Duplicate from groundhogLLM/ACC-dataset

Co-authored-by: SII-sqs <groundhogLLM@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,63 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.avro filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text
+search_agent_data.jsonl filter=lfs diff=lfs merge=lfs -text
+swe_agent_data.jsonl filter=lfs diff=lfs merge=lfs -text
+sql_agent_data.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,127 @@
+---
+language:
+- en
+license: apache-2.0
+size_categories:
+- 10K<n<100K
+task_categories:
+- question-answering
+- text-generation
+pretty_name: ACC Agent Context Compilation Dataset
+tags:
+- agent
+- long-context
+- question-answering
+- text-generation
+- software-engineering
+- sql
+configs:
+- config_name: search_agent
+  data_files:
+  - split: train
+    path: search_agent_data.jsonl
+- config_name: swe_agent
+  data_files:
+  - split: train
+    path: swe_agent_data.jsonl
+- config_name: sql_agent
+  data_files:
+  - split: train
+    path: sql_agent_data.jsonl
+---
+
+# ACC: Agent Context Compilation Dataset
+
+## Overview
+
+This dataset contains **10,802 compiled long-context QA pairs** derived from multi-turn agent trajectories, introduced in the paper [ACC: Compiling Agent Trajectories for Long-Context Training](https://huggingface.co/papers/2605.21850).
+
+Standard agent SFT masks tool responses and only supervises turn-level tool selection, leaving scattered evidence signals unused. **Agent Context Compilation (ACC)** converts trajectories from Search, Software Engineering (SWE), and SQL agents into long-context QA pairs by assembling the original question with tool responses and environment observations gathered across multiple turns. This makes dependencies between the question and distant evidence explicit, enabling direct supervision of long-context reasoning without additional human annotation.
+
+## Dataset Composition
+
+| Config | Examples | Description |
+|--------|----------|-------------|
+| `search_agent` | 3,369 | Web search trajectories compiled into long-context document QA |
+| `swe_agent` | 4,368 | Software engineering trajectories compiled into codebase context QA |
+| `sql_agent` | 3,065 | SQL agent trajectories compiled into table-context relational QA |
+| **Total** | **10,802** | |
+
+Each config is stored as a separate JSONL file and exposed as an independent Hugging Face dataset configuration.
+
+## Data Fields
+
+### `search_agent`
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `dialogs` | `list` | Two-turn conversation with `role` (`user` / `assistant`) and `content` |
+| `ground_truth_answer` | `string` | Reference answer extracted from the compiled trajectory |
+| `context_length` | `string` | Target long-context scale for the compiled example (e.g. `100k`) |
+
+### `swe_agent`
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `dialogs` | `list` | Two-turn conversation with `role` (`user` / `assistant`) and `content` |
+| `instance_id` | `string` | SWE-bench-style instance identifier |
+| `repo` | `string` | Source repository in `owner/repo` format |
+| `base_commit` | `string` | Git commit hash for the codebase context |
+
+### `sql_agent`
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `id` | `string` | Unique example identifier |
+| `dialogs` | `list` | Two-turn conversation with `role` (`user` / `assistant`) and `content` |
+| `ground_truth_answer` | `string` | Reference answer, typically wrapped in `<answer></answer>` tags |
+| `task_type` | `string` | SQL task category |
+| `metadata` | `object` | Additional fields such as `sql`, `scale`, `complexity`, and `task_type` |
+
+Each example follows a user-assistant dialog format suitable for supervised fine-tuning on long-context reasoning.
+
+## Usage
+
+```python
+from datasets import load_dataset
+
+# Load one configuration
+search_ds = load_dataset("groundhogLLM/acc-agent-context-compilation", "search_agent", split="train")
+swe_ds = load_dataset("groundhogLLM/acc-agent-context-compilation", "swe_agent", split="train")
+sql_ds = load_dataset("groundhogLLM/acc-agent-context-compilation", "sql_agent", split="train")
+
+# Inspect one example
+print(search_ds[0]["dialogs"])
+print(search_ds[0]["ground_truth_answer"])
+```
+
+
+## Dataset Viewer Notes
+
+Some examples contain very long `content` fields because they are compiled long-context QA pairs. If the Hugging Face Dataset Viewer shows a preview error such as `TooBigContentError`, the dataset can still be downloaded and loaded locally with the `datasets` library.
+
+## Citation
+
+If you use this dataset, please cite:
+
+```bibtex
+@misc{su2026acccompilingagenttrajectories,
+      title={ACC: Compiling Agent Trajectories for Long-Context Training}, 
+      author={Qisheng Su and Zhen Fang and Shiting Huang and Yu Zeng and Yiming Zhao and Kou Shi and Ziao Zhang and Lin Chen and Zehui Chen and Lijun Wu and Feng Zhao},
+      year={2026},
+      eprint={2605.21850},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2605.21850}, 
+}
+```
+
+## License
+
+This dataset is released under the [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0).
+
+## Limitations
+
+- The three configs have different schemas and should be loaded separately.
+- Examples are compiled from agent trajectories and may inherit biases or errors from the source environments and tools.
+- Long-context samples may be too large for default online preview in the Hugging Face Dataset Viewer.

search_agent_data.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59c276d230c61dc7d98eb111b3b964b42bf8c8c5401d8a300b231f4fcda96984
+size 1203288509

sql_agent_data.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9acc6fe3e71ead3d58fcbf94daea7a6b9fde42de438f26a13b57d09b272afadd
+size 517936951

swe_agent_data.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:998d652b718f1c0811ecf3126b023dbde025e48d1602380b7c09b7cc1e3d3bd6
+size 970664404