Improve dataset card with paper, composition, loading examples, and citation

README.md

@@ -1,6 +1,6 @@
 ---
 license: other
-pretty_name: Omni-Edu (Core V6)
+pretty_name: 'OmniEdu: Instruction Data for Learning and Teaching'
 language:
 - en
 - zh
@@ -15,6 +15,10 @@ tags:
 - sft
 - multimodal
 - sharegpt
+- curriculum-grounding
+- diagnostic-reasoning
+- pedagogical-scaffolding
+- arxiv:2609.23088
 size_categories:
 - 10K<n<100K
 configs:
@@ -24,81 +28,242 @@ configs:
     path: train.jsonl
 ---
 
-# Omni-Edu — Core V6 SFT mixture
+# OmniEdu Dataset
 
-69,999 supervised instruction examples (~158M characters) covering K-12 subject
-competence, curriculum grounding, diagnostic reasoning, pedagogical action and
-general-purpose instruction. 12,146 rows (17.4%) are multimodal; every image
-referenced by the JSONL ships in this repository under `images/`.
+**Instruction data for K–12 learning and teaching**
 
-This is the *system-prompted* assembly of the v6 core mixture: every row carries
-an explicit `system` message, and the non-system turns are byte-identical to the
-unprompted v6 release. `assembly_report.json` records the audit that verified
-that property.
+[Paper](https://arxiv.org/abs/2609.23088) · [Project Page](https://haolpku.github.io/Omni-Edu/) · [GitHub](https://github.com/haolpku/Omni-Edu) · [Models](#model-family) · [Quick Start](#quick-start) · [Citation](#citation)
 
-## Composition
+OmniEdu is a capability-oriented instruction-tuning dataset for educational models that connect **subject knowledge, curriculum understanding, learner diagnosis, and instructional support**. This repository releases **69,999 training examples**, combining **60,951 education-specific examples** with **9,048 general-purpose examples**.
 
-| Capability family | Examples |
-| --- | ---: |
-| Subject competence | 31,855 |
-| Pedagogical action and scaffolding | 14,226 |
-| Curriculum grounding | 9,381 |
-| General-purpose instruction | 9,048 |
-| Diagnostic reasoning | 5,489 |
-| **Total** | **69,999** |
+The dataset accompanies [**OmniEdu: Open Foundation Models for Learning and Teaching**](https://arxiv.org/abs/2609.23088). The paper describes a mixture drawn from more than 100 educational resources and general instruction sources, containing **15.96M supervised response tokens**.
+
+Hao Liang · Qihan Lin · Meiyi Qiang · Linzhuang Sun · Hengyi Feng · Mingrui Chen · Sizhe Qiu · Wentao Zhang  
+Peking University · University of the Chinese Academy of Sciences · Zhongguancun Academy
+
+## Dataset at a glance
+
+| Property | Released data |
+| --- | --- |
+| Examples | 69,999 |
+| Education-specific / general-purpose | 60,951 / 9,048 |
+| Multimodal examples | 12,146 (17.4%) |
+| Languages | English and Chinese |
+| Modalities | Text; image and text |
+| Format | JSON Lines with ShareGPT-style messages |
+| Configuration | `core_v6_full_system_prompted` |
+| Split | `train` |
+| System prompt IDs in use | 20, including `general_native` |
+| Image assets | Included under `images/` |
+
+This is the **system-prompted Core V6 release**. Every example starts with a system message. The [assembly report](assembly_report.json) records that non-system message content and image references are unchanged from the unprompted Core V6 assembly. This repository provides a training mixture; it does not define validation or test splits.
+
+## Capability composition
+
+| Capability family | Examples | Share | Training focus |
+| --- | ---: | ---: | --- |
+| Subject competence | 31,855 | 45.51% | Solve school-level problems using textual and visual evidence. |
+| Pedagogical action and scaffolding | 14,226 | 20.32% | Provide hints, guiding questions, feedback, and adaptive instructional support. |
+| Curriculum grounding | 9,381 | 13.40% | Connect problems and concepts to knowledge points, curriculum standards, and prerequisite relations. |
+| Diagnostic reasoning | 5,489 | 7.84% | Identify learner errors, misconceptions, and gaps in understanding. |
+| General-purpose instruction | 9,048 | 12.93% | Maintain broad instruction-following behavior alongside educational specialization. |
+| **Total** | **69,999** | **100%** | |
+
+The four education-specific families account for 60,951 examples. General-purpose instruction supplies the remaining 9,048 examples. Percentages are rounded.
 
-The system prompt is never free text: each row points at a named entry in
-`system_prompt_registry.json` via `extra_info.system_prompt_id`. 20 prompt
-variants are in use (e.g. `solve_reasoned`, `multiturn_socratic`,
-`mathtutor_scaffolding`, `general_native`). The 9,048 general-purpose rows keep
-their upstream system prompt, which is why the `general_native` id is not in the
-registry.
+## Quick start
 
-## Format
+### Load the training split
 
-JSON Lines, ShareGPT style, exactly three top-level keys per row:
+```bash
+pip install -U datasets huggingface_hub pillow
+```
+
+```python
+import json
+from datasets import load_dataset
+
+DATASET_ID = "lhpku20010120/Omni-Edu"
+CONFIG = "core_v6_full_system_prompted"
+
+train = load_dataset(DATASET_ID, CONFIG, split="train")
+example = train[0]
+metadata = json.loads(example["extra_info"])
+
+print(len(train))  # 69999
+print(example["messages"])
+print(example["images"])
+print(metadata["category"], metadata["system_prompt_id"])
+```
+
+For sequential inspection without first downloading the entire JSONL file, pass `streaming=True` to `load_dataset` and retrieve an example with `next(iter(train))` instead of indexing.
+
+### Load an image from an example
+
+The `images` field contains **relative file paths**, not decoded image objects. Loading the training split does not automatically download and open these image assets. Continue from the example above:
+
+```python
+from huggingface_hub import hf_hub_download
+from PIL import Image
+
+loaded_images = []
+for relative_path in example["images"]:
+    local_path = hf_hub_download(
+        repo_id=DATASET_ID,
+        repo_type="dataset",
+        filename=relative_path,
+    )
+    with Image.open(local_path) as image:
+        loaded_images.append(image.convert("RGB"))
+
+print([image.size for image in loaded_images])
+```
+
+Preserve the order of the image paths and any repeated entries: they correspond to the `<image>` placeholders in the conversation. For text-only examples, `images` is an empty list.
+
+### Download the full release for local training
+
+This downloads the JSONL file, image assets, and accompanying metadata:
+
+```python
+from pathlib import Path
+from datasets import load_dataset
+from huggingface_hub import snapshot_download
+
+release_dir = Path(snapshot_download(
+    repo_id="lhpku20010120/Omni-Edu",
+    repo_type="dataset",
+    local_dir="./Omni-Edu",
+))
+train = load_dataset(
+    "json",
+    data_files={"train": str(release_dir / "train.jsonl")},
+    split="train",
+)
+image_paths = [release_dir / path for path in train[0]["images"]]
+print(image_paths)
+```
+
+For reproducible experiments, set `revision` to a specific dataset commit when loading or downloading. LLaMA-Factory users can use the supplied [dataset registration](dataset_info.json), which maps the message roles, content, and image fields.
+
+## Data format
+
+Each line of `train.jsonl` is a JSON object with exactly three top-level fields:
+
+| Field | Type | Meaning |
+| --- | --- | --- |
+| `messages` | List of objects | A system message followed by user and assistant turns; each object has `role` and `content`. |
+| `images` | List of strings | Image paths relative to the repository root; an empty list for text-only examples. |
+| `extra_info` | String containing JSON | Provenance, capability category, prompt ID, and optional selection or source metadata. Parse with `json.loads`. |
+
+The following is a schematic example; the text and image filename are placeholders, and the metadata is abbreviated:
 
 ```json
 {
   "messages": [
-    {"role": "system",    "content": "..."},
-    {"role": "user",      "content": "Problem:\n<image> ..."},
-    {"role": "assistant", "content": "..."}
+    {
+      "role": "system",
+      "content": "Task-specific instruction..."
+    },
+    {
+      "role": "user",
+      "content": "Problem:\n<image>Question text..."
+    },
+    {
+      "role": "assistant",
+      "content": "Target answer..."
+    }
   ],
-  "images": ["images/0a/0af6e8ac....png"],
-  "extra_info": "{\"category\": \"subject_competence\", \"dataset\": \"geometry3k-answer-only\", ...}"
+  "images": [
+    "images/0a/example.png"
+  ],
+  "extra_info": "{\"category\": \"subject_competence\", \"dataset\": \"geometry3k-answer-only\", \"uid\": \"geometry3k-513\", \"system_prompt_id\": \"solve_answer_only\"}"
 }
 ```
 
-* `messages` — `system`, then alternating `user` / `assistant`. Multi-turn rows
-  go up to 21 messages; 94.5% are single-turn.
-* `images` — 0 to 16 entries, relative to the repository root; 82.6% of rows are
-  text-only, 16.2% carry one image, and the rest carry between 2 and 16. The
-  `<image>` marker inside the user turn shows where each image belongs, and 131
-  rows reference the same image more than once.
-* `extra_info` — a **JSON-encoded string** (not an object); `json.loads` it to
-  reach `category`, `dataset`, `uid`, `system_prompt_id`, `system_prompt_version`
-  and, where present, `kcenter` / `source_license` / `language`.
+Conversations contain up to 21 messages; approximately 94.5% are single-turn. Examples contain zero to 16 image references: approximately 82.6% are text-only, 16.2% contain one image, and the remainder contain multiple images. In 131 rows, an image is referenced more than once.
 
-`dataset_info.json` is the llama-factory registration file for this mixture.
+After parsing `extra_info`, common keys include `category`, `dataset`, `uid`, `system_prompt_id`, and `system_prompt_version`. Optional fields include `kcenter`, `source_license`, and `language`; availability varies by source.
 
-## Loading
+## System prompts
 
-```python
-from datasets import load_dataset
+Task-specific instructions make the expected teaching or problem-solving behavior explicit. Examples of active prompt IDs include:
 
-ds = load_dataset("lhpku20010120/Omni-Edu", "core_v6_full_system_prompted", split="train")
-print(ds[0]["messages"], ds[0]["images"])
-```
+| Prompt ID | Intended behavior |
+| --- | --- |
+| `solve_answer_only` | Return an answer in the requested format without an explanation. |
+| `solve_reasoned` | Produce a self-contained solution with a clear final answer. |
+| `knowledge_point` | Identify concepts, standards, prerequisites, or curriculum relations. |
+| `diagnose_and_correct` | Identify the learner's first material error and provide a useful correction. |
+| `mathtutor_scaffolding` | Give a concise next scaffold. |
+| `multiturn_socratic` | Use dialogue history to guide the learner with focused questions. |
+| `general_native` | Retain the upstream system prompt for general-purpose examples. |
+
+There are **20 prompt IDs in use** in the [assembly report](assembly_report.json). Education-specific instructions are defined in [system_prompt_registry.json](system_prompt_registry.json). The **9,048 `general_native` examples retain their upstream prompts**, so `general_native` is not an entry in that registry. The registry also contains `general_assistant`, which is not used in this release's prompt counts.
+
+## Construction and provenance
+
+The [paper](https://arxiv.org/abs/2609.23088) describes a pipeline combining deterministic cleaning, semantic auditing and rewriting, task-specific quality scoring, token-budgeted diversity selection, and pedagogical instruction assignment. Per-example metadata identifies the contributing source and, where provided upstream, its license.
+
+This release was assembled as `training_core_v6/releases/v6_full_system_prompted`. Its [assembly report](assembly_report.json) records:
+
+| Assembly check | Result |
+| --- | --- |
+| Every row has a system message | Yes |
+| Non-system content unchanged from the unprompted assembly | Yes |
+| Image references unchanged | Yes |
+| Tokenizer used for length audit | `Qwen/Qwen3.5-9B-Base` |
+| Maximum audited sequence length | 12,786 tokens |
+| 99th-percentile sequence length | 8,830 tokens |
+| Audit cutoff | 16,384 tokens |
+| Examples above the cutoff | 0 |
 
-Images resolve against the repository root, so either download the full snapshot
-or prefix `images/...` with the repository URL when serving rows.
+These are dataset assembly statistics under the named tokenizer. The paper separately reports a **32,768-token maximum training sequence length** for model fine-tuning; it is a training configuration rather than the longest sequence in this release. The **15.96M supervised response tokens** reported in the paper measure response-token volume, not total input-plus-output sequence length.
+
+## Repository contents
+
+| Path | Contents |
+| --- | --- |
+| [train.jsonl](https://huggingface.co/datasets/lhpku20010120/Omni-Edu/blob/main/train.jsonl) | Training examples |
+| [images/](https://huggingface.co/datasets/lhpku20010120/Omni-Edu/tree/main/images) | Referenced image assets |
+| [system_prompt_registry.json](system_prompt_registry.json) | Task-specific instruction definitions |
+| [assembly_report.json](assembly_report.json) | Prompt counts and assembly audit |
+| [dataset_info.json](dataset_info.json) | LLaMA-Factory dataset registration |
+
+## Model family
+
+Explore the accompanying models for inference examples and evaluation results:
+
+| Model | Scale |
+| --- | ---: |
+| [OmniEdu-4B](https://huggingface.co/lhpku20010120/Omni-Edu-4B) | 4B |
+| [OmniEdu-9B](https://huggingface.co/lhpku20010120/Omni-Edu-9B) | 9B |
+| [OmniEdu-27B](https://huggingface.co/lhpku20010120/Omni-Edu-27B) | 27B |
+
+## Intended use and limitations
+
+The mixture supports supervised fine-tuning, analysis of educational instruction data, and research on curriculum grounding, learner diagnosis, and tutoring. Coverage reflects the contributing sources and does not establish uniform quality across subjects, languages, curricula, or learner populations. Inspect examples for suitability before using them in educational applications. Model benchmark results reported in the paper do not establish classroom learning gains.
+
+## Source licenses and usage terms
+
+This repository retains the `other` license designation. Licenses and usage terms of the constituent datasets and source materials continue to apply. Check `extra_info.dataset` to identify the source and `extra_info.source_license` where available; consult the upstream terms before reuse or redistribution. Missing per-row license metadata does not imply unrestricted permission.
+
+## Citation
+
+If you use the dataset or accompanying models in your research, please cite:
+
+```bibtex
+@misc{liang2026omniedu,
+  title         = {OmniEdu: Open Foundation Models for Learning and Teaching},
+  author        = {Hao Liang and Qihan Lin and Meiyi Qiang and Linzhuang Sun and Hengyi Feng and Mingrui Chen and Sizhe Qiu and Wentao Zhang},
+  year          = {2026},
+  eprint        = {2609.23088},
+  archivePrefix = {arXiv},
+  primaryClass  = {cs.CL},
+  url           = {https://arxiv.org/abs/2609.23088}
+}
+```
 
-## Provenance and scope
+## Feedback
 
-Built from the `training_core_v6/releases/v6_full_system_prompted` assembly,
-tokenised against `Qwen/Qwen3.5-9B-Base`: max sequence length 12,786 tokens,
-p99 8,830, and zero rows above the 16,384 cutoff. Sources are public K-12 and
-general corpora; per-row `extra_info.source_license` is present where the
-upstream source declares one. Licences of the constituent corpora still apply —
-check `extra_info.dataset` before redistribution.
+For questions about the data, source attribution, or educational use cases, open a discussion in the [dataset Community tab](https://huggingface.co/datasets/lhpku20010120/Omni-Edu/discussions) or an [issue on GitHub](https://github.com/haolpku/Omni-Edu/issues).