e22e6e3a586a5b93f2d6c1ec431eaf298291", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/pytorch-model-recovery:20260430", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "pytorch_parameter_reconstruction_and_inference", "classification_reason": "Infers model structure from parameter tensors, repairs only permitted output-layer weights and exports working TorchScript inference.", "transfer_tags": [ "PyTorch", "state-dict", "architecture-inference", "restricted-fitting", "TorchScript" ], "split": "training" }, { "task_id": "qemu-alpine-ssh", "task_version_id": "1d359085-6d05-4de6-87c0-b7c668cef8f4", "task_revision": 3, "source_path": "prepared/tasks/qemu-alpine-ssh", "instruction_sha256": "8348c60490cb90bc690cba259dfe98996da4c3294c6d42bba1119b00c64de4de", "task_config_sha256": "10a43ec6ee732eeba852c377f12e23178f7fa5a4868143f3699a15565ffb64c7", "feature_text_sha256": "cb2e1f2276e7e216303a1cbe05e0ad74b2f41d34b9ab5af3cd3526561b31d5f1", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/qemu-alpine-ssh:20251031", "cpus": 1, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C04", "class_id": "C04", "class_label": "Runtime control, emulation and executable interfaces", "fine_family": "qemu_guest_boot_and_access", "classification_reason": "Boots a Linux guest, configures its SSH service and authentication, and verifies host-to-guest forwarded access.", "transfer_tags": [ "QEMU", "Alpine", "guest-SSH", "port-forwarding", "readiness" ], "split": "test" }, { "task_id": "qemu-startup", "task_version_id": "7ee8d5aa-08ca-461c-a550-70d46d0c4e44", "task_revision": 3, "source_path": "prepared/tasks/qemu-startup", "instruction_sha256": "7917cf2746328057b85d399c794bbf8bf328b2bd90502eaa3287170e25e6267b", "task_config_sha256": "dc46292fff83fba50192f4c30606a5f9c092a9fb82139686df7dabf6580937d4", "feature_text_sha256": "dbacff4c758ba18871a5979542b22014c5842d03811ce36aa91a9b8067d54c77", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/qemu-startup:20251031", "cpus": 1, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C04", "class_id": "C04", "class_label": "Runtime control, emulation and executable interfaces", "fine_family": "qemu_guest_boot_and_access", "classification_reason": "Boots an ISO as a persistent VM and exposes a ready guest console over the requested network interface.", "transfer_tags": [ "QEMU", "ISO", "serial-console", "telnet", "readiness" ], "split": "training" }, { "task_id": "query-optimize", "task_version_id": "94e5184d-6e52-417b-b72b-ca041fcd36f1", "task_revision": 4, "source_path": "prepared/tasks/query-optimize", "instruction_sha256": "473e12d320d2732f679fb631c2a4dab4a01deea6775686718ae0c417b8a39447", "task_config_sha256": "6e88f2a86df699594bcb1bf81df4ed694505f79c9420f671335b02524bbe3868", "feature_text_sha256": "2e6a0f3c18762b7c4e3f48053e224f907211a9752ad193a7bf65536380d67d7a", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/query-optimize:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C09", "class_id": "C09", "class_label": "Structured data transformations, queries and text extraction", "fine_family": "declarative_data_queries", "classification_reason": "Inspects and rewrites a database query for equivalent output and better performance while leaving the database unchanged.", "transfer_tags": [ "SQLite", "SQL", "joins", "query-plan", "equivalent-results" ], "split": "training" }, { "task_id": "raman-fitting", "task_version_id": "743258cb-0a8d-41b9-8da3-a787e59fae56", "task_revision": 3, "source_path": "prepared/tasks/raman-fitting", "instruction_sha256": "1fa59a89d88daf6a4b2d422955f53779b461880b527f40d79c71b16e0ee52433", "task_config_sha256": "0f3f6b580d7435954b3a114e77cdfb6cd0a856f15cc0ad83cd1b69c48492dfed", "feature_text_sha256": "412282ebd49e152cac31fcbc03aab30d861a9a37aa0f5bf56a268ce187454b49", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/raman-fitting:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "spectral_peak_fitting", "classification_reason": "Fits a parametric spectral model and exports the recovered peak and background parameters in a prescribed schema.", "transfer_tags": [ "spectral-data", "peak-fitting", "optimization", "parameter-estimation", "JSON" ], "split": "training" }, { "task_id": "regex-chess", "task_version_id": "1e49a9a6-9f58-4d27-ae7b-1cc6839ace80", "task_revision": 5, "source_path": "prepared/tasks/regex-chess", "instruction_sha256": "a6ed3fcc54fc92d71a84af921942422d299ca89b410dcd92cd05a53d0d06436c", "task_config_sha256": "07fe52636fcc4bc9fff25d5630af145b97a5331219d719cc8decb4ca67ea21d4", "feature_text_sha256": "ee84017c42869cd785a0f3782f1a503a226c3fae685ef8576ed6f9e9c0ea6f51", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/regex-chess:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C12", "class_id": "C12", "class_label": "Constrained language and program synthesis", "fine_family": "regex_rewrite_move_generation", "classification_reason": "Encodes legal move generation as a bounded ordered regex-replacement program; the core task is restricted computation rather than visual chess analysis.", "transfer_tags": [ "regex", "rewrite-program", "chess-rules", "JSON", "resource-bound" ], "split": "training" }, { "task_id": "regex-log", "task_version_id": "3463862e-e1c1-4baa-979f-508fb7cc155c", "task_revision": 3, "source_path": "prepared/tasks/regex-log", "instruction_sha256": "4f7ac05e70cf9220ea0f1e5a052c5f908cd0fa884e847d80b0bd51bae2e96f9c", "task_config_sha256": "6bc98525852755caf1fe34c174c2f10f4328adb4ce6c1968dcafe1c877e4e160", "feature_text_sha256": "b916147847e66e5fa0bfaa73f75a3e80080c781e5e8ba9b30f2618a8ab829730", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/regex-log:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C09", "class_id": "C09", "class_label": "Structured data transformations, queries and text extraction", "fine_family": "constrained_text_transformation", "classification_reason": "Builds a regex extracting the last valid date from qualifying log lines with exact date/address and token-boundary rules.", "transfer_tags": [ "Python-regex", "date-validation", "IPv4", "boundaries", "multiline" ], "split": "training" }, { "task_id": "reshard-c4-data", "task_version_id": "86d62c2f-c41c-4123-84a1-31b6ad6c1b4f", "task_revision": 4, "source_path": "prepared/tasks/reshard-c4-data", "instruction_sha256": "1810abbf666e51e410950b4f495d3dae88d392ad056afa392e4c687ff31e3992", "task_config_sha256": "c2f146a1e00f798f7486ff60686ad52749bb73c7b71e984d52716d859e0f7dd8", "feature_text_sha256": "64df7a41d7b8b6a56a60ec4f6c1eb340a17ad85fcf3bc97c3b2d66266845bef7", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/reshard-c4-data:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C09", "class_id": "C09", "class_label": "Structured data transformations, queries and text extraction", "fine_family": "nlp_corpus_processing", "classification_reason": "Repackages a text corpus under file-size and directory-fanout limits while preserving exact reconstructability and a fixed CLI/environment contract.", "transfer_tags": [ "dataset-sharding", "filesystem", "streaming", "reversible-transform", "uv" ], "split": "test" }, { "task_id": "rstan-to-pystan", "task_version_id": "8786d2d6-2c11-4814-94a1-57d08661cf40", "task_revision": 3, "source_path": "prepared/tasks/rstan-to-pystan", "instruction_sha256": "7b6dd84eb47579469556915a5eab6d303217bba438424eaee4db52375d5add2a", "task_config_sha256": "ed0cd4382cb353e73cb575aa7a20b624e35f09ddd8769c332f40f8986980e13f", "feature_text_sha256": "c6225b0a8ba47dd38fd9f5527b8c0f06ebcf61e6e262a07a3c89e35d763238e1", "official_agent_timeout_seconds": 1800.0, "official_verifier_timeout_seconds": 1800.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/rstan-to-pystan:20251031", "cpus": 4, "memory_mb": 8192, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "stan_posterior_workflows", "classification_reason": "Ports a Stan sampling workflow to a mandated Python interface and reproduces its posterior summaries without the original R runtime.", "transfer_tags": [ "RStan", "PyStan", "Stan", "API-migration", "posterior-validation" ], "split": "test" }, { "task_id": "sam-cell-seg", "task_version_id": "51e6e507-8dda-4aeb-a917-2896ced442da", "task_revision": 3, "source_path": "prepared/tasks/sam-cell-seg", "instruction_sha256": "370629c8a1b22896e5bb85d13411a860c496d294e70d0ec7ab98c2911c922150", "task_config_sha256": "f0eb6af2acb87cb6fee67d77e8fdd55e3b331baf48999260bc5b6c4ac466c128", "feature_text_sha256": "ed54fc1eee0f3fe3cb1ed354e7d53ffff781a5d08da3df6f7ce4add54d4451fd", "official_agent_timeout_seconds": 7200.0, "official_verifier_timeout_seconds": 7200.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/sam-cell-seg:20251031", "cpus": 1, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "model_guided_image_processing", "classification_reason": "Integrates a pretrained segmentation model with mask prompts and deterministic geometric cleanup/serialization while preserving table fields.", "transfer_tags": [ "MobileSAM", "CPU", "segmentation", "masks", "connected-components", "polygons" ], "split": "training" }, { "task_id": "sanitize-git-repo", "task_version_id": "7ecc1f73-d336-4f7a-a78c-0dc61f5c4f44", "task_revision": 3, "source_path": "prepared/tasks/sanitize-git-repo", "instruction_sha256": "d9e10a56e94e51592c6016efb02adef852d4fd08f9ba23dc304d2673d1478e8a", "task_config_sha256": "ff209feb7d07c68b8bd0d2e9abee96a1431c332095a1bbe0a532c44a833a264d", "feature_text_sha256": "33b48c4a4c7efc02275bd820fd7ffb83691d919eeeb3761383dc6a518320a0d6", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/sanitize-git-repo:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C02", "class_id": "C02", "class_label": "Git history, recovery, merging and deployment", "fine_family": "git_history_recovery_and_hygiene", "classification_reason": "Locates and consistently replaces repository credentials while preserving uncontaminated contents; history rewriting is not assumed unless the task requires it.", "transfer_tags": [ "Git", "secret-scanning", "consistent-replacement", "content-preservation" ], "split": "training" }, { "task_id": "schemelike-metacircular-eval", "task_version_id": "06bf378d-e9ea-4636-8d80-ddfcb514efb1", "task_revision": 4, "source_path": "prepared/tasks/schemelike-metacircular-eval", "instruction_sha256": "3af5f9567b62f93683db7518ec6fcb609f83796258aa02166b5a7091c5b69031", "task_config_sha256": "8dae8eb226183b0ca92331c1106152a7abc93b03b4e5a95291b86a4ddc9b44e0", "feature_text_sha256": "bd8dbfee7bdefbca350fa9b2fe1a89ba53fce10672798f6f929750ee72bd4dfb", "official_agent_timeout_seconds": 2400.0, "official_verifier_timeout_seconds": 2400.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/schemelike-metacircular-eval:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C12", "class_id": "C12", "class_label": "Constrained language and program synthesis", "fine_family": "interpreter_semantics", "classification_reason": "Implements a self-interpreting evaluator matching an existing language's semantics and nested interpreter I/O behavior.", "transfer_tags": [ "Scheme", "evaluation", "self-hosting", "stdin-stdout", "language-semantics" ], "split": "training" }, { "task_id": "sparql-university", "task_version_id": "c785aab0-29c1-48b0-a63f-375a1de861ba", "task_revision": 3, "source_path": "prepared/tasks/sparql-university", "instruction_sha256": "22eeab47d3b27b06df2627b3c21cfbd579ed7dcc85ebcb7c4b3a241c3c35f449", "task_config_sha256": "47a6f6dd9ce0fc7631ef607e900fcd039b2d064ea90545645b49c678cddc9585", "feature_text_sha256": "da9289af7b72818acd77d077f1a044c47de4b2c38ce04d31ea157b2472bbabc9", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/sparql-university:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C09", "class_id": "C09", "class_label": "Structured data transformations, queries and text extraction", "fine_family": "declarative_data_queries", "classification_reason": "Inspects an RDF schema and writes a query combining joins, temporal predicates and correctly scoped aggregates.", "transfer_tags": [ "RDF", "SPARQL", "ontology", "joins", "aggregation", "date-filters" ], "split": "test" }, { "task_id": "sqlite-db-truncate", "task_version_id": "914a83de-4d7c-4402-b073-5acbfce95aa0", "task_revision": 3, "source_path": "prepared/tasks/sqlite-db-truncate", "instruction_sha256": "465c135cd80be49023ea3ecf8f1d840c679b3a960efe425d510894e417b9d802", "task_config_sha256": "56061acd75d9933aa554aeb249b85d7d081227fca1b71a3da7e3dc97a675723e", "feature_text_sha256": "93a53b92afbf075d717ce62b40f143320ab254462b1ff51850398e388f6a70ef", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/sqlite-db-truncate:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C10", "class_id": "C10", "class_label": "SQLite storage recovery", "fine_family": "sqlite_file_recovery", "classification_reason": "Recovers structured rows from a physically truncated SQLite file and exports them, sharing low-level storage inspection with WAL recovery.", "transfer_tags": [ "SQLite", "binary-pages", "truncation", "JSON", "row-recovery" ], "split": "training" }, { "task_id": "sqlite-with-gcov", "task_version_id": "bc4dbfd4-d93c-4bb6-83a6-af87351a7df8", "task_revision": 3, "source_path": "prepared/tasks/sqlite-with-gcov", "instruction_sha256": "b6ba9acf62292bc2a0bab40d4805b3abceededbb0ccd4477a1ce9ccc42c14dd9", "task_config_sha256": "9a0da96990ee2ecbf4d92682a4c5301a53db45fc6e2d46ac7d45d3b8ee7fe523", "feature_text_sha256": "502c7ac1632273a32caef449c70660ab68c1ba277813c005c3d88e0ec3522a4d", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/sqlite-with-gcov:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C01", "class_id": "C01", "class_label": "Source builds, compatibility and code repair", "fine_family": "native_source_build", "classification_reason": "Builds the supplied SQLite source with coverage instrumentation and installs the executable, using compiler/linker flags and artifact checks.", "transfer_tags": [ "SQLite", "C", "gcov", "instrumentation", "source-build" ], "split": "training" }, { "task_id": "torch-pipeline-parallelism", "task_version_id": "19263c2a-c2db-432d-b355-629f27c67786", "task_revision": 3, "source_path": "prepared/tasks/torch-pipeline-parallelism", "instruction_sha256": "989bfcf0c54baf6c8cdc391743dbb2f9e858c5779dd85112536875d4d1b2dcd2", "task_config_sha256": "8b4f9f525c08afbbc4eca9168e46238a61ccb903486773e9ca4da64bb7ffe857", "feature_text_sha256": "c56e88282f77860ec04b3ea6b0c5fa5e34a553ac018cb364153d79d2a15e4f23", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/torch-pipeline-parallelism:20251031", "cpus": 1, "memory_mb": 8192, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "distributed_tensor_training", "classification_reason": "Partitions a neural model into balanced stages and implements microbatch forward/backward communication with correct loss and gradient semantics.", "transfer_tags": [ "PyTorch", "LLaMA", "pipeline-stages", "microbatches", "distributed-autograd" ], "split": "test" }, { "task_id": "torch-tensor-parallelism", "task_version_id": "044b02da-4f7d-4301-98e5-5a04319955e3", "task_revision": 3, "source_path": "prepared/tasks/torch-tensor-parallelism", "instruction_sha256": "13bde2ab165d72a311e2bf066d5d6c3c8761bf9d3dd39d1b23eda23bc546b910", "task_config_sha256": "72fd2a5500d1fe8795dcae76cccb1a29fb25894881a7475f2fa519c87d425a8d", "feature_text_sha256": "366531970e97b6e674925bf09a2ed1b5dac946037f7886e3d88158571bf6fc03", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/torch-tensor-parallelism:20251031", "cpus": 1, "memory_mb": 8192, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "distributed_tensor_training", "classification_reason": "Implements tensor-sharded linear layers and their communication/backward behavior across distributed ranks.", "transfer_tags": [ "PyTorch", "sharding", "collectives", "row-column-linear", "distributed-autograd" ], "split": "training" }, { "task_id": "train-fasttext", "task_version_id": "6cee61f6-578c-47b4-b71b-fbf925918ec0", "task_revision": 3, "source_path": "prepared/tasks/train-fasttext", "instruction_sha256": "b701c31f4c5eb6f7e9a6b74d62a52eb696dcae9548a8252c78c0fa850714f998", "task_config_sha256": "994fda7c3081bbdd9f6d05d6279ef64faefa94c214d1e08fbcea656540e4fe91", "feature_text_sha256": "5ae0cfd27af1570fcf2890a783324a25a00fbf8c30860b232b59f8548cc29c2d", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/train-fasttext:20251031", "cpus": 1, "memory_mb": 4096, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C11", "class_id": "C11", "class_label": "Neural model, embedding and inference workflows", "fine_family": "supervised_model_training", "classification_reason": "Trains a supervised classifier subject to both predictive quality and serialized model-size constraints.", "transfer_tags": [ "fastText", "text-classification", "training", "model-size", "accuracy" ], "split": "test" }, { "task_id": "tune-mjcf", "task_version_id": "f802cbd8-ed91-46e5-910b-c1a849e3626f", "task_revision": 3, "source_path": "prepared/tasks/tune-mjcf", "instruction_sha256": "c7fb299b6c6d6c9a2dcd95c5938e6f902ddad1cb02aee875d369cb7eae584b1f", "task_config_sha256": "27b6d4e5a23761cc1a24a695125d891be74a2930fb83a02a66221db0244b08b3", "feature_text_sha256": "bd00aac26b3dbb084874053e5caa81df63c088b283dad026055c3be6bdc1aa8a", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/tune-mjcf:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C07", "class_id": "C07", "class_label": "Statistical modeling, numerical inference and optimization", "fine_family": "simulation_performance_optimization", "classification_reason": "Optimizes a physics model's execution configuration while preserving its state trajectory within tolerance and meeting runtime requirements.", "transfer_tags": [ "MuJoCo", "MJCF", "simulation", "numerical-equivalence", "benchmarking" ], "split": "test" }, { "task_id": "video-processing", "task_version_id": "7d8118c9-905f-45cb-b9da-137a8f06446e", "task_revision": 3, "source_path": "prepared/tasks/video-processing", "instruction_sha256": "15936f5e61823ae4794994d6e1b7e4e776b40968ea382405b868c501dfe032db", "task_config_sha256": "be4fb42fccc5faa4862789cedd2c560eaa5758a2971989649a6dafeaeccc17aa", "feature_text_sha256": "cd545fc5a5105dc46700f9c57a15f871e5dc41416d75295737d3e66331cd3bcb", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/video-processing:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C13", "class_id": "C13", "class_label": "Document and media interpretation or transformation", "fine_family": "video_frame_interpretation", "classification_reason": "Identifies takeoff and landing frames through classical fixed-camera video analysis under explicit library restrictions.", "transfer_tags": [ "OpenCV", "NumPy", "motion-analysis", "foreground", "temporal-events" ], "split": "training" }, { "task_id": "vulnerable-secret", "task_version_id": "5c126f8c-7edc-42b9-8511-c581714da868", "task_revision": 3, "source_path": "prepared/tasks/vulnerable-secret", "instruction_sha256": "d43c72e6c39cdeb1141a9c8eae8357df1b0901383ee2beedacf828bf017d7a2f", "task_config_sha256": "51079626e4c76445e616eff8723efc53af7f1cf8d9c64d15aab2f4f0ecd6df3c", "feature_text_sha256": "adebf6f7966035bcbc38876086f30bd0e127fe8d87ddd0791f56dff2807e4b68", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/vulnerable-secret:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C06", "class_id": "C06", "class_label": "Cryptanalysis and secret recovery", "fine_family": "binary_secret_extraction", "classification_reason": "Inspects and interacts with a supplied executable to recover its embedded secret; binary evidence is central rather than web input sanitization.", "transfer_tags": [ "executable", "reverse-engineering", "vulnerability", "flag-extraction" ], "split": "training" }, { "task_id": "winning-avg-corewars", "task_version_id": "d6bca3f2-e4db-48e3-ba4c-f58f19f5f28f", "task_revision": 4, "source_path": "prepared/tasks/winning-avg-corewars", "instruction_sha256": "f8198a1644547b116b4b67384f6bb6b3a78e85076f8cda20079a77f2f3499242", "task_config_sha256": "a9b48765d03a66d966f6a162be203bc31e1d9d6e4abe616a2f4b42bac5dc246f", "feature_text_sha256": "fb5237644423da3f086b6bd6d8afa47061ac8b1396025ffb8478ae197d930d75", "official_agent_timeout_seconds": 3600.0, "official_verifier_timeout_seconds": 3600.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/winning-avg-corewars:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C12", "class_id": "C12", "class_label": "Constrained language and program synthesis", "fine_family": "redcode_strategy_search", "classification_reason": "Constructs and benchmarks a program in a restricted virtual instruction set against fixed opponents while preserving their source files.", "transfer_tags": [ "Redcode", "pMARS", "strategy-search", "simulation", "immutable-opponents" ], "split": "test" }, { "task_id": "write-compressor", "task_version_id": "029e85eb-7109-46ec-8605-6c59909d2182", "task_revision": 3, "source_path": "prepared/tasks/write-compressor", "instruction_sha256": "c4fc3e73e44deeac95f1cd4a34a2d87e853b22ceec2b2824767f5f087b470bb2", "task_config_sha256": "6066a509066433c97c417fea68c4e3e73996f698dcdcbabe07718835eea355a6", "feature_text_sha256": "b07798ef5be7cea387a0ab812bcfdcdb8b91320de50ef73bc6021016e19e5b8a", "official_agent_timeout_seconds": 900.0, "official_verifier_timeout_seconds": 900.0, "official_environment": { "build_timeout_sec": 600.0, "docker_image": "alexgshaw/write-compressor:20251031", "cpus": 1, "memory_mb": 2048, "storage_mb": 10240, "gpus": 0, "allow_internet": true, "mcp_servers": [], "env": {} }, "cluster_id": "C12", "class_id": "C12", "class_label": "Constrained language and program synthesis", "fine_family": "inverse_codec_synthesis", "classification_reason": "Infers the supplied decoder's representation and constructs a compatible compressed stream that round-trips exactly under a size limit.", "transfer_tags": [ "decompressor", "encoding", "round-trip", "size-bound", "reverse-engineering" ], "split": "training" } ] }