Hugging Face's logo

Datasets:
mayflowergmbh
/
boldt-dc-1b-orpo-onpolicy-de

Tasks:
Text Generation
Modalities:
Tabular
Text
Formats:
parquet
Languages:
German
Size:
10K - 100K
Tags:
german
deutsch
preference-data
orpo
dpo
on-policy
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
boldt-dc-1b-orpo-onpolicy-de / README.md
johannhartmann's picture
johannhartmann
Upload dataset
bbfb5ef verified
preview code
|
Raw
History Blame Contribute Delete
7.78 kB
metadata 
license: apache-2.0
language:
  - de
task_categories:
  - text-generation
size_categories:
  - 10K<n<100K
tags:
  - german
  - deutsch
  - preference-data
  - orpo
  - dpo
  - on-policy
  - rejected-sampling
  - llm-alignment
pretty_name: Boldt-DC-1B On-Policy ORPO (German)
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
dataset_info:
  features:
    - name: row_id
      dtype: string
    - name: source
      dtype: string
    - name: prompt
      dtype: string
    - name: chosen
      dtype: string
    - name: rejected
      dtype: string
    - name: original_rejected
      dtype: string
    - name: prompt_tok_len
      dtype: int64
    - name: gen_model
      dtype: string
    - name: gen_engine
      dtype: string
    - name: gen_seed
      dtype: int64
    - name: gen_temperature
      dtype: float64
    - name: timestamp_utc
      dtype: string
    - name: is_refusal
      dtype: bool
    - name: chosen_tok_len
      dtype: int64
    - name: rejected_tok_len
      dtype: int64
  splits:
    - name: train
      num_bytes: 276050043
      num_examples: 54421
  download_size: 139892644
  dataset_size: 276050043

Boldt-DC-1B On-Policy ORPO (German)

A 56,413-row German preference dataset in which the rejected field is the actual output of the mayflowergmbh/boldt-dc-1b-german-it-16k SFT model under greedy decoding, rather than an off-the-shelf "alternative" answer.

The intent is on-policy ORPO/DPO: training against a model's own current failure modes typically gives a much sharper gradient signal than training against a generic weak baseline, because every preference pair documents a concrete gap between what this specific model does today (rejected) and what we want it to do (chosen).

Quick start 

from datasets import load_dataset
ds = load_dataset("mayflowergmbh/boldt-dc-1b-orpo-onpolicy-de", split="train")
print(ds[0]["prompt"][-300:])
print("CHOSEN:",   ds[0]["chosen"][:200])
print("REJECTED:", ds[0]["rejected"][:200])

Drop straight into TRL's ORPOTrainer / DPOTrainer (the prompt field is already formatted with the Boldt chat tokens <|system|> / <|user|> / <|assistant|> and primed with a trailing <|assistant|>\n, so no chat-template application is required).

Schema

Column Type Description
prompt string Full conversation history + system prompt, formatted with Boldt's chat tokens, ending with `<
chosen string The preferred assistant response (from the source dataset)
rejected string Greedy continuation of prompt produced by mayflowergmbh/boldt-dc-1b-german-it-16k
original_rejected string The off-the-shelf rejected text from the source dataset, kept for provenance/diffing
source string "orpo-dpo-mix-40k-de" or "intel-orca-dpo-de"
row_id string 24-char SHA-256 of source + "::" + prompt, for stable dedup and resumable builds
prompt_tok_len int Token count of the prompt (Boldt tokenizer)
chosen_tok_len int Whitespace word count of chosen
rejected_tok_len int Whitespace word count of rejected
is_refusal bool Flagged by a small German+English refusal pattern. Refusals are not dropped — when chosen is a real answer, a refusal in rejected is valid contrastive signal
gen_model string Generation model identifier
gen_engine string "hf" (transformers + manual batched generation)
gen_seed int RNG seed
gen_temperature float 0.0 (greedy)
timestamp_utc string When the row was generated

How it was built

  1. Source unification. Two German preference datasets were merged and unified into a (prompt, chosen, original_rejected) shape:

    Source Rows ingested Rows kept after unify + cross-source dedup
    johannhartmann/orpo-dpo-mix-40k-llama3-de 44,240 43,947
    mayflowergmbh/intel_orca_dpo_pairs_de 12,855 12,466

    The prompt string is built with the exact same helper (build_orpo_prompt in boldt_it.dataset_formatting) that ORPO training uses, so the model sees byte-identical conditioning at training time and at the moment its rejected was generated.

  2. Generation. mayflowergmbh/boldt-dc-1b-german-it-16k was loaded in bf16 on a single RTX A6000. For each prompt, the model emitted a greedy continuation with do_sample=False, max_new_tokens=384, max_seq_length=4096. Generation was batched with padding_side="left" and length-bucketed within shuffled 256-row chunks; throughput was ~2.78 prompts/s end-to-end (≈5.6 h total).

  3. Filtering. Post-generation drops:

    Reason Count
    empty rejected 12
    too short (< 20 chars) 29
    equal to chosen 0
    near-duplicate (5-gram Jaccard ≥ 0.92) 0
    prompt would have been truncated (> 3,584 tokens) 39
    kept 56,413

    No equal_to_chosen and no jaccard_dup survived → there is real gradient signal in every published row.

Length distributions

chosen (words) rejected (words)
p10 30 108
median 216 237
mean 235 218
p90 457 294
p99 674 329
max 1,561 384

rejected capped at 384 tokens by the generation budget; chosen tail extends because the source datasets contain long-form answers. For ORPO this is fine — the contrast happens primarily in the first few hundred tokens.

Refusal rate on rejected: 1.2 % (flagged in is_refusal).

Intended use

  • ORPO/DPO/IPO/SimPO post-training of mayflowergmbh/boldt-dc-1b-german-it-16k and close descendants. The on-policy property is specific to this base model; using this dataset to train a different model gives you off-policy data again.
  • Diagnostic analysis of failure modes in the SFT-only Boldt model: every rejected row is a concrete example of where the model currently underperforms.

Limitations and considerations

  • Quality of rejected ≠ minimum-quality reference. Some rejected outputs are actually reasonable answers that just differ stylistically from chosen. ORPO will still learn from these, but the gradient strength varies per row.
  • chosen quality inherits from source datasets. orpo-dpo-mix-40k is a community-curated mix; intel_orca_dpo_pairs_de is a German translation of Intel's Orca DPO pairs. Neither has been hand-verified for factual accuracy.
  • Repetition / hallucination patterns in rejected are characteristic of 1.25 B-parameter Llama-family SFT-only behavior. Don't assume these patterns will transfer to a different scale.
  • Licensing. This dataset is released under Apache 2.0, but the upstream license of johannhartmann/orpo-dpo-mix-40k-llama3-de and mayflowergmbh/intel_orca_dpo_pairs_de (which seed the prompts and chosen answers) should be reviewed before commercial use.

Provenance

  • Generation script: scripts/16_build_orpo_onpolicy_de.py in the Boldt training repo
  • Build report: reports/orpo_on_policy_build.json
  • Spot-check sample: reports/orpo_on_policy_spotcheck.md (20 rows of prompt / chosen / rejected / original_rejected)
  • Engine: transformers 5.5.0 + torch 2.10.0+cu128, bf16, NVIDIA RTX A6000

Citation 

@misc{boldt-onpolicy-orpo-de-2026,
  title  = {Boldt-DC-1B On-Policy ORPO (German)},
  author = {Mayflower GmbH},
  year   = {2026},
  url    = {https://huggingface.co/datasets/mayflowergmbh/boldt-dc-1b-orpo-onpolicy-de}
}