Upload
A dataset folder, or log files. They appear on the Corpus Overview and stay in this browser.
Drop a dataset folder or log files here or
Claude Code · Codex · Qwen Code · SWE-agent · OpenHands · mini-swe-agent · ATIF · Markdown (SpecStory, Claude Code /export)
How a folder is read
<dataset>/
<bug>/ e.g. bug-1130
<agent>/ e.g. claude
run1/ <log>
run2/ <log>
outcomes.csv optional: bug, agent, outcome (external validation)
Each run folder holds one trajectory; other files beside it (metadata, sandbox logs) are not processed. The outcome CSV may name the run (a run column) or give one column per run (run1, run2…); its values are pass / fail or 1 / 0. The outcome is the external validation: refers to test case passed & verified externally.