| --- |
| pretty_name: DeliverBench |
| license: cc-by-4.0 |
| language: |
| - zh |
| - en |
| tags: |
| - agents |
| - benchmark |
| - multi-step |
| - document-generation |
| - finance |
| - medicine |
| - law |
| size_categories: |
| - n<1K |
| configs: |
| - config_name: projects |
| data_files: |
| - split: test |
| path: index/projects.jsonl |
| --- |
| <p align="center"> |
| <a href="https://deliverbench.github.io/"><img src="assets/logo-folio.svg" width="76" alt="DeliverBench project mark"></a> |
| </p> |
| <h1 align="center">DeliverBench</h1> |
| <p align="center"><strong>Benchmarking AI agents on continuous professional project delivery</strong></p> |
| <p align="center"><a href="https://arxiv.org/abs/2611.03412"><strong>📄 Paper</strong></a> · <a href="https://deliverbench.github.io/"><strong>🌐 Project website</strong></a> · <a href="https://github.com/theseus-labs-rsi/deliverbench"><strong>💻 Code</strong></a> · <a href="#download-and-run"><strong>📦 Download & run</strong></a></p> |
|
|
| Professional work unfolds through shared files, new inputs, feedback, and revision. |
| DeliverBench evaluates that continuous project lifecycle, from intermediate submissions |
| to a complete and usable final package. |
|
|
| [](assets/figure-01-motivation.png) |
|
|
| <sub>The evaluation unit expands from individual tasks to a complete professional project.</sub> |
|
|
| ## Three public cases |
|
|
| | Case | Professional workflow | Stages | Explore | |
| |:---|:---|:---:|:---| |
| | **Case 01** | Project Mars Refinancing | 13 | [📁 Files](https://huggingface.co/datasets/deliverbench/DeliverBench/tree/main/projects/case-01_mars_refinancing) · [📦 ZIP](https://huggingface.co/datasets/deliverbench/DeliverBench/resolve/main/archives/DeliverBench-Case01.zip?download=true) | |
| | **Case 02** | Pharmacovigilance ICSR | 12 | [📁 Files](https://huggingface.co/datasets/deliverbench/DeliverBench/tree/main/projects/case-02_pv_icsr) · [📦 ZIP](https://huggingface.co/datasets/deliverbench/DeliverBench/resolve/main/archives/DeliverBench-Case02.zip?download=true) | |
| | **Case 03** | StarPulse Legal Delivery | 8 | [📁 Files](https://huggingface.co/datasets/deliverbench/DeliverBench/tree/main/projects/case-03_starpulse_legal) · [📦 ZIP](https://huggingface.co/datasets/deliverbench/DeliverBench/resolve/main/archives/DeliverBench-Case03.zip?download=true) | |
|
|
| Each case also has one independent FINAL evaluation. This public collection contains |
| three complete projects; it is not the entire research collection. |
|
|
| ## Download and run |
|
|
| Clone the code repository, install its dependencies, and run: |
|
|
| ```sh |
| python scripts/download_data.py --project all --output-dir data |
| ``` |
|
|
| The code pins this dataset to a full commit SHA and checks every downloaded project |
| file. Each project is also browsable under `projects/`; `archives/` provides a complete |
| ZIP for each project. SHA256 lists cover the expanded files and archives. |
|
|
| Install shared extension 0.7.2 with stock Harbor 0.20.0 in Python 3.12. Docker and |
| Compose are required. Each project README explains execution and an optional local |
| synthetic pipeline check that does not call paid model APIs. Real Candidate, Judge, |
| and Responder calls require the user's own provider configuration and may incur cost. |
|
|
| ### Install the bundled Harbor extension |
|
|
| If you downloaded this Hugging Face repository, it also includes |
| [`deliverbench-harbor/`](deliverbench-harbor), the same shared extension distributed |
| on GitHub. From the downloaded repository root, with `uv` available: |
|
|
| ```sh |
| uv venv --python 3.12 .venv-deliverbench |
| uv pip install --python .venv-deliverbench/bin/python uv ./deliverbench-harbor |
| source .venv-deliverbench/bin/activate |
| harbor --version |
| harbor plugins list |
| ``` |
|
|
| Then follow the configuration and run instructions in the chosen project's README |
| under `projects/`. The individual case ZIPs contain task files; the shared extension |
| is provided separately in this repository. Python dependencies and container images |
| still require downloads. |
|
|
| GitHub is the canonical development repository. This copy is a source snapshot of |
| extension **0.7.2**, compatible with Harbor **0.20.0**, from |
| [commit `76dcc68`](https://github.com/theseus-labs-rsi/deliverbench/tree/76dcc68c464e7dace943e3871509b2d672dce896/deliverbench-harbor). |
| [`code_snapshot.json`](code_snapshot.json) records its origin and file hashes; |
| [`SHA256SUMS`](SHA256SUMS) covers the published files. Future snapshots should be |
| updated from GitHub together with their compatibility metadata. |
|
|
| ## How delivery works |
|
|
| [](assets/figure-03-execution.png) |
|
|
| <sub>The same agent carries its work across stages. Checkpoints and the final delivery package are evaluated separately.</sub> |
|
|
| ## A project in practice |
|
|
| [](assets/figure-08-dependencies.png) |
|
|
| <sub>Case 01: earlier analyses, models, and lender feedback feed into a six-file final handoff. Original paper artwork.</sub> |
|
|
| ## Dataset structure |
|
|
| The viewer contains one index row per project, with its directory, domain, stage count, |
| version, and criterion counts. The `test` split denotes public evaluation tasks; it is |
| not a training split or a set of independent question-answer records. Full documents, |
| staged inputs, rubrics, response rules, authored references, and environment files are |
| in the project directories. Office and PDF attachments retain their original formats. |
|
|
| An agent continues its own session across stages. Later inputs arrive at defined |
| times; response rules generate feedback from its frozen deliverables. The background |
| opening is unscored. The 33 process stages and three FINAL units are scored separately; |
| FINAL does not average process scores or inherit earlier hurdle hits. Incomplete |
| judgments and runtime failures do not become valid zero scores. |
|
|
| ## Materials and intended use |
|
|
| The tasks reconstruct professional workflows with source documents, reconstructed |
| scenario materials, staged client feedback, rubrics, and authored reference fixtures. |
| Original notices and business-relevant document comments and revisions are retained. |
| The publishers have confirmed permission to redistribute this release's materials. |
|
|
| References, rubrics, simulator internals, and future-stage materials are public for |
| research reproducibility but remain isolated from the Candidate runtime. Evaluated |
| agents must not retrieve these hidden materials from this public repository. These |
| are public evaluation tasks; exposure or training contamination may affect comparisons. |
| The three projects are not representative of every profession or language. Their |
| source material mixes Chinese and English, with output-language requirements defined |
| by the individual tasks. Agent and Judge performance can depend on provider and tool |
| configuration. No universal assessment of professional competence is implied. |
|
|
| ## Compatibility and licenses |
|
|
| Shared extension: 0.7.2. Harbor: 0.20.0. |
| The Case 01 language protocol treats a prohibited Chinese body as a violation; English |
| body text does not trigger that hurdle. Structured language observations are checked |
| for consistency. All task thresholds, weights, dependencies, and FINAL definitions |
| are fixed in this release. |
|
|
| Dataset content and [presentation figures](assets/README.md) are [CC BY 4.0](LICENSE). Executable code and environment/configuration |
| files are [Apache-2.0](LICENSE-CODE); see [LICENSING.md](LICENSING.md) for the boundary. |
| Use [CITATION.bib](CITATION.bib) when referencing this release. |
|
|
| This repository contains no model submissions, scores, judgments, evaluation reports, |
| execution logs, or trajectories. The reference fixtures are authored task components. |
|
|
