Data catalog
Tasks & Environments
Compare task coverage and delivery options by domain.
Software & terminal
4-
01
SWE-Marathon
Long-horizon software engineering
- Benchmark / task domain
- SWE-Marathon
- Tasks
- System builds, rewrites and recovery
- Delivery format
- Harbor-format tasks, Docker environments, reference solutions, acceptance tests
-
02
DeepSWE
Repository engineering tasks
- Benchmark / task domain
- DeepSWE
- Tasks
- Repository-level implementation and repair
- Delivery format
- Harbor-format tasks, Docker environments, reference solutions, evaluation code
-
03
Terminal-Bench 3.0
Repository repair and migration
- Benchmark / task domain
- Terminal-Bench 3.0 format
- Tasks
- Config migration, transaction recovery, persistence
- Delivery format
- Harbor-format tasks, Docker environments, reference solutions, evaluation code
-
04
Terminal-Bench 2.1
Terminal tools and systems
- Benchmark / task domain
- Terminal-Bench 2.1 format
- Tasks
- Binary parsing, concurrent services, persistent state, memory safety
- Delivery format
- Harbor-format tasks, Docker environments, reference solutions, evaluation code
Cybersecurity
2-
05
ExploitBench V8
Security task environments
- Benchmark / task domain
- ExploitBench / bench-v8
- Tasks
- V8 vulnerability research and exploitation-capability evaluation
- Delivery format
- JSON / YAML task configurations, container images, MCP interfaces, evaluation components
-
06
ExploitGym
CVE task packages
- Benchmark / task domain
- ExploitGym
- Tasks
- Vulnerability analysis and remediation assessment
- Delivery format
- Task metadata, Dockerfiles, binaries, evaluation code
Desktop agents (CUA)
1-
07
CUA RL Desktop Tasks
Custom desktop tasks
- Benchmark / task domain
- Windows / macOS
- Tasks
- Cross-application work, spreadsheets, document layouts, data entry, recovery
- Delivery format
- Desktop tasks, VM environments and initial snapshots, reference workflows, evaluation tools
Professional workflows
3-
08
AutomationBench
Simulated API environments
- Benchmark / task domain
- AutomationBench
- Tasks
- Invoice posting, access review, meeting conflicts
- Delivery format
- API simulations, business policies and initial data, reference workflows, evaluation tools
-
09
GDPVal
Document and spreadsheet tasks
- Benchmark / task domain
- GDPVal
- Tasks
- Quality engineering, credit analysis, integration and operations
- Delivery format
- Docker environments, business attachments, DOCX / XLSX / PDF, scoring criteria
-
10
ALE
File deliverable tasks
- Benchmark / task domain
- ALE
- Tasks
- Financial analysis, spreadsheet editing, imaging/video QC, log triage
- Delivery format
- JSON task cards, input materials, CSV / JSON / ODS reference files, evaluation code
Deep research
1-
11
STEM Research QA
Research question-answer data
- Benchmark / task domain
- STEM Research QA
- Tasks
- Literature analysis, experimental interpretation, scientific derivations
- Delivery format
- JSON QA data, PDF / XLSX attachments, reference answers, scoring criteria
Finance & quant
2-
12
Financial Research Trajectories
Research dialogues and tool traces
- Benchmark / task domain
- Financial research
- Tasks
- Disclosure checks, comparable valuation, basic portfolio risk
- Delivery format
- Multi-turn research dialogues, tool-use traces, calculations and analytical outputs
-
13
QuantEnv
Prediction and scoring tasks
- Benchmark / task domain
- QuantEnv
- Tasks
- Direction forecasts, sector ranking, leader selection, factor analysis
- Delivery format
- Docker environments, JSON / JSONL research data, submission schemas, evaluation tools
AI for science
2-
14
GeneBench-Pro
Life-sciences analysis and experimental decisions
- Benchmark / task domain
- GeneBench-Pro
- Tasks
- RNA-seq analysis, candidate screening, experiment decisions
- Delivery format
- Harbor-format tasks, research inputs, reference solutions, evaluation code
-
15
Terminal-Bench-Science
Scientific numerical tasks
- Benchmark / task domain
- Terminal-Bench format
- Tasks
- Model calibration, experimental design, uncertainty, prediction envelopes
- Delivery format
- Harbor-format tasks, Docker environments, research inputs, reference solutions and evaluation code
Visual QA (VQA)
1-
16
Chartography
Chart question-answer data
- Benchmark / task domain
- Chartography
- Tasks
- Chart reading, cross-panel matching, graph reasoning
- Delivery format
- Chart images, Parquet / YAML QA data, answer explanations, evaluation tools
Design workspace
1-
17
Design workspace
Design tasks and editing records
- Benchmark / task domain
- Design workflows
- Tasks
- Product retouching, packaging, portraits
- Delivery format
- Source assets, editing records, editable PSD, PNG
No matching programs. Try another search or clear the filters.
WORK WITH ASTRAFLUX
What should your
model be able to do?
Share the capability, task types and delivery requirements. We’ll discuss the right program and representative tasks.
Contact us norman383@astraflux.space