task stringlengths 20 20 | num_rollouts int64 2 85 | num_passed int64 1 18 | num_failed int64 1 80 | pass_rate float64 0.05 0.94 | prior_num_rollouts int64 0 85 | prior_num_passed int64 0 18 | prior_num_failed int64 0 80 | current_num_rollouts int64 0 8 | current_num_passed int64 0 8 | current_num_failed int64 0 7 | has_prior_rollouts bool 2
classes | has_current_rollouts bool 2
classes | instruction stringlengths 925 4.06k | task_toml stringlengths 754 1.21k | file_paths listlengths 6 11 | file_contents listlengths 6 11 | file_encodings listlengths 6 11 | source_repo stringclasses 1
value | source_revision stringclasses 1
value | source_split stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
task_000000_f8baca82 | 4 | 3 | 1 | 0.75 | 0 | 0 | 0 | 4 | 3 | 1 | false | true | You are a data analyst. You have been provided with a CSV file at `/home/user/network_data.csv` containing transaction logs for a network of users.
The CSV has the following columns:
`tx_id`, `sender`, `receiver`, `amount`, `timestamp`
Your task is to analyze this data by performing both analytical aggregations and g... | schema_version = "1.1"
[task]
name = "tmax/task_000000_f8baca82"
description = "data analyst processing CSV files"
[metadata]
source = "legacy"
base_image = ""
domain = "data_querying"
skill_type = "Data Comprehension"
primitive_skills = ["Output schema validation", "Window functions and analytical aggregation", "Gra... | [
"environment/Dockerfile",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\n\nCOPY post_install.sh /tmp/post_install.sh\nRUN bash /tmp/post_install.sh && rm /tmp/post_install.sh\n",
"apt-get update && apt-get install -y python3 python3-pip\n pip3 install pytest networkx pandas scipy\n\n mkdir -p /home/user\n\n cat << 'EOF... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000001_e67d5628 | 4 | 1 | 3 | 0.25 | 0 | 0 | 0 | 4 | 1 | 3 | false | true | You are a backup operator tasked with testing the restoration of our critical network infrastructure configurations. The backup archive contains routing tables and firewall rules that need to be parsed and validated using our legacy C libraries.
To complete this scenario, you must perform the following steps:
1. **M... | schema_version = "1.1"
[task]
name = "tmax/task_000001_e67d5628"
description = "backup operator testing restores"
[metadata]
source = "v2"
base_image = "intricate"
domain = "system_administration"
skill_type = "Network"
primitive_skills = ["Locale and timezone configuration", "Text processing pipelines (awk, sed, gre... | [
"environment/Dockerfile",
"environment/_fixtures/app/fixtures/minicalc/Makefile",
"environment/_fixtures/app/fixtures/minicalc/README.md",
"environment/_fixtures/app/fixtures/minicalc/_perturbation.json",
"environment/_fixtures/app/fixtures/minicalc/minicalc.c",
"environment/base_install.sh",
"environme... | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\nENV LC_ALL=C.UTF-8\nENV LANG=C.UTF-8\nENV GOPATH=/home/user/go\nENV PATH=/home/user/go/bin:/home/user/.cargo/bin:$PATH\n\nCOPY base_install.sh /tmp/base_install.sh\nRUN bash /tmp/base_install.sh && rm /tmp/base_install.sh\n\nCOPY _fixtures/app/fixtures/minic... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000004_aeddda76 | 2 | 1 | 1 | 0.5 | 0 | 0 | 0 | 2 | 1 | 1 | false | true | You are assisting a storage administrator in managing and securing user uploads. Users frequently upload `.tar.gz` archives to the `/home/user/incoming` directory. However, some of these archives are malicious and contain "zip slip" payloads—files with absolute paths (e.g., `/etc/passwd`) or directory traversal sequenc... | schema_version = "1.1"
[task]
name = "tmax/task_000004_aeddda76"
description = "storage administrator managing disk space"
[metadata]
source = "legacy"
base_image = ""
domain = "file_operations"
skill_type = "Transformation"
primitive_skills = ["Bulk file renaming", "File watching and change detection", "Compressed s... | [
"environment/Dockerfile",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\n\nCOPY post_install.sh /tmp/post_install.sh\nRUN bash /tmp/post_install.sh && rm /tmp/post_install.sh\n",
"apt-get update && apt-get install -y python3 python3-pip\n pip3 install pytest\n\n useradd -m -s /bin/bash user || true\n\n mkdir -p /home/u... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000005_ceefb305 | 3 | 1 | 2 | 0.333333 | 0 | 0 | 0 | 3 | 1 | 2 | false | true | You are a backup administrator tasked with archiving active system binaries and legacy logs. You need to write a C program to safely extract metadata and convert log encodings, then use shell commands to create a multi-part archive.
Step 1: Write a C program
Create a C program at `/home/user/archiver.c` and compile it... | schema_version = "1.1"
[task]
name = "tmax/task_000005_ceefb305"
description = "backup administrator archiving data"
[metadata]
source = "legacy"
base_image = ""
domain = "file_operations"
skill_type = "Transformation"
primitive_skills = ["Character encoding conversion", "Nested and multi-part archive handling", "Fil... | [
"environment/Dockerfile",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\n\nCOPY post_install.sh /tmp/post_install.sh\nRUN bash /tmp/post_install.sh && rm /tmp/post_install.sh\n",
"apt-get update && apt-get install -y python3 python3-pip gcc libc6-dev\n pip3 install pytest\n\n useradd -m -s /bin/bash user || true\n\n mk... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000006_74a36316 | 2 | 1 | 1 | 0.5 | 0 | 0 | 0 | 2 | 1 | 1 | false | true | **PagerDuty Alert: CRITICAL**
**Time:** 03:14 AM
**Service:** VisionMath-Analytics
**Issue:** The real-time video analytics service is returning `NaN` in its JSON payload, dropping TCP connections, and randomly crashing under load. Downstream clients are failing to parse the coordinates.
You are the on-call engineer. ... | schema_version = "1.1"
[task]
name = "tmax/task_000006_74a36316"
description = "on-call engineer responding to a 3am page"
[metadata]
source = "v2"
base_image = "intricate"
domain = "debugging"
skill_type = "Mathematical"
primitive_skills = ["Logging and traceback analysis", "Precision loss tracking", "Race condition... | [
"environment/Dockerfile",
"environment/_fixtures/app/fixtures/video.label.txt",
"environment/_fixtures/app/fixtures/video.mp4",
"environment/base_install.sh",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\nENV LC_ALL=C.UTF-8\nENV LANG=C.UTF-8\nENV GOPATH=/home/user/go\nENV PATH=/home/user/go/bin:/home/user/.cargo/bin:$PATH\n\nCOPY base_install.sh /tmp/base_install.sh\nRUN bash /tmp/base_install.sh && rm /tmp/base_install.sh\n\nCOPY _fixtures/app/fixtures/video... | [
"utf-8",
"utf-8",
"base64",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000006_93ea43e6 | 3 | 1 | 2 | 0.333333 | 0 | 0 | 0 | 3 | 1 | 2 | false | true | You are managing a custom configuration tracker that stores incremental configuration changes in Write-Ahead Log (WAL) formats, securely packaged in nested archives.
Your task is to reconstruct the final configuration state for a specific service by extracting and parsing these fragmented log files.
The system store... | schema_version = "1.1"
[task]
name = "tmax/task_000006_93ea43e6"
description = "configuration manager tracking changes"
[metadata]
source = "legacy"
base_image = ""
domain = "file_operations"
skill_type = "Archives"
primitive_skills = ["Archive creation and extraction", "Domain-specific format parsing (GCode, ELF, WA... | [
"environment/Dockerfile",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\n\nCOPY post_install.sh /tmp/post_install.sh\nRUN bash /tmp/post_install.sh && rm /tmp/post_install.sh\n",
"apt-get update && apt-get install -y python3 python3-pip jq\n pip3 install pytest\n\n useradd -m -s /bin/bash user || true\n\n cat << 'EOF' ... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000008_e94be48f | 4 | 3 | 1 | 0.75 | 0 | 0 | 0 | 4 | 3 | 1 | false | true | You are a data scientist working on a sensor network. You need to clean and join two datasets using a high-performance C program, as the real datasets will be massive.
The datasets are located in `/home/user/data/`:
1. `sensors.csv`: Contains sensor metadata.
Format: `sensor_id,location,calibration_factor`
2. `read... | schema_version = "1.1"
[task]
name = "tmax/task_000008_e94be48f"
description = "data scientist cleaning datasets"
[metadata]
source = "legacy"
base_image = ""
domain = "data_science"
skill_type = "Systems"
primitive_skills = ["Large-scale data storage management", "Data schema enforcement", "Numerical accuracy testin... | [
"environment/Dockerfile",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\n\nCOPY post_install.sh /tmp/post_install.sh\nRUN bash /tmp/post_install.sh && rm /tmp/post_install.sh\n",
"apt-get update && apt-get install -y python3 python3-pip gcc libc6-dev\n pip3 install pytest\n\n useradd -m -s /bin/bash user || true\n\n mk... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000014_3d9789ba | 4 | 3 | 1 | 0.75 | 0 | 0 | 0 | 4 | 3 | 1 | false | true | You are an AI assistant helping a climate researcher organize a messy, deeply nested dataset of sensor readings.
The raw data is located in `/home/user/raw_datasets/`. Over time, the data has been dumped into various archive formats (`.tar.gz`, `.tar`, `.zip`) scattered across multiple subdirectories. Inside these ar... | schema_version = "1.1"
[task]
name = "tmax/task_000014_3d9789ba"
description = "researcher organizing datasets"
[metadata]
source = "legacy"
base_image = ""
domain = "file_operations"
skill_type = "Archives"
primitive_skills = ["Archive creation and extraction", "Filesystem mount and path manipulation", "Recursive di... | [
"environment/Dockerfile",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\n\nCOPY post_install.sh /tmp/post_install.sh\nRUN bash /tmp/post_install.sh && rm /tmp/post_install.sh\n",
"apt-get update && apt-get install -y python3 python3-pip zip unzip tar\n pip3 install pytest\n\n useradd -m -s /bin/bash user || true\n\n mk... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000016_3ffaf505 | 9 | 3 | 6 | 0.333333 | 8 | 2 | 6 | 1 | 1 | 0 | true | true | You are a data engineer responsible for a new time-series ETL pipeline. We receive IoT sensor data from two different systems, both outputting CSV files. You need to write a Rust program to read these files, validate the data, deduplicate it using a cryptographic hash, and load it into a SQLite database.
I have create... | schema_version = "1.1"
[task]
name = "tmax/task_000016_3ffaf505"
description = "data engineer building ETL pipelines"
[metadata]
source = "legacy"
base_image = ""
domain = "data_processing"
skill_type = "Time Series"
primitive_skills = ["Validation checkpoints and quality gates", "Hash-based deduplication", "Database... | [
"environment/Dockerfile",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\n\nCOPY post_install.sh /tmp/post_install.sh\nRUN bash /tmp/post_install.sh && rm /tmp/post_install.sh\n",
"apt-get update && apt-get install -y python3 python3-pip curl build-essential sqlite3\n pip3 install pytest\n\n export RUSTUP_HOME=/opt/rustup\... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000019_9b7bd652 | 2 | 1 | 1 | 0.5 | 0 | 0 | 0 | 2 | 1 | 1 | false | true | You are a QA engineer setting up a test environment for our physics simulation pipeline. We are migrating our trajectory prediction module to C, but the current project is broken.
You need to complete the following tasks:
1. **Video Analysis**: We lost the original parameters of our physics model. However, we have a... | schema_version = "1.1"
[task]
name = "tmax/task_000019_9b7bd652"
description = "QA engineer setting up test environments"
[metadata]
source = "v2"
base_image = "intricate"
domain = "software_engineering"
skill_type = "Mathematical"
primitive_skills = ["Custom data structure design", "Property-based testing", "gRPC an... | [
"environment/Dockerfile",
"environment/_fixtures/app/fixtures/video.label.txt",
"environment/_fixtures/app/fixtures/video.mp4",
"environment/base_install.sh",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\nENV LC_ALL=C.UTF-8\nENV LANG=C.UTF-8\nENV GOPATH=/home/user/go\nENV PATH=/home/user/go/bin:/home/user/.cargo/bin:$PATH\n\nCOPY base_install.sh /tmp/base_install.sh\nRUN bash /tmp/base_install.sh && rm /tmp/base_install.sh\n\nCOPY _fixtures/app/fixtures/video... | [
"utf-8",
"utf-8",
"base64",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000019_d40ae014 | 12 | 9 | 3 | 0.75 | 4 | 3 | 1 | 8 | 6 | 2 | true | true | You are an AI assistant acting as a backup administrator. We need to consolidate and archive legacy data from different servers into a unified format before storing it long-term.
The raw data is located in `/home/user/legacy_data/` and has two subdirectories: `win_logs/` and `csv_metrics/`.
Here is what you need to ... | schema_version = "1.1"
[task]
name = "tmax/task_000019_d40ae014"
description = "backup administrator archiving data"
[metadata]
source = "legacy"
base_image = ""
domain = "file_operations"
skill_type = "Archives"
primitive_skills = ["Format conversion between file types", "Character encoding conversion", "Bulk file r... | [
"environment/Dockerfile",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\n\nCOPY post_install.sh /tmp/post_install.sh\nRUN bash /tmp/post_install.sh && rm /tmp/post_install.sh\n",
"apt-get update && apt-get install -y python3 python3-pip jq file\n pip3 install pytest\n\n useradd -m -s /bin/bash user || true\n\n mkdir -p... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000022_21389603 | 7 | 6 | 1 | 0.857143 | 4 | 3 | 1 | 3 | 3 | 0 | true | true | You are a red-team operator simulating a local privilege escalation attack. You have obtained the source code of a custom authentication tool located at `/home/user/validator.go`.
This tool validates a custom three-part token (similar to a JWT) provided as a command-line argument. If the token is valid and belongs to... | schema_version = "1.1"
[task]
name = "tmax/task_000022_21389603"
description = "red-team operator crafting an evasion payload"
[metadata]
source = "legacy"
base_image = ""
domain = "security"
skill_type = "Algorithmic"
primitive_skills = ["Privilege escalation auditing", "Reverse engineering and disassembly", "Encryp... | [
"environment/Dockerfile",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\n\nCOPY post_install.sh /tmp/post_install.sh\nRUN bash /tmp/post_install.sh && rm /tmp/post_install.sh\n",
"apt-get update && apt-get install -y python3 python3-pip golang\n pip3 install pytest\n\n useradd -m -s /bin/bash user || true\n\n cat << 'E... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
task_000031_bae35058 | 2 | 1 | 1 | 0.5 | 0 | 0 | 0 | 2 | 1 | 1 | false | true | You are acting as a support engineer investigating a bug reported by a customer. The customer uses our Rust-based job submission tool, `rusty-dispatcher`, to send files to a cluster of worker containers. They are reporting that files are randomly being dropped or failing to process.
You have been provided with two thi... | schema_version = "1.1"
[task]
name = "tmax/task_000031_bae35058"
description = "support engineer collecting diagnostics"
[metadata]
source = "legacy"
base_image = ""
domain = "debugging"
skill_type = "Systems"
primitive_skills = ["Regression test construction", "Log timeline reconstruction across services", "Containe... | [
"environment/Dockerfile",
"environment/post_install.sh",
"instruction.md",
"task.toml",
"tests/test.sh",
"tests/test_final_state.py"
] | [
"FROM ubuntu:22.04\n\nENV DEBIAN_FRONTEND=noninteractive\n\nCOPY post_install.sh /tmp/post_install.sh\nRUN bash /tmp/post_install.sh && rm /tmp/post_install.sh\n",
"apt-get update && apt-get install -y python3 python3-pip cargo\n pip3 install pytest\n\n useradd -m -s /bin/bash user || true\n\n mkdir -p /... | [
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8",
"utf-8"
] | PrimeIntellect-ai/research-tasks | 8b38d35b53271a5f955dfc5dd8197d562cebf46e | tmax@2026-07-01 |
End of preview. Expand in Data Studio
README.md exists but content is empty.
- Downloads last month
- -