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
backend: string
train_errors: struct<>
train_metrics: struct<aligntune_result: struct<training_time: double, final_loss: double, model_path: string, steps (... 437 chars omitted)
  child 0, aligntune_result: struct<training_time: double, final_loss: double, model_path: string, steps: int64, epochs: int64, t (... 17 chars omitted)
      child 0, training_time: double
      child 1, final_loss: double
      child 2, model_path: string
      child 3, steps: int64
      child 4, epochs: int64
      child 5, task_type: string
  child 1, final_eval: struct<eval_loss: double>
      child 0, eval_loss: double
  child 2, train_loss_log: list<item: list<item: double>>
      child 0, item: list<item: double>
          child 0, item: double
  child 3, eval_loss_by_epoch: list<item: list<item: double>>
      child 0, item: list<item: double>
          child 0, item: double
  child 4, train_summary: struct<train_runtime: double, train_samples_per_second: double, train_steps_per_second: double, tota (... 63 chars omitted)
      child 0, train_runtime: double
      child 1, train_samples_per_second: double
      child 2, train_steps_per_second: double
      child 3, total_flos: double
      child 4, train_loss: double
      child 5, epoch: double
      child 6, step: int64
  child 5, best_model_checkpoint: string
  child 6, oom_skipped_microbatches: list<item: null>
      child 0, item: null
validation_miro_bandai: struct<base: struct<n_pairs: int64, overall: struct<correct: int64, wrong: in
...
   child 6, estimated_reductions: struct<n: int64, correct: int64, wrong: int64, conflict-unresolved: int64, correctly-abstained: int6 (... 122 chars omitted)
              child 0, n: int64
              child 1, correct: int64
              child 2, wrong: int64
              child 3, conflict-unresolved: int64
              child 4, correctly-abstained: int64
              child 5, hallucinated: int64
              child 6, accuracy_incl_abstention: double
              child 7, stated_n: int64
              child 8, stated_correct: int64
              child 9, stated_accuracy: double
note: string
loftq_config: struct<>
fan_in_fan_out: bool
alpha_pattern: struct<>
layers_pattern: null
eva_config: null
r: int64
megatron_config: null
corda_config: null
arrow_config: null
lora_bias: bool
use_qalora: bool
use_dora: bool
lora_alpha: int64
lora_dropout: double
rank_pattern: struct<>
velora_config: null
bias: string
lora_ga_config: null
megatron_core: string
alora_invocation_tokens: null
exclude_modules: null
layer_replication: null
task_type: string
ensure_weight_tying: bool
use_bdlora: null
layers_to_transform: null
revision: null
kasa_config: null
peft_version: string
modules_to_save: null
monteclora_config: null
target_parameters: null
init_lora_weights: bool
target_modules: list<item: string>
  child 0, item: string
auto_mapping: null
qalora_group_size: int64
use_rslora: bool
peft_type: string
base_model_name_or_path: string
trainable_token_indices: null
inference_mode: bool
to
{'alora_invocation_tokens': Value('null'), 'alpha_pattern': {}, 'arrow_config': Value('null'), 'auto_mapping': Value('null'), 'base_model_name_or_path': Value('string'), 'bias': Value('string'), 'corda_config': Value('null'), 'ensure_weight_tying': Value('bool'), 'eva_config': Value('null'), 'exclude_modules': Value('null'), 'fan_in_fan_out': Value('bool'), 'inference_mode': Value('bool'), 'init_lora_weights': Value('bool'), 'kasa_config': Value('null'), 'layer_replication': Value('null'), 'layers_pattern': Value('null'), 'layers_to_transform': Value('null'), 'loftq_config': {}, 'lora_alpha': Value('int64'), 'lora_bias': Value('bool'), 'lora_dropout': Value('float64'), 'lora_ga_config': Value('null'), 'megatron_config': Value('null'), 'megatron_core': Value('string'), 'modules_to_save': Value('null'), 'monteclora_config': Value('null'), 'peft_type': Value('string'), 'peft_version': Value('string'), 'qalora_group_size': Value('int64'), 'r': Value('int64'), 'rank_pattern': {}, 'revision': Value('null'), 'target_modules': List(Value('string')), 'target_parameters': Value('null'), 'task_type': Value('string'), 'trainable_token_indices': Value('null'), 'use_bdlora': Value('null'), 'use_dora': Value('bool'), 'use_qalora': Value('bool'), 'use_rslora': Value('bool'), 'velora_config': Value('null')}
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 544, 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 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, 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 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, 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 361, 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 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              backend: string
              train_errors: struct<>
              train_metrics: struct<aligntune_result: struct<training_time: double, final_loss: double, model_path: string, steps (... 437 chars omitted)
                child 0, aligntune_result: struct<training_time: double, final_loss: double, model_path: string, steps: int64, epochs: int64, t (... 17 chars omitted)
                    child 0, training_time: double
                    child 1, final_loss: double
                    child 2, model_path: string
                    child 3, steps: int64
                    child 4, epochs: int64
                    child 5, task_type: string
                child 1, final_eval: struct<eval_loss: double>
                    child 0, eval_loss: double
                child 2, train_loss_log: list<item: list<item: double>>
                    child 0, item: list<item: double>
                        child 0, item: double
                child 3, eval_loss_by_epoch: list<item: list<item: double>>
                    child 0, item: list<item: double>
                        child 0, item: double
                child 4, train_summary: struct<train_runtime: double, train_samples_per_second: double, train_steps_per_second: double, tota (... 63 chars omitted)
                    child 0, train_runtime: double
                    child 1, train_samples_per_second: double
                    child 2, train_steps_per_second: double
                    child 3, total_flos: double
                    child 4, train_loss: double
                    child 5, epoch: double
                    child 6, step: int64
                child 5, best_model_checkpoint: string
                child 6, oom_skipped_microbatches: list<item: null>
                    child 0, item: null
              validation_miro_bandai: struct<base: struct<n_pairs: int64, overall: struct<correct: int64, wrong: in
              ...
                 child 6, estimated_reductions: struct<n: int64, correct: int64, wrong: int64, conflict-unresolved: int64, correctly-abstained: int6 (... 122 chars omitted)
                            child 0, n: int64
                            child 1, correct: int64
                            child 2, wrong: int64
                            child 3, conflict-unresolved: int64
                            child 4, correctly-abstained: int64
                            child 5, hallucinated: int64
                            child 6, accuracy_incl_abstention: double
                            child 7, stated_n: int64
                            child 8, stated_correct: int64
                            child 9, stated_accuracy: double
              note: string
              loftq_config: struct<>
              fan_in_fan_out: bool
              alpha_pattern: struct<>
              layers_pattern: null
              eva_config: null
              r: int64
              megatron_config: null
              corda_config: null
              arrow_config: null
              lora_bias: bool
              use_qalora: bool
              use_dora: bool
              lora_alpha: int64
              lora_dropout: double
              rank_pattern: struct<>
              velora_config: null
              bias: string
              lora_ga_config: null
              megatron_core: string
              alora_invocation_tokens: null
              exclude_modules: null
              layer_replication: null
              task_type: string
              ensure_weight_tying: bool
              use_bdlora: null
              layers_to_transform: null
              revision: null
              kasa_config: null
              peft_version: string
              modules_to_save: null
              monteclora_config: null
              target_parameters: null
              init_lora_weights: bool
              target_modules: list<item: string>
                child 0, item: string
              auto_mapping: null
              qalora_group_size: int64
              use_rslora: bool
              peft_type: string
              base_model_name_or_path: string
              trainable_token_indices: null
              inference_mode: bool
              to
              {'alora_invocation_tokens': Value('null'), 'alpha_pattern': {}, 'arrow_config': Value('null'), 'auto_mapping': Value('null'), 'base_model_name_or_path': Value('string'), 'bias': Value('string'), 'corda_config': Value('null'), 'ensure_weight_tying': Value('bool'), 'eva_config': Value('null'), 'exclude_modules': Value('null'), 'fan_in_fan_out': Value('bool'), 'inference_mode': Value('bool'), 'init_lora_weights': Value('bool'), 'kasa_config': Value('null'), 'layer_replication': Value('null'), 'layers_pattern': Value('null'), 'layers_to_transform': Value('null'), 'loftq_config': {}, 'lora_alpha': Value('int64'), 'lora_bias': Value('bool'), 'lora_dropout': Value('float64'), 'lora_ga_config': Value('null'), 'megatron_config': Value('null'), 'megatron_core': Value('string'), 'modules_to_save': Value('null'), 'monteclora_config': Value('null'), 'peft_type': Value('string'), 'peft_version': Value('string'), 'qalora_group_size': Value('int64'), 'r': Value('int64'), 'rank_pattern': {}, 'revision': Value('null'), 'target_modules': List(Value('string')), 'target_parameters': Value('null'), 'task_type': Value('string'), 'trainable_token_indices': Value('null'), 'use_bdlora': Value('null'), 'use_dora': Value('bool'), 'use_qalora': Value('bool'), 'use_rslora': Value('bool'), 'velora_config': Value('null')}
              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.

Tally: training logs and aggregate scores

Logs and aggregate results for Tally, a quote-grounded Verra carbon-credit extractor fine-tuned from CohereLabs/tiny-aya-global (Lexsi AI Safety Hackathon, Track 1). Code and write-up: github.com/davissekai/tally-lexsi, including WRITEUP.md. Submission adapter: tally-v8-lora, whose repo also holds the v8 training metrics and provenance.

File Contents
heldout/scores_a100_bf16_aggregate.json Final held-out scores, A100 bf16, merge v1 and merge v2, five models
heldout/scores_aggregate.json Earlier fp16 / 4-bit held-out runs
benchmark/bench_a100.py Script for the A100 benchmark (ships no held-out data)
benchmark/bench_aggregate.json Earlier 4-bit benchmark settings and scores
auditkit/auditkit_v8_aggregate.json AuditKIT 1.1.12 aggregate scores for v8 vs base on a 40-example validation slice
auditkit/auditkit_v8.py Script for the AuditKIT run (ships no prompts, predictions or document text)
v7_training/ v7 training metrics, trainer log, LoRA config, validation comparison
provenance/lexsi_provenance.json AlignTune provenance for v7

Final held-out scores

Three unseen projects, 126 chunks, 20 stated values. One A100, bf16, greedy, max_new_tokens=900. Merge v1 and v2 use the same saved outputs.

Model Merge v1 Merge v2 Hallucinated Conflicts open (v2)
Tally v8 6/20 8/20 0 2
Qwen3-8B 4/20 6/20 0 10
Tally v7 3/20 5/20 0 6
Base Tiny Aya 2/20 3/20 0 9
Gemma 3 12B 3/20 3/20 0 13

Merge v2 rules were written after error analysis on the held-out set; on validation documents they lost no correct answers. One answer moves the score by 5 points.

AuditKIT (validation slice)

AuditKIT 1.1.12, 40 validation examples (not the held-out documents), same A100, greedy decoding. Aggregate metrics only; no prompts or outputs are in this dataset.

Metric Base Tiny Aya Tally v8
Field F1 0.12 0.71
Exact match 0/40 17/40
Valid JSON 37/40 40/40
Lexical groundedness 0.82 0.75

Red-team checks on v8: 39 of 47 passed (jailbreak 16/16, refusal 5/5, encoding 8/10, prompt injection 10/16).

Kept private: the held-out answer key and per-answer outputs, so the test stays clean. Training data: tally-sft-data.

Downloads last month
58

Collection including renhuang001/tally-eval-logs