58k
Tasks
Task dataset
Real engineering problems with cleaned prompts and execution summaries. The foundation for SFT on problem comprehension and solution planning.
The full agent trajectory. Not just the code.
58k real engineering tasks captured end-to-end: reasoning traces, tool calls, code edits, and explicit human acceptance signals. Built for labs training the next generation of coding agents.
Trajectories · sample
| Task | Repo | Type | Steps | Tool calls | Reward | Status |
|---|---|---|---|---|---|---|
| T-58021 | api-gateway | bugfix | 23 | 6 | 0.91 | accepted |
| T-58018 | auth-service | feature | 31 | 9 | 0.87 | accepted |
| T-58014 | billing-core | refactor | 18 | 5 | 0.74 | Scoring... |
| T-58011 | ui-kit | bugfix | 12 | 4 | 0.95 | accepted |
| T-58007 | data-pipeline | feature | 27 | 11 | 0.82 | accepted |
| T-58003 | ml-serving | bugfix | 41 | 14 | 0.68 | Scoring... |
| T-57998 | search-index | perf | 19 | 7 | 0.89 | accepted |
| T-57994 | api-gateway | bugfix | 15 | 5 | 0.93 | accepted |
| T-57990 | worker-pool | refactor | 22 | 8 | 0.79 | accepted |
| T-57986 | auth-service | feature | 34 | 12 | 0.71 | Scoring... |
| T-57981 | cli-tools | bugfix | 9 | 3 | 0.96 | accepted |
- 58kreal engineering tasks
- 833ktrajectory steps
- 29kaccepted code actions
- 22tools logged with context
How this goes beyond the benchmarks
| Capability | SWE-bench | HumanEval | CodeSearchNet | Zstate SWE |
|---|---|---|---|---|
| Full reasoning traces | ✕ | ✕ | ✕ | ✓ |
| Tool use captured | ✕ | ✕ | ✕ | ✓ |
| Human acceptance signals | ✕ | ✕ | ✕ | ✓ |
| Real production tasks | ✓ | ✕ | ~ | ✓ |
| Action-level reward signal | ✕ | ✕ | ✕ | ✓ |
| Scale (tasks) | 300 | 164 | ~100k | 58k |
One corpus. Three lenses.
833k
Steps
Trajectory dataset
Full step-by-step agent traces: reasoning, tool usage, and code generation at every decision point. The complete picture of how an expert agent solves problems.
29k
Signals
Reward dataset
Explicit user acceptance signals at the action level. 50% acceptance rate. Supports iterative multi-accept workflows and action-level reward modelling.
Every tool interaction logged with context.
Semantic search
Call graph analysis
File edits
CLI execution
Directory traversal
Test runner
Symbol lookup
Dependency graph
Code commenting
Diff generation
Static analysis
more tools across the corpus
Ready to see the SWE dataset schema?
We're packaging a sample and schema for AI lab outreach now. Get in touch to be first in line, or to discuss a curated subset built for your training pipeline.