Hugging Face's logo

Datasets:
tts-sft
/
round3-fb-se

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
problems: list<item: struct<orig_prompt: string, gt: null, candidates: list<item: string>, candidate_groups: l (... 85 chars omitted)
  child 0, item: struct<orig_prompt: string, gt: null, candidates: list<item: string>, candidate_groups: list<item: l (... 73 chars omitted)
      child 0, orig_prompt: string
      child 1, gt: null
      child 2, candidates: list<item: string>
          child 0, item: string
      child 3, candidate_groups: list<item: list<item: null>>
          child 0, item: list<item: null>
              child 0, item: null
      child 4, routing_details: null
      child 5, question: string
      child 6, options: null
metrics: struct<loop: int64, note: string>
  child 0, loop: int64
  child 1, note: string
units_total_problems: int64
units: list<item: struct<unit: string, bucket: string, phase: int64, max_evolution_loops: int64, LOOPS: int (... 59 chars omitted)
  child 0, item: struct<unit: string, bucket: string, phase: int64, max_evolution_loops: int64, LOOPS: int64, n_probl (... 47 chars omitted)
      child 0, unit: string
      child 1, bucket: string
      child 2, phase: int64
      child 3, max_evolution_loops: int64
      child 4, LOOPS: int64
      child 5, n_problems: int64
      child 6, n_spj: int64
      child 7, problem_loops: int64
phase2_units: list<item: struct<unit: string, bucket: string, phase: int64, max_evolution_loops: int64, LOOPS: int (... 59 chars omitted)
  child 0, item: struct<unit: string, bucket: string, phase: int64, max_evolution_loops: int64, LOOPS: int64, n_probl (... 47 chars omitted)
      child 0, unit: string
      child 1, bucket: string
      child 2, phase: int64
      child 3, max_evolution_loops: int64
      child 4, LOOPS: int64
      child 5, n_problems: int64
      child 6, n_spj: int64
      child 7, problem_loops: int64
phase2_units_total_problems: int64
units_total_problem_loops: int64
phase2_units_total_problem_loops: int64
note: string
to
{'units': List({'unit': Value('string'), 'bucket': Value('string'), 'phase': Value('int64'), 'max_evolution_loops': Value('int64'), 'LOOPS': Value('int64'), 'n_problems': Value('int64'), 'n_spj': Value('int64'), 'problem_loops': Value('int64')}), 'phase2_units': List({'unit': Value('string'), 'bucket': Value('string'), 'phase': Value('int64'), 'max_evolution_loops': Value('int64'), 'LOOPS': Value('int64'), 'n_problems': Value('int64'), 'n_spj': Value('int64'), 'problem_loops': Value('int64')}), 'units_total_problems': Value('int64'), 'units_total_problem_loops': Value('int64'), 'phase2_units_total_problems': Value('int64'), 'phase2_units_total_problem_loops': Value('int64'), 'note': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, 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 129, 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 489, 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 2818, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2355, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2380, 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 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 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              problems: list<item: struct<orig_prompt: string, gt: null, candidates: list<item: string>, candidate_groups: l (... 85 chars omitted)
                child 0, item: struct<orig_prompt: string, gt: null, candidates: list<item: string>, candidate_groups: list<item: l (... 73 chars omitted)
                    child 0, orig_prompt: string
                    child 1, gt: null
                    child 2, candidates: list<item: string>
                        child 0, item: string
                    child 3, candidate_groups: list<item: list<item: null>>
                        child 0, item: list<item: null>
                            child 0, item: null
                    child 4, routing_details: null
                    child 5, question: string
                    child 6, options: null
              metrics: struct<loop: int64, note: string>
                child 0, loop: int64
                child 1, note: string
              units_total_problems: int64
              units: list<item: struct<unit: string, bucket: string, phase: int64, max_evolution_loops: int64, LOOPS: int (... 59 chars omitted)
                child 0, item: struct<unit: string, bucket: string, phase: int64, max_evolution_loops: int64, LOOPS: int64, n_probl (... 47 chars omitted)
                    child 0, unit: string
                    child 1, bucket: string
                    child 2, phase: int64
                    child 3, max_evolution_loops: int64
                    child 4, LOOPS: int64
                    child 5, n_problems: int64
                    child 6, n_spj: int64
                    child 7, problem_loops: int64
              phase2_units: list<item: struct<unit: string, bucket: string, phase: int64, max_evolution_loops: int64, LOOPS: int (... 59 chars omitted)
                child 0, item: struct<unit: string, bucket: string, phase: int64, max_evolution_loops: int64, LOOPS: int64, n_probl (... 47 chars omitted)
                    child 0, unit: string
                    child 1, bucket: string
                    child 2, phase: int64
                    child 3, max_evolution_loops: int64
                    child 4, LOOPS: int64
                    child 5, n_problems: int64
                    child 6, n_spj: int64
                    child 7, problem_loops: int64
              phase2_units_total_problems: int64
              units_total_problem_loops: int64
              phase2_units_total_problem_loops: int64
              note: string
              to
              {'units': List({'unit': Value('string'), 'bucket': Value('string'), 'phase': Value('int64'), 'max_evolution_loops': Value('int64'), 'LOOPS': Value('int64'), 'n_problems': Value('int64'), 'n_spj': Value('int64'), 'problem_loops': Value('int64')}), 'phase2_units': List({'unit': Value('string'), 'bucket': Value('string'), 'phase': Value('int64'), 'max_evolution_loops': Value('int64'), 'LOOPS': Value('int64'), 'n_problems': Value('int64'), 'n_spj': Value('int64'), 'problem_loops': Value('int64')}), 'units_total_problems': Value('int64'), 'units_total_problem_loops': Value('int64'), 'phase2_units_total_problems': Value('int64'), 'phase2_units_total_problem_loops': Value('int64'), 'note': 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.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

round3-fb-se — round-3 feedback-SE run units (gpt-oss-120B only)

Feedback-SE evolution loops on the round-3 screen pool, restarting from the screen run's clean loop-0 populations (16 candidates/problem). Loop 0 is NOT re-run.

  • Buckets: checker-corrected 120B screen buckets (screen_per_problem.spjfix.csv); bucket 16 (saturated) excluded. 2,833 problems total.
  • Loop caps (evidence: round-2 rerun vs matched independent): bucket 0 → 8 evolution loops, 1-3 / 4-7 / 8-11 → 6, 12-15 → 4. LOOPS in unit.env = 1 + evo.
  • Phases: phase 1 = r3fb_120b_{b00,b13,b47,b811}_u* (13 units, 2,345 problems, 16,290 problem-loops). Phase 2 = r3fb_120b_b1215_u00..01 (488 problems, 1,952 problem-loops) — run AFTER all phase-1 units.
  • SPJ-aware feedback (new): 217 problems carry a certified special-judge checker inline in their fb_tests.jsonl row; the shipped lcb_public_probe_harness_spj.py consults it on stdout mismatch, so any-valid-answer solutions no longer receive bogus "wrong answer" feedback. Non-SPJ rows behave bit-identically to the fixed (5d921c4) harness. Feedback tests = first 8 of the 50 grading tests (LCB_FB_FULLTESTS unset — subset wording).
  • Runner: run_shard_r3_fb.sh (copy into <tts-sft>/handoff/round2/, and copy lcb_public_probe_harness_spj.py into <tts-sft>/scripts/). Requires the round-3 minipatch (stdin-buffer fix) already applied. One node = one unit: UNIT=<DATA>/120b/<name> RUN_DIR=<out>/<name> bash run_shard_r3_fb.sh. Resume = re-run the same command.
  • Returns: upload each whole RUN_DIR (loops.out.jsonl + ck_nonsat/ + *.raw.json + fb_audit.jsonl + run.log). Grading happens on the cutting side (checker-aware).

The compute-matched independent-sampling arm is cut AFTER these runs return, from this run's token ledger (same harvest-at-16 method as round 2).

Unit layout is identical to tts-sft/round2-rerun-fb (seeds.jsonl, fb_tests.jsonl, ck_nonsat/_loop0.json, unit.env, unit.json), prefix r3fb_120b_.
Downloads last month
318
Total file size:
17.8 GB