{"data":{"kind":"file","path":"README.md","version_id":"omxhf4x8g1cxl3plwfpsrm0x","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":2552,"modified_at":"2026-09-13T04:06:06.957000","content_hash":"49422b21e3d688e4079068730e143f764cf01736ce24ae4a575b9fbef1437a6e"},"entries":[],"content":"# swelego\n\nSWE-Lego real-data issue-resolving tasks drawn from real GitHub issues, each with a public image carrying the repo. Scored 1.0 iff every FAIL_TO_PASS and PASS_TO_PASS test id passes after the held-out test patch is re-applied and the task's canonical test command is run.\n\n## Taskset\n\n- **Source:** [`PrimeIntellect/SWE-Lego-Real-Data-Verified`](https://huggingface.co/datasets/PrimeIntellect/SWE-Lego-Real-Data-Verified)\n- **Size:** 4,323 tasks (`resolved` split)\n\n## Changelog\n\n- 2026-09-04: Add a solver system prompt (adapted from FrontierCode 1.1's fair-internet-use prompt) that allows documentation lookups but forbids retrieving the task's upstream fix online or from non-current-branch git history; complements #795, which restored network access.\n- 2026-09-03: Restore default solver network access by removing the `network_allow=[]` override introduced in #780; training rollouts need outbound network.\n- 2026-08-31: Yield task records on demand so bounded evaluations construct only the requested prefix.\n- 2026-07-17: `patch_capture` now imports from `verifiers.v1` (`capture_patch`/`resolve_head` upstreamed in verifiers#2054); the copied module is removed. Requires `verifiers>=0.2.2.dev5`.\n- 2026-07-16: `finalize` persists the agent's final patch to `trace.info[\"patch\"]` (truncated at 2 MB with `patch_truncated`; failures record `patch_error` instead of failing the rollout) via the copied `patch_capture` helper.\n- 2026-07-10: Ported to the task-centric verifiers API: the `solved` reward and the `setup` hook now live on `SWELegoTask` (a `SWELegoData` row + behavior split), and the taskset exposes `load` instead of `load_tasks`. Load-time config (`dataset_name`, `split`, `filter_fn`) stays on `--env.taskset.*`. Requires `verifiers>=0.2.0` and Python `>=3.11`.\n- 2026-07-08: Default dataset is now the gold-patch-validated `PrimeIntellect/SWE-Lego-Real-Data-Verified` (4,323 rows); the fixed-fork base `SWE-Lego-Real-Data` stays available as a typed override.\n- 2026-07-06: `PrimeIntellect/SWE-Lego-Real-Data-Validated` renamed to `PrimeIntellect/SWE-Lego-Real-Data-Verified` in the typed dataset options; the old name redirects.\n- 2026-07-01: Added `filter_fn`, applied directly with `datasets.Dataset.filter` to raw HF rows before task construction.\n- 2026-06-30: Setup no longer applies `test_patch`. Scoring applies the patch after resetting touched files from `base_commit` when available, matching the original SWE-Lego/SWE-bench eval pattern rather than the verifiers composable port's setup-time application.\n","encoding":"utf-8","truncated":false,"total_bytes":2552},"status":null}