Harness adapters: Safe RL for efficient agentic development
harness-hub
β’ β’ 14trajectory stringlengths 79 126 | gate stringclasses 1
value | task_id stringlengths 29 33 | family stringclasses 3
values | source_trace stringlengths 83 115 | kept stringclasses 1
value | reason stringclasses 1
value | request stringlengths 41 200 β | hidden_tests stringclasses 16
values | mutant_kill_rate stringclasses 18
values | solve_rate stringlengths 3 6 | shard stringclasses 4
values | difficulty stringclasses 3
values | rewards stringlengths 10 16 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
scale/trajectories/shard-00/TeichAI__DeepSeek-v4-Pro-Agent__026-05-08T02-12-02-072Z_019e055b-2d98-72bb-bb74-3a3a63938563.json | skipped (--no-scan) | 019e055b-2d98-72bb-bb74-build | build | hf://datasets/TeichAI/DeepSeek-v4-Pro-Agent/2026-05-08T02-12-02-072Z_019e055b-2d98-72bb-bb74-3a3a63938563.jsonl | True | ok | I'm building a text adventure / RPG in Python (no external libraries) and I want to implement a ghost enemy with a proper state machine. I've started coding a Ghost class but I'm running into a proble | 16 | 0.688 | 0.5625 | shard-00 | medium | [0.5, 0.625] |
scale/trajectories/shard-00/AletheiaResearch__GLM-5.2-Agent__026-06-20T05-10-58-153Z_019ee370-7368-7700-918a-9296213f41d1.json | skipped (--no-scan) | 019ee370-7368-7700-918a-build | build | hf://datasets/AletheiaResearch/GLM-5.2-Agent/2026-06-20T05-10-58-153Z_019ee370-7368-7700-918a-9296213f41d1.jsonl | True | ok | Debug this scenario: a background job retry system retries forever for some permanent failures. Fix the retry policy and add clear dead-letter behavior. | 19 | 0.75 | 0.8947 | shard-00 | easy | [0.8947, 0.8947] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T03-00-56-676Z_01a031b6-c4e3-7421-b297-ff82057c58ff.json | skipped (--no-scan) | 01a031b6-c4e3-7421-b297-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T03-00-56-676Z_01a031b6-c4e3-7421-b297-ff82057c58ff.jsonl | True | ok | Compressed radix tree (Patricia trie) for byte strings in Python. radix.py: class RadixSet with add(key: bytes), contains, discard, and `keys_with_prefix(prefix)` returning sorted results; edges hold | 14 | 1.0 | 0.2857 | shard-00 | hard | [0.0714, 0.5] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T03-10-48-004Z_01a031bf-cac4-74a0-926b-33da42fdb9c2.json | skipped (--no-scan) | 01a031bf-cac4-74a0-926b-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T03-10-48-004Z_01a031bf-cac4-74a0-926b-33da42fdb9c2.jsonl | True | ok | Expert geometry task: a working Bentley-Ottmann sweep line in Python reporting all intersection points among n segments. Package sweep/ with geometry.py (exact orientation and segment-intersection pre | 16 | 0.812 | 0.4688 | shard-00 | medium | [0.4375, 0.5] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T03-15-02-708Z_01a031c3-adb3-74ef-b552-ad9c25d132f7.json | skipped (--no-scan) | 01a031c3-adb3-74ef-b552-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T03-15-02-708Z_01a031c3-adb3-74ef-b552-ad9c25d132f7.jsonl | True | ok | Rotating calipers for point-set diameter in Python. calipers.py: `diameter(points)` β build the convex hull (monotone chain, include it in the file), then rotate antipodal pairs to find the farthest p | 17 | 0.875 | 0.7647 | shard-00 | easy | [0.5294, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T03-23-16-897Z_01a031cb-3820-71fe-90b7-4e84884571c0.json | skipped (--no-scan) | 01a031cb-3820-71fe-90b7-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T03-23-16-897Z_01a031cb-3820-71fe-90b7-4e84884571c0.jsonl | True | ok | Implement HyperLogLog from the paper, Python stdlib only. hll.py: class HLL(p) with 2^p registers, add(item) hashing with blake2b (64-bit), estimate() including the standard bias correction for small | 16 | 0.625 | 0.7188 | shard-00 | easy | [1.0, 0.4375] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-18-40-143Z_01a031fd-ed8e-73e2-b75e-a80579af5333.json | skipped (--no-scan) | 01a031fd-ed8e-73e2-b75e-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-18-40-143Z_01a031fd-ed8e-73e2-b75e-a80579af5333.jsonl | True | ok | From a click log CSV with columns user_id,timestamp_iso,url write sessionize.py: group each user's events into sessions that split when the gap between consecutive events exceeds 30 minutes, then prin | 18 | 0.812 | 0.4722 | shard-00 | medium | [0.2778, 0.6667] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-26-34-154Z_01a03205-2929-7461-a96f-75dd5e32a29e.json | skipped (--no-scan) | 01a03205-2929-7461-a96f-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-26-34-154Z_01a03205-2929-7461-a96f-75dd5e32a29e.jsonl | True | ok | Write collector.py, a long-running process that reads lines from stdin, buffers up to 100, and flushes the buffer as one JSON array per line appended to out.jsonl every 100 lines or every 0.5 seconds, | 8 | 0.75 | 0.375 | shard-00 | medium | [0.0, 0.75] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-31-36-242Z_01a03209-c531-76da-81c7-0341756b6afd.json | skipped (--no-scan) | 01a03209-c531-76da-81c7-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-31-36-242Z_01a03209-c531-76da-81c7-0341756b6afd.jsonl | True | ok | csv2json.py converts CSV (first row headers) to a JSON array of objects with per-column type inference: int, float, true/false to booleans, empty string to null, everything else string; --no-infer kee | 17 | 0.812 | 0.8235 | shard-00 | easy | [0.8235, 0.8235] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-37-17-883Z_01a0320e-fbba-727e-b7bb-44c72a3a8c72.json | skipped (--no-scan) | 01a0320e-fbba-727e-b7bb-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-37-17-883Z_01a0320e-fbba-727e-b7bb-44c72a3a8c72.jsonl | True | ok | treelite.py prints a directory tree in the style of tree(1) with box-drawing prefixes for branch, last-branch, and vertical continuation, directories listed before files and each group alphabetical, h | 18 | 1.0 | 0.6666 | shard-00 | easy | [0.8889, 0.4444] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-44-47-696Z_01a03215-d8cf-7425-9597-5f909b36e2fb.json | skipped (--no-scan) | 01a03215-d8cf-7425-9597-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-44-47-696Z_01a03215-d8cf-7425-9597-5f909b36e2fb.jsonl | True | ok | Write your own timeout context manager for asyncio: class deadline(seconds) usable as 'async with deadline(0.05):' that cancels the enclosed work and raises TimeoutError, built on loop.call_later plus | 18 | 0.6 | 0.5833 | shard-00 | medium | [0.3333, 0.8333] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-47-09-418Z_01a03218-0269-75af-95e0-902e96a1b342.json | skipped (--no-scan) | 01a03218-0269-75af-95e0-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-47-09-418Z_01a03218-0269-75af-95e0-902e96a1b342.jsonl | True | ok | Per-key rate limiter with idle eviction: KeyedLimiter(rate, burst, max_keys=1000, clock) maintains an independent token bucket per key (user id) but must not grow without bound - evict the least-recen | 13 | 0.938 | 0.9231 | shard-00 | easy | [0.8462, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-16-08-705Z_01a03232-8c80-7584-bd8a-13b453c647f3.json | skipped (--no-scan) | 01a03232-8c80-7584-bd8a-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-16-08-705Z_01a03232-8c80-7584-bd8a-13b453c647f3.jsonl | True | ok | Quartile bucketing report in SQLite. orders(id, customer, amount_cents) seeded with 48 deterministic rows spanning 500 to 250000 cents. Using NTILE(4) OVER (ORDER BY amount_cents), produce quartile su | 17 | 1.0 | 0.7059 | shard-00 | easy | [0.4118, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-18-24-355Z_01a03234-9e62-708c-a599-270a5b1a9b01.json | skipped (--no-scan) | 01a03234-9e62-708c-a599-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-18-24-355Z_01a03234-9e62-708c-a599-270a5b1a9b01.jsonl | True | ok | Phone-number-keyed fuzzy dedup job. Generate data/leads.csv with 80 rows where the same real-world person appears up to 3 times with the phone formatted differently ((555) 123-4567 vs 5551234567 vs +1 | 19 | 0.688 | 0.8684 | shard-00 | easy | [0.7895, 0.9474] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-20-31-760Z_01a03236-900f-7577-b18a-f5cabc76c11c.json | skipped (--no-scan) | 01a03236-900f-7577-b18a-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-20-31-760Z_01a03236-900f-7577-b18a-f5cabc76c11c.jsonl | True | ok | Change-detection merge for a product catalog. Generator writes data/catalog_monday.csv and data/catalog_tuesday.csv (sku, name, price_cents, category; 50 skus Monday; Tuesday drops 3 skus, adds 4, cha | 19 | 0.938 | 0.4474 | shard-00 | medium | [0.8947, 0.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-23-41-143Z_01a03239-73d6-7602-80e6-03eaf775fa4f.json | skipped (--no-scan) | 01a03239-73d6-7602-80e6-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-23-41-143Z_01a03239-73d6-7602-80e6-03eaf775fa4f.jsonl | True | ok | Classic watermark-driven incremental load, simulated over three runs. Source: SQLite source.db table events(id INTEGER PRIMARY KEY, created_at TEXT, payload TEXT). Target: warehouse.db with events_cop | 19 | 1.0 | 0.2369 | shard-00 | hard | [0.0, 0.4737] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-27-19-844Z_01a0323c-ca23-75c7-9edd-41538cab6134.json | skipped (--no-scan) | 01a0323c-ca23-75c7-9edd-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-27-19-844Z_01a0323c-ca23-75c7-9edd-41538cab6134.jsonl | True | ok | Out-of-order sessionization. Generate a JSONL log where each user's events are shuffled within a 2-hour arrival jitter (seeded) so file order is NOT time order, and 5 events have identical (user, ts) | 19 | 0.688 | 0.7894 | shard-00 | easy | [0.5789, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-32-13-146Z_01a03241-43d9-7461-a328-dc7c1b8e3277.json | skipped (--no-scan) | 01a03241-43d9-7461-a328-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-32-13-146Z_01a03241-43d9-7461-a328-dc7c1b8e3277.jsonl | True | ok | In-memory B-tree of order 4 (max 3 keys per node) with insert, search, delete, and in-order iteration, in btree.py under 400 lines, stdlib only. Deletion must handle borrow-from-sibling and merge case | 16 | 0.938 | 0.1875 | shard-00 | hard | [0.0, 0.375] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-36-21-124Z_01a03245-0c83-70cf-8ebf-a2878420d1d9.json | skipped (--no-scan) | 01a03245-0c83-70cf-8ebf-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-36-21-124Z_01a03245-0c83-70cf-8ebf-a2878420d1d9.jsonl | True | ok | Consolidation migration: merge legacy tables customers_us(id, name, email, state) and customers_intl(id, name, email, country) into one customers table with a region column and NEW surrogate ids, pres | 20 | 1.0 | 0.65 | shard-00 | easy | [0.75, 0.55] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-37-31-060Z_01a03246-1db3-72f4-a7f0-4c22ef3c0e4a.json | skipped (--no-scan) | 01a03246-1db3-72f4-a7f0-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-37-31-060Z_01a03246-1db3-72f4-a7f0-4c22ef3c0e4a.jsonl | True | ok | Event-time windows with watermarks and allowed lateness. Generator emits 10k events with event_time and PROCESSING order shuffled: 90 percent near-in-order, 8 percent late within 30s, 2 percent late b | 20 | 0.75 | 0.5 | shard-00 | medium | [0.3, 0.7] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-47-38-563Z_01a0324f-62c2-71bb-9d34-abc455fe8b39.json | skipped (--no-scan) | 01a0324f-62c2-71bb-9d34-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-47-38-563Z_01a0324f-62c2-71bb-9d34-abc455fe8b39.jsonl | True | ok | Tenant config bleed. Put this in /workspace/tenantcfg.py: BASE_CONFIG = { "features": {"beta": False, "export": True}, "limits": {"rate": 100, "burst": 20}, "theme": "light", } def config_for(tenant_o | 17 | 1.0 | 0.8529 | shard-00 | easy | [0.7059, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-49-51-814Z_01a03251-6b45-7783-9ded-8f9790c57366.json | skipped (--no-scan) | 01a03251-6b45-7783-9ded-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-49-51-814Z_01a03251-6b45-7783-9ded-8f9790c57366.jsonl | True | ok | Make this rate limiter testable without a single sleep. Save /workspace/ratelimit.py: import time class SlidingWindowLimiter: def __init__(self, max_requests, window_seconds, clock=time.monotonic): se | 14 | 1.0 | 0.9643 | shard-00 | easy | [1.0, 0.9286] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-51-24-829Z_01a03252-d69c-712e-8024-2652d7ac2501.json | skipped (--no-scan) | 01a03252-d69c-712e-8024-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-51-24-829Z_01a03252-d69c-712e-8024-2652d7ac2501.jsonl | True | ok | Legacy refactor, characterization-first. Save the following as /workspace/sales_report.py exactly as given, then lock in its behavior with tests BEFORE changing a single line: import sys REGIONS = {} | 20 | 0.812 | 0.6 | shard-00 | easy | [0.6, 0.6] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-52-33-985Z_01a03253-e4c0-7705-b546-c853bddee697.json | skipped (--no-scan) | 01a03253-e4c0-7705-b546-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-52-33-985Z_01a03253-e4c0-7705-b546-c853bddee697.jsonl | True | ok | Boundary mutants in pricing tiers. Save /workspace/tiers.py: def unit_price_cents(qty): if qty <= 0: raise ValueError("qty must be positive") if qty >= 1000: return 70 if qty >= 100: return 80 if qty | 17 | 1.0 | 0.7941 | shard-00 | easy | [0.7647, 0.8235] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-56-02-296Z_01a03257-1277-7339-948e-85113e52b3e5.json | skipped (--no-scan) | 01a03257-1277-7339-948e-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-56-02-296Z_01a03257-1277-7339-948e-85113e52b3e5.jsonl | True | ok | Characterize, then refactor this bank reconciliation script. Save as /workspace/reconcile.py exactly: import sys import csv MATCHED = [] LEDGER_ONLY = [] BANK_ONLY = [] AMOUNT_MISMATCH = [] def run(le | 19 | 0.938 | 0.8947 | shard-00 | easy | [0.8947, 0.8947] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-57-51-622Z_01a03258-bd85-70d0-bcd0-1d44fd7ade92.json | skipped (--no-scan) | 01a03258-bd85-70d0-bcd0-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-57-51-622Z_01a03258-bd85-70d0-bcd0-1d44fd7ade92.jsonl | True | ok | Stateful property-based testing: the cache has a planted bug that only shows through operation SEQUENCES. Save /workspace/lru.py: from collections import OrderedDict class LRUCache: def __init__(self, | 14 | 0.8 | 0.8214 | shard-00 | easy | [0.8571, 0.7857] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T06-03-05-369Z_01a0325d-8718-75e1-8d7b-bd23a1fa60ac.json | skipped (--no-scan) | 01a0325d-8718-75e1-8d7b-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T06-03-05-369Z_01a0325d-8718-75e1-8d7b-bd23a1fa60ac.jsonl | True | ok | Write a custom lint rule as a standalone tool: nodup_imports.py walks Python files given as CLI args, parses them with the ast module, and reports (a) duplicate imports of the same module, (b) imports | 18 | 1.0 | 0.6945 | shard-00 | easy | [0.6111, 0.7778] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T06-03-06-583Z_01a0325d-8bd7-7080-aa1e-ac3f7e0d8ab6.json | skipped (--no-scan) | 01a0325d-8bd7-7080-aa1e-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T06-03-06-583Z_01a0325d-8bd7-7080-aa1e-ac3f7e0d8ab6.jsonl | True | ok | Implement a Markdown linter mdlint.py with three rules: MD001 heading levels must not skip (h1 then h3 is a violation), MD009 no trailing whitespace, MD040 fenced code blocks must declare a language. | 18 | 0.733 | 0.8889 | shard-00 | easy | [0.8889, 0.8889] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T06-07-54-756Z_01a03261-f183-7186-932c-716e537a8596.json | skipped (--no-scan) | 01a03261-f183-7186-932c-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T06-07-54-756Z_01a03261-f183-7186-932c-716e537a8596.jsonl | True | ok | Write install_hooks.py, a hook manager: it symlinks every executable file from a repo's hooks.d/ directory into .git/hooks/, backing up any pre-existing non-symlink hook to <name>.bak, and supports `- | 18 | 0.75 | 0.6945 | shard-00 | easy | [0.6667, 0.7222] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T06-12-01-107Z_01a03265-b3d3-7130-8858-7689d1377eec.json | skipped (--no-scan) | 01a03265-b3d3-7130-8858-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T06-12-01-107Z_01a03265-b3d3-7130-8858-7689d1377eec.jsonl | True | ok | Implement release_notes.py: given a git repo, it finds the latest two version tags matching vX.Y.Z (sort them by semver, not lexically, so v0.10.0 beats v0.9.0), collects commits between them, and wri | 17 | 0.875 | 0.8236 | shard-00 | easy | [0.8824, 0.7647] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T06-20-29-174Z_01a0326d-7475-71ca-a86c-4f82c1f447b8.json | skipped (--no-scan) | 01a0326d-7475-71ca-a86c-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T06-20-29-174Z_01a0326d-7475-71ca-a86c-4f82c1f447b8.jsonl | True | ok | Small task: gitignore_gen.py takes template names as args (support python, node, go with sensible hardcoded pattern sets embedded in the script) and writes a combined .gitignore with section header co | 19 | 0.75 | 0.8947 | shard-00 | easy | [0.8947, 0.8947] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T06-55-15-182Z_01a0328d-48ed-738c-8cb0-73cc3dffe2d6.json | skipped (--no-scan) | 01a0328d-48ed-738c-8cb0-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T06-55-15-182Z_01a0328d-48ed-738c-8cb0-73cc3dffe2d6.jsonl | True | ok | Small library plus CLI in one file: glob2re.py exposes translate(pattern) turning a shell glob with *, ?, [abc], [!abc], and [a-z] into an anchored Python regex, and a main() that reads candidate stri | 15 | 0.938 | 0.6 | shard-00 | easy | [0.6, 0.6] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T07-02-06-635Z_01a03293-902a-7594-a386-521f62d92fe4.json | skipped (--no-scan) | 01a03293-902a-7594-a386-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T07-02-06-635Z_01a03293-902a-7594-a386-521f62d92fe4.jsonl | True | ok | Molar mass calculator with a recursive-descent formula parser supporting nested parentheses and dot-hydrate notation, e.g. molar_mass('CuSO4.5H2O'). Inline atomic weight table: H 1.008, C 12.011, N 14 | 20 | 0.812 | 0.975 | shard-00 | easy | [1.0, 0.95] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T07-04-08-313Z_01a03295-6b78-719f-bdb0-0f53ea695656.json | skipped (--no-scan) | 01a03295-6b78-719f-bdb0-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T07-04-08-313Z_01a03295-6b78-719f-bdb0-0f53ea695656.jsonl | True | ok | Multi-sample VCF genotype summarizer. Write gt_stats.py parsing a VCF whose records carry GT as the first FORMAT key. Create this fixture inline (tabs between columns): header '#CHROM POS ID REF ALT Q | 15 | 0.875 | 0.8667 | shard-00 | easy | [0.8667, 0.8667] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T07-06-07-430Z_01a03297-3cc5-72ae-b026-6defde9b97bd.json | skipped (--no-scan) | 01a03297-3cc5-72ae-b026-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T07-06-07-430Z_01a03297-3cc5-72ae-b026-6defde9b97bd.jsonl | True | ok | Gas-law solver CLI in Python. `gas.py ideal --solve V --n 1 --T 273.15 --P 1` solves PV=nRT with R=0.082057 LΒ·atm/(molΒ·K) for whichever variable is omitted; a --units si mode uses R=8.314 with Pa and | 19 | 0.938 | 0.9737 | shard-00 | easy | [0.9474, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T07-06-58-745Z_01a03298-0538-72a5-ab0f-e0ad643975fc.json | skipped (--no-scan) | 01a03298-0538-72a5-ab0f-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T07-06-58-745Z_01a03298-0538-72a5-ab0f-e0ad643975fc.jsonl | True | ok | Primer QC tool. primer_check.py takes a primer sequence and prints a report with: length, GC fraction, Wallace Tm = 2(A+T)+4(G+C), salt-adjusted Tm = 81.5 + 16.6*log10([Na+]) + 0.41*(GC%) - 600/N with | 18 | 0.938 | 0.9445 | shard-00 | easy | [1.0, 0.8889] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T07-22-11-939Z_01a032a5-f462-75e9-9415-7658829979ed.json | skipped (--no-scan) | 01a032a5-f462-75e9-9415-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T07-22-11-939Z_01a032a5-f462-75e9-9415-7658829979ed.jsonl | True | ok | English draughts (checkers) move generator with forced captures and multi-jumps, Python. Movegen must enforce mandatory capture and chain jumps, with kinging on the back row ending the jump sequence. | 20 | 0.75 | 0.05 | shard-00 | hard | [0.05, 0.05] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T07-28-41-975Z_01a032ab-e7f6-77ea-9721-f55cd941300a.json | skipped (--no-scan) | 01a032ab-e7f6-77ea-9721-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T07-28-41-975Z_01a032ab-e7f6-77ea-9721-f55cd941300a.jsonl | True | ok | RFM over SQLite with a view, verified against Python. Create schema.sql defining orders(customer_id, order_date, amount) and a view rfm_view computing recency in days from '2025-08-20', frequency, and | 17 | 0.812 | 0.6765 | shard-00 | easy | [0.6471, 0.7059] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T07-33-57-655Z_01a032b0-b917-7563-9184-60808838dfba.json | skipped (--no-scan) | 01a032b0-b917-7563-9184-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T07-33-57-655Z_01a032b0-b917-7563-9184-60808838dfba.jsonl | True | ok | Build a sample-size-aware A/B analyzer with a simulation check. abkit/power.py: required_n_per_arm(p_base, mde_abs, alpha=0.05, power=0.8) using the standard two-proportion formula with normal quantil | 14 | 0.875 | 0.9285 | shard-00 | easy | [0.8571, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-06-28-858Z_01a032ce-7ef9-779e-b4b2-18820be6b2e1.json | skipped (--no-scan) | 01a032ce-7ef9-779e-b4b2-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-06-28-858Z_01a032ce-7ef9-779e-b4b2-18820be6b2e1.jsonl | True | ok | Security audit finding: our password-policy tests would still pass if the policy were badly broken. Save /workspace/policy.py: import string def check_password(pw): problems = [] if len(pw) < 10: prob | 17 | 1.0 | 0.853 | shard-00 | easy | [0.9412, 0.7647] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-14-58-831Z_01a032d6-470f-7188-8495-5467dc94ee7c.json | skipped (--no-scan) | 01a032d6-470f-7188-8495-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-14-58-831Z_01a032d6-470f-7188-8495-5467dc94ee7c.jsonl | True | ok | Quick tool: histogram_report.py reads a scores.csv with header 'student,score', prints count, mean, median, and population standard deviation each to 2 decimals, then a bucket histogram for 0-59, 60-6 | 19 | 0.812 | 0.7369 | shard-00 | easy | [1.0, 0.4737] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-21-11-447Z_01a032db-f696-75c6-b008-6072e24066f8.json | skipped (--no-scan) | 01a032db-f696-75c6-b008-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-21-11-447Z_01a032db-f696-75c6-b008-6072e24066f8.jsonl | True | ok | Pinyin tone converter: numbered syllables to diacritics. pinyin.py: convert(syllable) turns forms like 'ni3', 'hao3', 'lΓΌ4' (also accept 'lv4' as ΓΌ), 'xiong2', 'er2', 'ma' (no digit = neutral tone, un | 19 | 0.938 | 0.9474 | shard-00 | easy | [0.9474, 0.9474] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-22-45-460Z_01a032dd-65d4-7417-bd02-c59e18a8b018.json | skipped (--no-scan) | 01a032dd-65d4-7417-bd02-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-22-45-460Z_01a032dd-65d4-7417-bd02-c59e18a8b018.jsonl | True | ok | Minimal-pair finder for pronunciation drills. Inline phonemic lexicon (word -> phoneme tuple) of 40 entries, e.g. ship /Κ Ιͺ p/, sheep /Κ iΛ p/, chip /tΚ Ιͺ p/, cheap /tΚ iΛ p/, bat /b Γ¦ t/, pat /p Γ¦ t/ | 18 | 0.857 | 0.8611 | shard-00 | easy | [0.8333, 0.8889] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-27-26-870Z_01a032e1-b116-7616-bf37-fe61f75168f2.json | skipped (--no-scan) | 01a032e1-b116-7616-bf37-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-27-26-870Z_01a032e1-b116-7616-bf37-fe61f75168f2.jsonl | True | ok | Expert build: translation-memory system with fuzzy lookup. Package tm/: store.py (SQLite table of segment pairs source->target with a source_norm column), match.py (similarity(a, b) = 1 - levenshtein( | 14 | 0.625 | 0.8215 | shard-00 | easy | [0.9286, 0.7143] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-42-51-744Z_01a032ef-cde0-7266-8fb3-1ccd09b8e88a.json | skipped (--no-scan) | 01a032ef-cde0-7266-8fb3-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-42-51-744Z_01a032ef-cde0-7266-8fb3-1ccd09b8e88a.jsonl | True | ok | Fixed asset register with disposals. assets.py: register acquisitions (id, cost_cents, salvage_cents, life_years, acquired_year), compute straight-line book value at any year end (full-year convention | 19 | 0.875 | 0.9737 | shard-00 | easy | [0.9474, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T09-19-22-978Z_01a03311-3d61-74c0-8852-8eec08eddc96.json | skipped (--no-scan) | 01a03311-3d61-74c0-8852-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T09-19-22-978Z_01a03311-3d61-74c0-8852-8eec08eddc96.jsonl | True | ok | Path smoothing by line-of-sight string pulling, Python. First get an A* path on this map (4-connected, # wall, - open): S--------- -########- ---------- -########- ---------G Then smooth it: greedily | 21 | 0.625 | 0.8095 | shard-00 | easy | [0.9048, 0.7143] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T09-23-00-940Z_01a03314-90cb-758e-ae7d-4d1ce644a47d.json | skipped (--no-scan) | 01a03314-90cb-758e-ae7d-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T09-23-00-940Z_01a03314-90cb-758e-ae7d-4d1ce644a47d.jsonl | True | ok | Othello/Reversi engine verified by perft, Python. Implement legal move generation with flipping in all 8 directions and the pass rule (a player with no moves passes; two passes end the game). From the | 17 | 0.688 | 0.4117 | shard-00 | medium | [0.2353, 0.5882] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T10-08-10-968Z_01a0333d-ead7-74c7-a5a3-1f6a462a8666.json | skipped (--no-scan) | 01a0333d-ead7-74c7-a5a3-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T10-08-10-968Z_01a0333d-ead7-74c7-a5a3-1f6a462a8666.jsonl | True | ok | Grid-city ambulance dispatch simulator, educational emergency-ops modeling. City is a 20x20 grid; 4 ambulance bases at (2,2), (2,17), (17,2), (17,17); travel time = Manhattan distance in minutes. Call | 18 | 0.812 | 0.1111 | shard-00 | hard | [0.2222, 0.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T10-08-48-604Z_01a0333e-7ddb-7770-b5f5-b3f64edc21bb.json | skipped (--no-scan) | 01a0333e-7ddb-7770-b5f5-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T10-08-48-604Z_01a0333e-7ddb-7770-b5f5-b3f64edc21bb.jsonl | True | ok | Discharge-summary templating engine, educational document tooling with strict safety rails. Template syntax: {{field}} substitutes a value; {{field|upper}} and {{field|date:%d %b %Y}} apply filters; u | 17 | 0.938 | 0.7647 | shard-00 | easy | [null, 0.7647] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T10-10-30-066Z_01a03340-0a31-7273-aa98-f3d8d01a341e.json | skipped (--no-scan) | 01a03340-0a31-7273-aa98-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T10-10-30-066Z_01a03340-0a31-7273-aa98-f3d8d01a341e.jsonl | True | ok | Event-sourced immunization registry, expert educational build with an invented vaccine EXV. Append-only event log (JSON lines) with event types DoseRecorded {patient, date, lot}, DoseInvalidated {pati | 20 | 0.812 | 0.2 | shard-00 | hard | [0.2, 0.2] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T10-26-31-546Z_01a0334e-b5f9-70ac-a5e4-b3bce2604076.json | skipped (--no-scan) | 01a0334e-b5f9-70ac-a5e4-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T10-26-31-546Z_01a0334e-b5f9-70ac-a5e4-b3bce2604076.jsonl | True | ok | Local-time notice effectiveness. The invented rule: an emailed termination notice is effective the moment sent if sent before 17:00 in the RECIPIENT's timezone on a business day; otherwise it is effec | 15 | 0.812 | 0.9667 | shard-00 | easy | [1.0, 0.9333] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T10-29-09-873Z_01a03351-2070-755a-b716-78616bfdb2a5.json | skipped (--no-scan) | 01a03351-2070-755a-b716-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T10-29-09-873Z_01a03351-2070-755a-b716-78616bfdb2a5.jsonl | True | ok | Bates cross-check auditor. Given a directory of stamped text files (page footers hold stamps like DRX-0001042; pages separated by form-feed) and a manifest.csv claiming doc_id,first,last per file, xch | 20 | 0.875 | 0.275 | shard-00 | hard | [0.0, 0.55] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T11-15-49-040Z_01a0337b-d6b0-74b5-a0f8-c9415a07f2e2.json | skipped (--no-scan) | 01a0337b-d6b0-74b5-a0f8-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T11-15-49-040Z_01a0337b-d6b0-74b5-a0f8-c9415a07f2e2.jsonl | True | ok | Glob pattern compiler in Python: glob2re.py translating glob patterns to regular expressions. Support * (no slash crossing), ? (single non-slash char), [seq] and [!seq] character classes (with ranges, | 15 | 1.0 | 0.4334 | shard-00 | medium | [0.2, 0.6667] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T11-56-50-602Z_01a033a1-6629-7162-af73-b2aeec1994d5.json | skipped (--no-scan) | 01a033a1-6629-7162-af73-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T11-56-50-602Z_01a033a1-6629-7162-af73-b2aeec1994d5.jsonl | True | ok | Pendulum period beyond the small-angle approximation. Write pendulum_period.py computing T(theta0) two ways: (1) numeric quadrature of the exact elliptic integral T = 4*sqrt(L/g)*K(sin^2(theta0/2)) us | 18 | 0.688 | 0.8611 | shard-00 | easy | [0.7222, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T12-09-47-702Z_01a033ad-41b5-75cd-96ea-77b9a464f713.json | skipped (--no-scan) | 01a033ad-41b5-75cd-96ea-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T12-09-47-702Z_01a033ad-41b5-75cd-96ea-77b9a464f713.jsonl | True | ok | Effectiveness-NTU vs discretized simulation for a counterflow heat exchanger. Hot stream: water m_dot=0.5 kg/s, cp=4180, T_in=90 C; cold stream: m_dot=0.8 kg/s, cp=4180, T_in=20 C; UA=6000 W/K. Implem | 19 | 0.625 | 0.7369 | shard-00 | easy | [1.0, 0.4737] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T12-20-42-706Z_01a033b7-4051-75b7-b5d5-845c0668f880.json | skipped (--no-scan) | 01a033b7-4051-75b7-b5d5-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T12-20-42-706Z_01a033b7-4051-75b7-b5d5-845c0668f880.jsonl | True | ok | Maximum power transfer demonstrator. Using a DC nodal solver (write a minimal one supporting R and V elements, ground node 0), build source_with_internal_resistance(Vs=10, Rs=50) and sweep load resist | 19 | 0.812 | 0.3947 | shard-00 | medium | [0.7895, 0.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T12-23-33-702Z_01a033b9-dc46-75d5-8aeb-9dae1ac97205.json | skipped (--no-scan) | 01a033b9-dc46-75d5-8aeb-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T12-23-33-702Z_01a033b9-dc46-75d5-8aeb-9dae1ac97205.jsonl | True | ok | TDD: here is the complete test file, test_units_core.py β write units_core.py to satisfy it without touching the tests. import pytest from units_core import Q, DimensionError def test_add_like(): asse | 17 | 0.875 | 0.9412 | shard-00 | easy | [0.9412, 0.9412] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T14-14-33-740Z_01a0341f-7c0c-7449-9364-0d3cfbc28de4.json | skipped (--no-scan) | 01a0341f-7c0c-7449-9364-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T14-14-33-740Z_01a0341f-7c0c-7449-9364-0d3cfbc28de4.jsonl | True | ok | Write an auto-grading library grader.py plus a pytest suite, Python stdlib only. It grades three question types via grade(question_spec, student_answer) returning 1.0 or 0.0. Rules: (a) multiple_choic | 20 | 0.812 | 0.975 | shard-00 | easy | [1.0, 0.95] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T14-14-54-306Z_01a0341f-cc61-7322-9471-348efeb354ee.json | skipped (--no-scan) | 01a0341f-cc61-7322-9471-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T14-14-54-306Z_01a0341f-cc61-7322-9471-348efeb354ee.jsonl | True | ok | Build a mastery-progress tracker around a course prerequisite DAG. Python package layout: mastery/graph.py, mastery/progress.py, tests/. graph.py: add_course(id), add_prereq(course, prereq), and valid | 19 | 0.875 | 0.7369 | shard-00 | easy | [0.4737, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T14-14-56-868Z_01a0341f-d663-7225-9c83-8c4cfaf1e25f.json | skipped (--no-scan) | 01a0341f-d663-7225-9c83-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T14-14-56-868Z_01a0341f-d663-7225-9c83-8c4cfaf1e25f.jsonl | True | ok | Build an algebra worksheet generator with a machine-verified answer key. gen_algebra.py produces, from a seed, N problems of two kinds: 'ax + b = c' with a in 2-12, x in -9..9 nonzero, b in -20..20 (c | 16 | 0.875 | 0.9375 | shard-00 | easy | [1.0, 0.875] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T14-26-42-614Z_01a0342a-9b36-711d-a08c-3a946d93670b.json | skipped (--no-scan) | 01a0342a-9b36-711d-a08c-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T14-26-42-614Z_01a0342a-9b36-711d-a08c-3a946d93670b.jsonl | True | ok | Cost-volume-profit analysis tool. cvp.py: breakeven_units(fixed, price, var_cost) with ValueError when contribution margin <= 0; breakeven_revenue; units_for_target_profit; margin_of_safety(actual_uni | 17 | 1.0 | 0.9706 | shard-00 | easy | [0.9412, 1.0] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T16-34-39-166Z_01a0349f-bdbd-73cf-835b-991330ff14af.json | skipped (--no-scan) | 01a0349f-bdbd-73cf-835b-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T16-34-39-166Z_01a0349f-bdbd-73cf-835b-991330ff14af.jsonl | True | ok | Lab-panel validator with unit conversion, educational chemistry over synthetic panels. Conversion factors inline: glucose mg/dL to mmol/L divide by 18.0; creatinine mg/dL to umol/L multiply by 88.4; c | 10 | 0.938 | 0.9 | shard-00 | easy | [0.9, 0.9] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T17-08-03-815Z_01a034be-5466-7433-93d3-2e693203208b.json | skipped (--no-scan) | 01a034be-5466-7433-93d3-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T17-08-03-815Z_01a034be-5466-7433-93d3-2e693203208b.jsonl | True | ok | Build a TLV (tag-length-value) codec in Python modeled on DER structure. tlv.py: a tag byte where bit 5 (0x20) marks constructed, low bits are the tag number; length is definite only: short form for l | 18 | 0.812 | 0.5 | shard-00 | medium | [0.4444, 0.5556] |
scale/trajectories/shard-00/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T17-10-58-486Z_01a034c0-feb5-76b6-b20a-18fb28cd070f.json | skipped (--no-scan) | 01a034c0-feb5-76b6-b20a-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T17-10-58-486Z_01a034c0-feb5-76b6-b20a-18fb28cd070f.jsonl | True | ok | Mini append-only record log with crash recovery, Python stdlib. Package reclog/ with log.py and tests. Format: each record on disk is u32 LE payload length, u32 LE CRC32 of payload (zlib.crc32), paylo | 18 | 0.625 | 0.5 | shard-00 | medium | [0.7778, 0.2222] |
scale/trajectories/shard-00/choucsan__mimo-claude-code-traces-1k__c316b552.json | skipped (--no-scan) | 0b708fe1-5869-45f0-a499-build | build | hf://datasets/choucsan/mimo-claude-code-traces-1k/session/code_generation/c316b552.jsonl | True | ok | Write a Python function called `retry_with_backoff` that parses mathematical expressions with operator precedence. Include comprehensive test cases. | 19 | 0.75 | 0.7369 | shard-00 | easy | [0.6842, 0.7895] |
scale/trajectories/shard-00/choucsan__mimo-claude-code-traces-1k__5351643f.json | skipped (--no-scan) | 425cb74a-7853-4e3b-b72b-build | build | hf://datasets/choucsan/mimo-claude-code-traces-1k/session/code_generation/5351643f.jsonl | True | ok | Implement a Python class called `ConnectionPool` that provides publish/subscribe messaging with topic-based routing. It should support: insert, delete, and range_query. Include type hints and docstrin | 19 | 1.0 | 0.7368 | shard-00 | easy | [0.7368, 0.7368] |
scale/trajectories/shard-00/choucsan__mimo-claude-code-traces-1k__451d8e6d.json | skipped (--no-scan) | c9016c22-de70-49d4-9708-build | build | hf://datasets/choucsan/mimo-claude-code-traces-1k/session/algorithms/451d8e6d.jsonl | True | ok | Implement A* pathfinding algorithm in Python. Analyze its time and space complexity. | 20 | 0.625 | 0.8 | shard-00 | easy | [0.8, 0.8] |
scale/trajectories/shard-01/codex300__rollout-2026-07-03T06-23-53-019f26a5-e43f-7191-aad1-a9e251064bcd.json | skipped (--no-scan) | 019f26a5-e43f-7191-aad1-build | build | hf://datasets/AletheiaResearch/GPT-5.5-Codex/rollout-2026-07-03T06-23-53-019f26a5-e43f-7191-aad1-a9e251064bcd.jsonl | True | ok | Build a webhook receiver service that validates signatures, stores events, retries failed processing, and exposes a status endpoint. | 20 | 0.625 | 0.2 | shard-01 | hard | [0.4, 0.0] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-09-05-961Z_01a031f5-2aa8-773f-874f-ecc3a60b8349.json | skipped (--no-scan) | 01a031f5-2aa8-773f-874f-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-09-05-961Z_01a031f5-2aa8-773f-874f-ecc3a60b8349.jsonl | True | ok | Write wc.py, a Python 3 stdlib-only clone of wc supporting -l, -w, and -c flags plus multiple file arguments, reading stdin when no files are given. Counts are space-separated with the filename append | 19 | 1.0 | 0.6316 | shard-01 | easy | [0.4211, 0.8421] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-15-26-165Z_01a031fa-f7d5-77ac-aed4-72f57a437f6f.json | skipped (--no-scan) | 01a031fa-f7d5-77ac-aed4-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-15-26-165Z_01a031fa-f7d5-77ac-aed4-72f57a437f6f.jsonl | True | ok | Terse spec: commlite.py behaves like comm FILE1 FILE2 with -1 -2 -3 suppression flags and proper tab-indented columns. Inputs must be sorted; detect unsorted input, warn on stderr, and exit 1 after pr | 19 | 0.812 | 0.2105 | shard-01 | hard | [0.3684, 0.0526] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-25-21-722Z_01a03204-0e39-70f2-a265-f7c1cd49a5a5.json | skipped (--no-scan) | 01a03204-0e39-70f2-a265-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-25-21-722Z_01a03204-0e39-70f2-a265-f7c1cd49a5a5.jsonl | True | ok | Build kvlite.py, a persistent key-value CLI backed by an append-only log db.log with one record per line: operation, key, base64 value where applicable, tab-separated, with a tombstone record type for | 19 | 0.875 | 0.3158 | shard-01 | medium | [0.2632, 0.3684] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-28-42-143Z_01a03207-1d1e-70ca-87ad-8bf3c739a2dd.json | skipped (--no-scan) | 01a03207-1d1e-70ca-87ad-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-28-42-143Z_01a03207-1d1e-70ca-87ad-8bf3c739a2dd.jsonl | True | ok | teelite.py mirrors stdin to stdout and to each FILE argument, with -a to append. It must survive a downstream pipe closing: on BrokenPipeError from stdout, keep writing the remaining stream to the fil | 16 | 0.625 | 0.625 | shard-01 | easy | [0.625, 0.625] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-30-05-365Z_01a03208-6235-709c-8406-90169c53d885.json | skipped (--no-scan) | 01a03208-6235-709c-8406-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-30-05-365Z_01a03208-6235-709c-8406-90169c53d885.jsonl | True | ok | rotate.py CONF.json implements size-based log rotation: for each configured entry {path, max_bytes, keep, compress}, when path exceeds max_bytes shift path.2 to path.3 and so on up to keep (deleting t | 18 | 0.75 | 0.8611 | shard-01 | easy | [0.7222, 1.0] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-41-38-361Z_01a03212-f538-74f1-a0c9-87eaaf1e50d9.json | skipped (--no-scan) | 01a03212-f538-74f1-a0c9-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-41-38-361Z_01a03212-f538-74f1-a0c9-87eaaf1e50d9.jsonl | True | ok | A teammate left this buggy shell-style argument splitter; save it as shsplit.py: def split_args(s): out, cur, q = [], '', None for ch in s: if q: if ch == q: q = None else: cur += ch elif ch in '"\'': | 16 | 1.0 | 0.75 | shard-01 | easy | [0.75, 0.75] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T04-47-42-186Z_01a03218-8269-713c-b9dd-99ca4778b85b.json | skipped (--no-scan) | 01a03218-8269-713c-b9dd-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T04-47-42-186Z_01a03218-8269-713c-b9dd-99ca4778b85b.jsonl | True | ok | Accuracy shoot-out: sliding-window-counter approximation vs exact sliding log. Implement both (clock injected), then write harness.py that replays three deterministic traffic patterns (steady 8 rps, b | 15 | 0.625 | 0.8 | shard-01 | easy | [0.8, 0.8] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-24-11-091Z_01a03239-e8d2-7353-8ffb-c6e6652c63c6.json | skipped (--no-scan) | 01a03239-e8d2-7353-8ffb-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-24-11-091Z_01a03239-e8d2-7353-8ffb-c6e6652c63c6.jsonl | True | ok | File-arrival ledger loader. Directory landing/ receives CSV files; build a generator creating day1: a.csv, b.csv; day2: c.csv plus a MODIFIED b.csv (same name, new content, later mtime and different s | 18 | 0.875 | 0.5555 | shard-01 | medium | [0.6111, 0.5] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T05-52-23-691Z_01a03253-bc8a-7082-9e9a-129b15448e72.json | skipped (--no-scan) | 01a03253-bc8a-7082-9e9a-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T05-52-23-691Z_01a03253-bc8a-7082-9e9a-129b15448e72.jsonl | True | ok | Mutation-hardening drill. Save /workspace/leap.py: def is_leap(year): if year % 400 == 0: return True if year % 100 == 0: return False return year % 4 == 0 and /workspace/tests/test_leap.py containing | 13 | 0.875 | 0.9616 | shard-01 | easy | [1.0, 0.9231] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T06-02-18-012Z_01a0325c-ce1b-73a1-9f2a-9bab5860ec37.json | skipped (--no-scan) | 01a0325c-ce1b-73a1-9f2a-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T06-02-18-012Z_01a0325c-ce1b-73a1-9f2a-9bab5860ec37.jsonl | True | ok | Create changelog.py: a stdlib-only Python tool that generates CHANGELOG.md from conventional commits (feat:, fix:, docs:, chore:, feat!: or BREAKING CHANGE footer) via `git log`. Construct a test repo | 18 | 0.875 | 0.9166 | shard-01 | easy | [0.8333, 1.0] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T06-04-16-711Z_01a0325e-9dc7-77d2-a78d-6175bbf53a59.json | skipped (--no-scan) | 01a0325e-9dc7-77d2-a78d-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T06-04-16-711Z_01a0325e-9dc7-77d2-a78d-6175bbf53a59.jsonl | True | ok | Write dep_audit.py: parses a requirements.txt (handle ==, >=, ~= pins, comments, blank lines, environment markers after ';' which you may ignore for matching) and checks each pinned package against an | 19 | 0.75 | 0.7631 | shard-01 | easy | [0.7368, 0.7895] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T06-10-21-835Z_01a03264-300a-74de-a235-cbb8c042c974.json | skipped (--no-scan) | 01a03264-300a-74de-a235-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T06-10-21-835Z_01a03264-300a-74de-a235-cbb8c042c974.jsonl | True | ok | Implement a size-tiered log compactor: compact_logs.py scans a logs/ dir for app-YYYY-MM-DD.log files; files older than 7 days get gzipped in place, older than 30 days get merged per-month into app-YY | 16 | 0.625 | 0.5625 | shard-01 | medium | [0.5625, 0.5625] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T06-57-35-071Z_01a0328f-6b5e-743b-b2af-a4e9bda50a09.json | skipped (--no-scan) | 01a0328f-6b5e-743b-b2af-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T06-57-35-071Z_01a0328f-6b5e-743b-b2af-a4e9bda50a09.jsonl | True | ok | Implement translate(dna, frame=0, strict=False) in Python using the standard genetic code with the full 64-entry codon table written inline in the source as a dict. Translation stops at the first stop | 18 | 1.0 | 0.9445 | shard-01 | easy | [1.0, 0.8889] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T07-15-34-305Z_01a0329f-e321-7535-b3e0-bd1418fbd87e.json | skipped (--no-scan) | 01a0329f-e321-7535-b3e0-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T07-15-34-305Z_01a0329f-e321-7535-b3e0-bd1418fbd87e.jsonl | True | ok | Spatial hash broadphase in Python. `spatialhash.py` buckets circle centers into cell_size grid cells and yields candidate pairs from each cell and its 8 neighbors, deduplicated. Test harness: seed ran | 16 | 0.938 | 0.9688 | shard-01 | easy | [0.9375, 1.0] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T07-18-24-417Z_01a032a2-7ba0-75e9-a61e-66b6211954d3.json | skipped (--no-scan) | 01a032a2-7ba0-75e9-a61e-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T07-18-24-417Z_01a032a2-7ba0-75e9-a61e-66b6211954d3.jsonl | True | ok | This Euler-method SIR integrator misbehaves with its shipped settings (beta=1.5, gamma=0.5, dt=1.0, N=100, I0=1, 60 steps): def run_sir(beta=1.5, gamma=0.5, N=100, I0=1, dt=1.0, steps=60): S, I, R = N | 18 | 0.812 | 0.9166 | shard-01 | easy | [1.0, 0.8333] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T07-34-59-372Z_01a032b1-aa2b-72bb-94c8-b54f2b268f62.json | skipped (--no-scan) | 01a032b1-aa2b-72bb-94c8-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T07-34-59-372Z_01a032b1-aa2b-72bb-94c8-b54f2b268f62.jsonl | True | ok | Behavior tree for a guard NPC, Python. Implement Sequence, Selector, Inverter, and condition/action leaves returning SUCCESS/FAILURE/RUNNING with proper RUNNING resumption (restart-from-running-child | 17 | 0.938 | 0.2647 | shard-01 | hard | [0.2353, 0.2941] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-08-02-997Z_01a032cf-eeb4-7593-9b21-83cec54f5396.json | skipped (--no-scan) | 01a032cf-eeb4-7593-9b21-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-08-02-997Z_01a032cf-eeb4-7593-9b21-83cec54f5396.jsonl | True | ok | Write bench_guard.py, a performance-regression gate for CI: it runs `python3 -m timeit`-style measurements of named snippets defined in bench.json ({name, setup, stmt, budget_ms}) using timeit program | 18 | 0.625 | 0.8334 | shard-01 | easy | [0.7778, 0.8889] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-09-46-748Z_01a032d1-83fc-742b-b5fa-c7f67c0985b0.json | skipped (--no-scan) | 01a032d1-83fc-742b-b5fa-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-09-46-748Z_01a032d1-83fc-742b-b5fa-c7f67c0985b0.jsonl | True | ok | Write an auto-grading library grader.py plus a pytest suite, Python stdlib only. It grades three question types via grade(question_spec, student_answer) returning 1.0 or 0.0. Rules: (a) multiple_choic | 20 | 0.812 | 0.375 | shard-01 | medium | [0.0, 0.75] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-11-34-615Z_01a032d3-2956-74e3-9abc-b3aa30d412b9.json | skipped (--no-scan) | 01a032d3-2956-74e3-9abc-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-11-34-615Z_01a032d3-2956-74e3-9abc-b3aa30d412b9.jsonl | True | ok | I need a roster importer for a gradebook. Write make_roster.py that generates roster.csv with 200 rows and columns student_id,last_name,first_name,email,section, deliberately corrupting 12 known rows | 19 | 0.688 | 0.9737 | shard-01 | easy | [1.0, 0.9474] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-17-52-159Z_01a032d8-ec1f-7051-aaaf-b90d2e443c72.json | skipped (--no-scan) | 01a032d8-ec1f-7051-aaaf-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-17-52-159Z_01a032d8-ec1f-7051-aaaf-b90d2e443c72.jsonl | True | ok | Transliteration engine, Russian Cyrillic to Latin, with round-trip guarantees on the unambiguous core. Mapping table inline in translit.py: Π°>a Π±>b Π²>v Π³>g Π΄>d Π΅>e ΠΆ>zh Π·>z ΠΈ>i ΠΉ>j ΠΊ>k Π»>l ΠΌ>m Π½>n ΠΎ>o | 18 | 0.875 | 0.6667 | shard-01 | easy | [0.6667, 0.6667] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-19-51-992Z_01a032da-c037-7170-820d-d2719e0ac747.json | skipped (--no-scan) | 01a032da-c037-7170-820d-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-19-51-992Z_01a032da-c037-7170-820d-d2719e0ac747.jsonl | True | ok | Mini hyphenation engine in the Knuth-Liang style with an inline pattern set. hyphen.py: patterns are strings with digits marking break weights, e.g. '1na', 'n2at', '1tion', '2io', 'h1yph', 'hy3ph', 'h | 11 | 0.688 | 0.4546 | shard-01 | medium | [0.3636, 0.5455] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-21-09-965Z_01a032db-f0cd-70db-a2aa-a0e626ebc2ba.json | skipped (--no-scan) | 01a032db-f0cd-70db-a2aa-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-21-09-965Z_01a032db-f0cd-70db-a2aa-a0e626ebc2ba.jsonl | True | ok | Morphological analyzer for the invented agglutinative language 'Voru'. Word structure is STEM + case + number + possessive, in that order, all suffixes optional but order fixed. Inline suffix tables: | 17 | 1.0 | 0.9706 | shard-01 | easy | [1.0, 0.9412] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T08-40-26-164Z_01a032ed-9534-7223-a74c-841a110b6eb9.json | skipped (--no-scan) | 01a032ed-9534-7223-a74c-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T08-40-26-164Z_01a032ed-9534-7223-a74c-841a110b6eb9.jsonl | True | ok | Accounts receivable aging report. aging.py ingests invoices (id, customer, date, amount) and payments (invoice_id, date, amount) applied to invoices; open balance is amount minus applied payments. buc | 19 | 0.875 | 0.7106 | shard-01 | easy | [0.6316, 0.7895] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T09-13-44-606Z_01a0330c-139e-7084-ac91-63f287c55a0f.json | skipped (--no-scan) | 01a0330c-139e-7084-ac91-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T09-13-44-606Z_01a0330c-139e-7084-ac91-63f287c55a0f.jsonl | True | ok | Write a lockstep command-queue simulator in Python. `world.py` holds units with integer positions; commands (move dx,dy; halt) are scheduled for specific tick numbers. `step(world)` applies all comman | 19 | 0.812 | 0.1053 | shard-01 | hard | [0.1053, 0.1053] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T09-15-17-984Z_01a0330d-805f-7768-a468-dcc8a10dcfc2.json | skipped (--no-scan) | 01a0330d-805f-7768-a468-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T09-15-17-984Z_01a0330d-805f-7768-a468-dcc8a10dcfc2.jsonl | True | ok | AABB collision with minimal-penetration resolution, Python stdlib. `aabb.py`: Aabb(x, y, w, h), overlap test, and resolve(a, b) that pushes box a out of static box b along the axis of least penetratio | 19 | 0.875 | 0.9474 | shard-01 | easy | [0.9474, 0.9474] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T09-24-47-836Z_01a03316-325b-7095-b51e-f049ba4be213.json | skipped (--no-scan) | 01a03316-325b-7095-b51e-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T09-24-47-836Z_01a03316-325b-7095-b51e-f049ba4be213.jsonl | True | ok | Cribbage hand scorer, Python. Score a 4-card hand plus starter: fifteens (2 pts each), pairs (2), runs (length 3+, with multiplicities), flush (4 in hand, 5 with starter), his nobs (jack matching star | 17 | 0.875 | 0.7353 | shard-01 | easy | [0.7059, 0.7647] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T09-33-05-119Z_01a0331d-c8de-71df-bb84-3fcd1f2d2c5d.json | skipped (--no-scan) | 01a0331d-c8de-71df-bb84-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T09-33-05-119Z_01a0331d-c8de-71df-bb84-3fcd1f2d2c5d.jsonl | True | ok | Rollback netcode core, Python β the hard one. Deterministic sim of 2 fighters (integer positions, moves, a hit landing when in range on an attack frame). Keep a ring buffer of the last 30 tick snapsho | 18 | 0.688 | 0.3889 | shard-01 | medium | [0.3333, 0.4444] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T09-40-46-649Z_01a03324-d3b9-76db-9180-e6d5cac37847.json | skipped (--no-scan) | 01a03324-d3b9-76db-9180-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T09-40-46-649Z_01a03324-d3b9-76db-9180-e6d5cac37847.jsonl | True | ok | Data-driven adventure with winnability proof, Python β hard. Define a world in `world.json` (6+ rooms, items, locked doors keyed by items, one win room) and an engine that loads it. Then write a solve | 18 | 0.625 | 0.9444 | shard-01 | easy | [0.9444, 0.9444] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T10-05-31-735Z_01a0333b-7cd6-7757-a280-ed54450b3dd6.json | skipped (--no-scan) | 01a0333b-7cd6-7757-a280-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T10-05-31-735Z_01a0333b-7cd6-7757-a280-ed54450b3dd6.jsonl | True | ok | Syndromic-surveillance outbreak detector, an educational EARS-style C2 algorithm on synthetic counts. Daily counts for 120 days: baseline Poisson-like values you generate with random.Random(13) around | 19 | 0.625 | 0.9474 | shard-01 | easy | [1.0, 0.8947] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T10-06-03-228Z_01a0333b-f7db-778e-9979-d00cb43b86e7.json | skipped (--no-scan) | 01a0333b-f7db-778e-9979-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T10-06-03-228Z_01a0333b-f7db-778e-9979-d00cb43b86e7.jsonl | True | ok | Bed-occupancy forecaster with a holdout error budget, educational hospital-ops analytics. Data generation (seeded, random.Random(17)): 365 daily census values built as base 220 + weekly seasonality (s | 18 | 0.688 | 0.8889 | shard-01 | easy | [0.8889, 0.8889] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T10-06-39-666Z_01a0333c-8631-72a7-af15-afe609e32384.json | skipped (--no-scan) | 01a0333c-8631-72a7-af15-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T10-06-39-666Z_01a0333c-8631-72a7-af15-afe609e32384.jsonl | True | ok | Operating-theatre list packer, educational scheduling optimization. One theatre, day runs 08:00-18:00; every case needs its stated duration plus a fixed 20-minute turnover after it (no turnover needed | 19 | 0.938 | 0.6316 | shard-01 | easy | [0.7368, 0.5263] |
scale/trajectories/shard-01/TeichAI__Ox-Alpha-Pi-Traces__026-08-24T10-17-59-602Z_01a03346-e632-744c-ad91-3771af9230a7.json | skipped (--no-scan) | 01a03346-e632-744c-ad91-build | build | hf://datasets/TeichAI/Ox-Alpha-Pi-Traces/2026-08-24T10-17-59-602Z_01a03346-e632-744c-ad91-3771af9230a7.jsonl | True | ok | Write a license-compatibility checker. Inline knowledge base β save as licenses.json: compat matrix (row may link against column): MIT->everything yes; Apache-2.0->everything except GPL-2.0-only; GPL- | 20 | 0.812 | 0.825 | shard-01 | easy | [0.9, 0.75] |
End of preview. Expand in Data Studio
RL tasks generated with traj2envs
270 Harbor tasks under tasks/, each with its environment image pinned in task.toml ([environment].docker_image, with opencode preinstalled), a top-level difficulty and its solve rate in [metadata]. manifest.parquet lists them with their family. sources.json lists the trajectories they were built from: block fetching them when training with internet. Tasks from the same trajectory can be near-duplicates: split train/eval by source_trace.
- Downloads last month
- -