TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1

---
dataset_info:
- config_name: evals_eval_rl
features:
- name: question
dtype: string
- name: answer
dtype: string
- name: task_config
dtype: string
- name: task_source
dtype: string
- name: prompt
list:
- name: content
dtype: string
- name: role
dtype: string
- name: model_responses
list: 'null'
- name: model_responses__eval_is_correct
list: 'null'
- name: all_other_columns
dtype: string
- name: original_split
dtype: string
- name: metadata
dtype: string
- name: model_responses__best_of_n_atags
list: string
- name: model_responses__best_of_n_atags__finish_reason_length_flags
list: bool
- name: model_responses__best_of_n_atags__length_partial_responses
list: string
- name: prompt__best_of_n_atags__metadata
struct:
- name: api_url
dtype: string
- name: backend
dtype: string
- name: chat_template_applied
dtype: bool
- name: generation_params
struct:
- name: chat_template_applied
dtype: bool
- name: max_tokens
dtype: int64
- name: n
dtype: int64
- name: repetition_penalty
dtype: float64
- name: temperature
dtype: float64
- name: top_k
dtype: int64
- name: top_p
dtype: float64
- name: model_name
dtype: string
- name: prompt
list:
- name: content
dtype: string
- name: role
dtype: string
- name: model_responses__best_of_n_atags__metadata
struct:
- name: backend
dtype: string
- name: model_name
dtype: string
- name: n_responses
dtype: int64
- name: model_responses__best_of_n_atags__eval_is_correct
list: bool
- name: model_responses__best_of_n_atags__eval_extracted_answers
list: string
- name: model_responses__best_of_n_atags__eval_extraction_metadata
dtype: string
- name: model_responses__best_of_n_atags__eval_evaluation_metadata
dtype: string
- name: model_responses__best_of_n_atags__internal_answers__eval_is_correct
list:
list: bool
- name: model_responses__best_of_n_atags__internal_answers__eval_extracted_answers
list:
list: string
- name: model_responses__best_of_n_atags__internal_answers__eval_extraction_metadata
dtype: string
- name: model_responses__best_of_n_atags__internal_answers__eval_evaluation_metadata
dtype: string
- name: model_responses__best_of_n_atags__metrics
struct:
- name: flips_by
list: int64
- name: flips_total
dtype: int64
- name: num_correct
dtype: int64
- name: pass_at_n
dtype: int64
- name: percent_correct
dtype: float64
- name: total_responses
dtype: int64
- name: eval_date
dtype: string
- name: split
dtype: string
- name: revision_name
dtype: string
- name: model_path
dtype: string
- name: checkpoint_step
dtype: int64
- name: stage_name
dtype: string
- name: stage_number
dtype: int64
- name: timestamp
dtype: string
- name: eval_repo_id
dtype: string
splits:
- name: test
num_bytes: 35316600
num_examples: 1000
download_size: 4998744
dataset_size: 35316600
- config_name: logs__evaluation_eval_rl
features:
- name: timestamp
dtype: string
- name: end_timestamp
dtype: string
- name: stage_name
dtype: string
- name: stage_number
dtype: int64
- name: level
dtype: string
- name: message
dtype: string
- name: stdout_content
dtype: string
- name: stderr_content
dtype: string
- name: experiment_name
dtype: string
- name: elapsed_time_seconds
dtype: float64
- name: stage_complete
dtype: bool
splits:
- name: train
num_bytes: 7940166
num_examples: 1
download_size: 553870
dataset_size: 7940166
- config_name: metadata
features:
- name: experiment_name
dtype: string
- name: start_time
dtype: string
- name: description
dtype: string
- name: base_org
dtype: string
- name: stage_number
dtype: string
- name: stage_type
dtype: string
- name: status
dtype: string
splits:
- name: train
num_bytes: 9426
num_examples: 5
download_size: 9684
dataset_size: 9426
configs:
- config_name: evals_eval_rl
data_files:
- split: test
path: evals_eval_rl/test-*
- config_name: logs__evaluation_eval_rl
data_files:
- split: train
path: logs__evaluation_eval_rl/train-*
- config_name: metadata
data_files:
- split: train
path: metadata/train-*
---
# Experiment Tracker: FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg

**Experiment Description:** Evaluation experiment for task countdown_4arg from FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL

**Start Time:** 2025-12-01T22:40:29.187000

**Tracker Dataset:** [TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1](https://huggingface.co/datasets/TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1)

## Stages Completed

Total stages: 1

## Models Created

## Dataset Configurations

This tracker dataset contains the following configurations with **immediate upload** as stages complete:

### Training Data (Complete Datasets)

### Hyperparameters (Complete Configurations)

### Logs (Stage-Specific)

### Evaluation Results (Complete with Annotations)

### Metadata
- **experiment_metadata**: Timeline and stage information

## Usage

Load specific configurations with:

```python
from datasets import load_dataset

# Load experiment metadata
metadata = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1', 'experiment_metadata')

# Load complete training datasets
sft_data = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1', 'training_data__sft')
sft_metadata = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1', 'training_data__sft_metadata')

# Load complete configurations
sft_hyperparams = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1', 'hyperparameters__sft')
rl_hyperparams = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1', 'hyperparameters__rl')

# Load stage-specific logs
sft_logs = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1', 'logs__sft')
rl_logs = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1', 'logs__rl')

# Load evaluation results with annotations
sft_eval_results = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1', 'evals_eval_sft')
rl_eval_results = load_dataset('TAUR-dev/D-ExpTracker__FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg__v1', 'evals_eval_rl')
```

## Models

## Registry

All models from this experiment are automatically registered in the [SkillFactory Model Registry](https://huggingface.co/datasets/TAUR-dev/SkillFactory-Registration) with:

- **Complete training configuration** (hyperparameters, datasets, methods)
- **Experiment lineage** (links back to this tracker dataset)
- **Stage-specific metadata** (SFT vs RL training details)
- **Structured input data references** (training datasets and configurations)

Registry entries follow the naming pattern: `Model - FinEval_16k_fulleval_3arg_OLMO_RLONLY-RL-countdown_4arg - {stage_name} - {SFT/RL}`

---
*Generated by SkillFactory Experiment Management System*
*All artifacts uploaded immediately as stages complete with perfect data provenance*