The dataset viewer is not available for this split.
Error code: FeaturesError
Exception: ArrowInvalid
Message: JSON parse error: Invalid value. in row 0
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 324, in _generate_tables
df = pandas_read_json(f)
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 38, in pandas_read_json
return pd.read_json(path_or_buf, **kwargs)
~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 815, in read_json
return json_reader.read()
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1014, in read
obj = self._get_object_parser(self.data)
File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1040, in _get_object_parser
obj = FrameParser(json, **kwargs).parse()
File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1176, in parse
self._parse()
~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1392, in _parse
ujson_loads(json, precise_float=self.precise_float), dtype=None
~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: Trailing data
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/split/first_rows.py", line 244, in compute_first_rows_from_streaming_response
iterable_dataset = iterable_dataset._resolve_features()
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 4408, in _resolve_features
features = _infer_features_from_batch(self.with_format(None)._head())
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2679, in _head
return next(iter(self.iter(batch_size=n)))
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2861, in iter
for key, pa_table in ex_iterable.iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2395, in _iter_arrow
yield from 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 327, in _generate_tables
raise e
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 290, in _generate_tables
pa_table = paj.read_json(
io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
)
File "pyarrow/_json.pyx", line 342, in pyarrow._json.read_json
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: JSON parse error: Invalid value. in row 0Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
EvolveBench v11 — judge traces (Claude Opus 5.5, Qwen3.7-Plus)
Grounded-judge traces for the 78-task EvolveBench v11 suite
(tasks_version_v11_20260922). Every score in these archives was produced by the
same judge: the vendored record-rules-v12-access-recovery grounding harness
reached through PYTHONPATH=<task>/environment, scored by gpt-5.4.
Contents
| file | agent | tasks | notes |
|---|---|---|---|
v11_claude_opus55_rejudge_gpt54_redacted.zip |
Claude Code · Opus 5.5, effort High | 78 | host-side rejudge, late fallback rungs enabled |
v11_qwen37_judge_gpt54_redacted.zip |
Qwen Code · Qwen3.7-Plus (OpenRouter) | 78 | host-side judge; see caveats |
Each task directory holds judge_runs.json (full record-level verdicts and captured
evidence), judge_trajectory.json, judge_summary.json, reward.json,
feedback.txt, rejudge.log.
Headline numbers
Reward = overall_score / 5, averaged over all 78 tasks with agent failures counted
as zero.
| agent | overall | Lookup | Comparison | Planning | Analysis | Custom | runtime |
|---|---|---|---|---|---|---|---|
| Claude Code · Opus 5.5 (High) | 74.8 | 76.5 | 74.9 | 77.3 | 67.4 | 76.5 | 7:59 ± 6:34 |
| Qwen Code · Qwen3.7-Plus | 53.6 | 53.7 | 51.4 | 58.6 | 50.4 | 55.8 | 8:18 ± 5:55 |
Task-type columns key on metadata.primary_type in each task's task.toml
(n = 16 / 25 / 13 / 11 / 13).
Caveats that matter for comparison
- Reasoning effort is not matched. Claude ran at effort High. Qwen Code exposes no reasoning-effort dial, so its row carries none. Gemini's adapter caps turn 1 at 600s against Qwen's 2250s, a 3.75x turn-1 compute difference.
- Failure modes differ in kind, not just degree. Qwen lost 11/78 tasks to agent-side give-ups — 8 client-side tool-call repetition halts, 1 upstream repetition rejection, 1 turn-1 timeout, 1 unparseable summary — versus 1/78 for Claude. A large share of the 21-point gap is the agent stopping, not answering badly.
- One stale task. Qwen's
task-ar-010-job-application-trackerretains its pre-repair judge output (its rejudge exceeded the 3600s judge timeout twice) and is flagged with arejudge_notefield. It also holds 70 of that run's 113source_access_blockedrecords; excluding it the blocked residual is 43/1901 = 2.3%, against 3.4% for Claude.
Redaction
These archives are redacted for public release. Scores, record-level verdicts, claim text, URLs and evidence structure are unchanged — every one of the 156 task rewards is byte-identical to the unredacted originals, and all 588 JSON files parse strictly.
Replaced inside string values only:
| pattern | replacement | why |
|---|---|---|
AIzaSy… |
AIzaSy__THIRD_PARTY_KEY_REDACTED__ |
Google API keys belonging to third parties, scraped from captured pages (e.g. Maps tile URLs). 26k+ occurrences. |
| email addresses | __EMAIL_REDACTED__ |
third-party addresses appearing in captured page content |
/Users/icy |
/home/user |
local filesystem paths |
provider key shapes (sk-…, hf_…) |
placeholder | defensive; none of ours were present |
No credentials belonging to this project appear in these traces.
Gemini
A Gemini CLI row is not included. The first attempt was cut short when the provider route ran out of quota, and a clean 78-task run is pending. A partial result would invite citation of a number that should not be cited.
- Downloads last month
- 141