Hugging Face's logo

Datasets:
Ilikemechuri
/
ikp_agent

ArXiv:
License:
Dataset card Files
xet
 Community
Dataset Viewer
Duplicate
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:    CastError
Message:      Couldn't cast
at: string
type: string
call_id: string
command: string
elapsed_seconds: double
exit_code: int64
timed_out: bool
output: string
output_clipped: bool
response_id: string
answer: string
task_id: string
tool_calls: int64
run_id: string
question: string
answer_key: struct<probe_id: string, tier: string, domain: string, gold_answer: string>
  child 0, probe_id: string
  child 1, tier: string
  child 2, domain: string
  child 3, gold_answer: string
model: string
to
{'run_id': Value('string'), 'task_id': Value('string'), 'question': Value('string'), 'model': Value('string'), 'tool_calls': Value('int64'), 'elapsed_seconds': Value('float64'), 'answer': Value('string'), 'answer_key': {'probe_id': Value('string'), 'tier': Value('string'), 'domain': Value('string'), 'gold_answer': Value('string')}}
because column names don't match
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 2951, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2461, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2486, 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 547, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 430, 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 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              at: string
              type: string
              call_id: string
              command: string
              elapsed_seconds: double
              exit_code: int64
              timed_out: bool
              output: string
              output_clipped: bool
              response_id: string
              answer: string
              task_id: string
              tool_calls: int64
              run_id: string
              question: string
              answer_key: struct<probe_id: string, tier: string, domain: string, gold_answer: string>
                child 0, probe_id: string
                child 1, tier: string
                child 2, domain: string
                child 3, gold_answer: string
              model: string
              to
              {'run_id': Value('string'), 'task_id': Value('string'), 'question': Value('string'), 'model': Value('string'), 'tool_calls': Value('int64'), 'elapsed_seconds': Value('float64'), 'answer': Value('string'), 'answer_key': {'probe_id': Value('string'), 'tier': Value('string'), 'domain': Value('string'), 'gold_answer': Value('string')}}
              because column names don't match

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.

 
https://github.com/mechuchuchu

IKP terminal research prototype

This prototype gives an OpenAI model one custom run_terminal function. The host process handles OpenAI API calls; shell commands run in a separate Docker container with no API key and only a per-run workspace mounted. The host removes the key from its environment after creating the API client, so Docker CLI subprocesses do not inherit it. It does not enable OpenAI's built-in web-search tool.

The sample tasks are a reproducible random sample of five questions from /root/ikp/data/probes/final_probe_set_v8.json. tasks/tasks.json is sent to the model. tasks/answer_key.json stays in the host-side project and is only copied to the result after the model finishes.

Setup

From this directory:

uv pip install -r requirements.txt
docker build -t ikp-terminal-agent:latest .
export OPENAI_API_KEY='your-key'

Select a different random task sample with:

uv run python select_tasks.py --count 5 --seed 20261016

Run one task with a model available to your API project:

uv run python agent.py --task task-01 --model YOUR_MODEL_ID

Set CONTAINER_ENGINE=podman to use a compatible Podman installation.

Prototype limits

  • The system prompt allows public search-engine webpages (including DuckDuckGo) and source-site search pages, but bans search APIs and instructs the model to follow robots.txt. This first version records commands for later review but does not enforce those rules itself.
  • Commands have a 30-second default timeout, tool calls are capped at 15 per task, and model-visible command output is clipped at 16,000 characters.
  • The container can access the public internet. This prototype does not yet filter private-network destinations or enforce a per-request HTTP budget.
  • Each task run gets a fresh workspace. The run trace and answer are saved under runs/; these files are ignored by Git.
cication:https://arxiv.org/abs/2604.24827
Downloads last month
-
Total file size:
471 kB

Paper for Ilikemechuri/ikp_agent