Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: JSONDecodeError
Message: Expecting property name enclosed in double quotes: line 1 column 2 (char 1)
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 179, in _generate_tables
training_examples = convert_traces_to_training_data(trace_file)
File "/usr/local/lib/python3.14/site-packages/teich/converter.py", line 3635, in convert_traces_to_training_data
rows.extend(_convert_jsonl_file_to_training_rows(path))
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^
File "/usr/local/lib/python3.14/site-packages/teich/converter.py", line 3596, in _convert_jsonl_file_to_training_rows
rows = load_trace_file(jsonl_file)
File "/usr/local/lib/python3.14/site-packages/teich/converter.py", line 3029, in load_trace_file
events.append(json.loads(line))
~~~~~~~~~~^^^^^^
File "/usr/local/lib/python3.14/json/__init__.py", line 352, in loads
return _default_decoder.decode(s)
~~~~~~~~~~~~~~~~~~~~~~~^^^
File "/usr/local/lib/python3.14/json/decoder.py", line 345, in decode
obj, end = self.raw_decode(s, idx=_w(s, 0).end())
~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/json/decoder.py", line 361, in raw_decode
obj, end = self.scan_once(s, idx)
~~~~~~~~~~~~~~^^^^^^^^
json.decoder.JSONDecodeError: Expecting property name enclosed in double quotes: line 1 column 2 (char 1)Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Distillation stream traces
Tool-call traces from open models working on SWE-bench Verified tasks. The AI Alignment @ Illinois distillation stream records them to test whether a student model copies how its teacher acts with tools. Each run is one fresh attempt at one task in pi 0.85.1, run through Harbor at commit 99218a4.
Runs
| Folder | Model | Tasks | Solved | Date |
|---|---|---|---|---|
qwen3.8-27b/first-traces-v0 |
Qwen/Qwen3.8-27B, revision 1d4bf0f2 |
10 | 9 | 2026-10-01 |
The 10 tasks were drawn by a fixed rule: the SWE-bench Verified tasks marked "<15 min fix", sorted by id, then random.Random(0).sample(tasks, 10).
Files in each run folder
records.jsonl: one line per task withtask_id,model,harness,steps(each tool call with its arguments, result and timing),n_steps, andresolvedfrom Harbor's grader.trajectories/<task>/trajectory.json: the full run in ATIF, Harbor's trajectory format.trajectories/<task>/result.json: Harbor's trial result, with the grader's reward.trajectories/<task>/pi.txtandpi/: pi's raw event log and session files, including the model's thinking.run.md: every setting of the run and what went wrong.analysis.md: what the traces show.
These are first runs with no baseline, one run per task. The model ran on our own server with thinking on and its default sampling settings. SWE-bench tasks and the repositories in them keep their own licenses.
- Downloads last month
- -