Upload folder using huggingface_hub

README.md

@@ -24,8 +24,7 @@ configs:
 
 # TEBench: Project-Level Test Evolution Benchmark
 
-> **TODO before publishing:** set the final `license:` (see *Licensing*), fill in
-> the BibTeX authors, and add the harness / Docker links.
+📄 **Paper:** https://arxiv.org/abs/2605.06125  |  💻 **Code:** https://github.com/iSEngLab/TEBench
 
 **TEBench** is the first **project-level** benchmark for *test evolution*: given a
 project repository and a commit that changes production code, a system must
@@ -53,20 +52,21 @@ A task may carry multiple types simultaneously.
 
 Each instance is built around three versions of the project:
 
-- **V−1** (`base_commit`) — parent commit, before any change.
+- **V−1** (`v_minus_1_commit`) — parent commit, before any change.
 - **V−0.5** — V−1 **plus all non-test changes** from the commit (production code,
-  build config, resources), with the **original tests untouched**. This is the
+  build config, resources), with the **original tests untouched**. V−0.5 is *not*
+  a commit; it is reconstructed by applying `source_patch` to V−1. This is the
   state given to the system: *the source has moved on, the tests have not.*
-- **V0** (`commit`) — the full commit including the developer's actual test
+- **V0** (`v0_commit`) — the full commit including the developer's actual test
   changes. This is the **ground truth**.
 
 In this dataset:
 
-- `source_patch` applies to **V−1** and reconstructs **V−0.5** (the agent's input).
+- `source_patch` applies to **V−1** and reconstructs **V−0.5** (the system input).
 - `test_patch` is the **gold answer**: the developer's test changes (V−0.5 → V0).
 
 Both patches were generated with `git diff --binary --full-index` and are verified
-to apply cleanly to `base_commit` with `git apply` (see `validate_patches.py`):
+to apply cleanly to `v_minus_1_commit` with `git apply` (see `validate_patches.py`):
 **314/314 source patches and 314/314 test patches apply.**
 
 ## Data Fields
@@ -76,14 +76,11 @@ to apply cleanly to `base_commit` with `git apply` (see `validate_patches.py`):
 | `instance_id` | string | Stable id, `<project>__<short_hash>`. |
 | `project` | string | Project name. |
 | `repo` | string | Upstream GitHub slug (e.g. `apache/commons-lang`). |
-| `base_commit` | string | V−1 commit hash (parent). |
-| `commit` | string | V0 commit hash (ground-truth commit). |
-| `commit_date` | string | V0 commit date. |
-| `commit_message` | string | V0 commit message. |
-| `type` | string | Internal source category (`type 1`/`type 2`). |
-| `difficulty` | string | Curated difficulty label. |
-| `labels` | list[string] | Evolution types: `breaking` / `stale` / `missing`. |
-| `test_change_type` | string | Primary test-change category. |
+| `v_minus_1_commit` | string | **V−1** commit hash (parent state, before the change). |
+| `v0_commit` | string | **V0** commit hash (full ground-truth commit). |
+| `v0_date` | string | V0 commit date. |
+| `v0_message` | string | V0 commit message. |
+| `labels` | list[string] | Evolution types: `breaking` / `stale` / `missing` (≥1 per task). |
 | `changed_test_methods` | list[string] | GT modified/added test methods (`Class.method`). |
 | `new_test_files` | list[string] | Test files where GT adds new methods. |
 | `new_test_methods` | string | JSON map `{TestClass: [methods]}` of GT-added methods. |
@@ -92,7 +89,6 @@ to apply cleanly to `base_commit` with `git apply` (see `validate_patches.py`):
 | `test_changed_lines` | int | Test lines changed. |
 | `source_patch` | string | **V−1 → V−0.5** diff (non-test; given to the system). |
 | `test_patch` | string | **V−0.5 → V0** diff (gold test changes). |
-| `label_reasoning` | string | Rationale for the assigned labels. |
 
 There is a single **`test`** split with 314 instances.
 
@@ -110,7 +106,7 @@ Reconstruct the agent input (V−0.5) for an instance:
 
 ```bash
 git clone https://github.com/<repo>.git && cd <repo>
-git checkout <base_commit>
+git checkout <v_minus_1_commit>
 git apply source.patch       # now at V-0.5: new source, old tests
 # ... system produces test changes ...
 git apply test.patch         # gold V0 test changes, for reference/eval
@@ -144,6 +140,14 @@ TEBench evaluates two dimensions (full harness and Docker image referenced below
 | jsoup | 56 | 48 | 42 | 50 |
 | **Total** | **314** | **172** | **207** | **199** |
 
+**Label distribution** (a task may carry several types, so shares do not sum to 100%):
+
+| Type | Tasks | Share of 314 |
+|------|------:|-------------:|
+| Breaking | 172 | 54.8% |
+| Stale | 207 | 65.9% |
+| Missing | 199 | 63.4% |
+
 - **Multi-label:** 219 tasks (69.7%) carry ≥2 types; 45 (14.3%) carry all three;
   95 (30.3%) are single-type.
 - **Complexity (median):** 4 changed files, 34 source lines, ~32 test lines changed.
@@ -159,14 +163,45 @@ projects**.
 
 ## Companion Artifacts
 
-- **Evaluation harness:** *(GitHub repo — TODO link)* — identification & update
-  scoring, baselines.
-- **Reproducible environment:** a Docker image bundling all 10 projects, Maven
-  3.6.3, JDK 8/11/17, and the evaluated agents. *(registry — TODO link)*
+- **Evaluation harness & code:** https://github.com/iSEngLab/TEBench —
+  identification & update scoring, baselines, dataset build scripts.
+- **Reproducible environment:** a prebuilt Docker image, shipped as
+  `docker/tebench-env.tar` in this repository (see below).
+
+## Docker Environment
+
+The Docker image bundles everything needed to run and evaluate TEBench — all 10
+project repositories, Maven 3.6.3, OpenJDK 8 / 11 / 17 (default JDK 8), and the
+evaluated coding agents (Claude Code, Codex CLI, OpenCode CLI) — so you do not
+have to configure per-project toolchains.
+
+The image archive (`docker/tebench-env.tar`) is stored in this dataset repo as an
+LFS file. HuggingFace is **not** a container registry, so you download the tar and
+load it locally:
+
+```bash
+# 1. Download the image archive from this dataset repo
+hf download iSEngLab/TEBench docker/tebench-env.tar --repo-type=dataset --local-dir .
+
+# 2. Load it into your local Docker
+docker load -i docker/tebench-env.tar
+
+# 3. Start a container
+docker run --rm -it tebench-env
+
+# 4. Inside the container
+cd /root/TestUpdate
+java -version && mvn -version          # JDK 8 + Maven 3.6.3
+claude --version && codex --version && opencode --version
+```
+
+Switch JDK inside the container via `JAVA_HOME` (`JAVA11_HOME`, `JAVA17_HOME` are
+also available). Agent worktrees are large derived artifacts and are **not**
+bundled; regenerate them inside the container with the harness when needed.
 
 ## Licensing
 
-- **Annotations / patches metadata in this dataset:** CC-BY-4.0 *(proposed — confirm)*.
+- **Annotations / patch metadata in this dataset:** CC-BY-4.0.
 - **Embedded source code** inside `source_patch` / `test_patch` remains under the
   **original license of each upstream project** and must be used accordingly:
 
@@ -174,7 +209,7 @@ projects**.
 |------------|---------|
 | commons-* (cli, codec, collections, compress, csv, lang, math), gson | Apache-2.0 |
 | jsoup | MIT |
-| jfreechart | LGPL-2.1 *(verify before redistribution)* |
+| jfreechart | LGPL-2.1 |
 
 Attribution to the original authors is preserved through commit hashes and the
 `repo` field. Redistribution of the Docker image (which contains full source
@@ -183,10 +218,10 @@ trees) must likewise honor these licenses.
 ## Citation
 
 ```bibtex
-@inproceedings{tebench,
-  title     = {TEBench: Benchmarking LLM Agents on Project-Level Test Evolution},
-  author    = {TODO},
-  booktitle = {TODO},
-  year      = {2026}
+@article{tebench,
+  title   = {TEBench: Benchmarking LLM Agents on Project-Level Test Evolution},
+  journal = {arXiv preprint arXiv:2605.06125},
+  year    = {2026},
+  url     = {https://arxiv.org/abs/2605.06125}
 }
 ```

build_dataset.py

@@ -97,34 +97,31 @@ def load_labels(labels_csv: Path) -> dict:
         labels = [LABEL_MAP[t] for t in to_list(r.get("new_labels")) if t in LABEL_MAP]
         out[key] = {
             "labels": labels,
-            "label_reasoning": r.get("reasoning", ""),
             "new_test_methods": r.get("new_test_methods", ""),
             "new_test_files": to_list(r.get("new_test_files")),
         }
     return out
 
 
+# Field names follow the paper's version notation (V-1 / V0). V-0.5 is not a
+# commit: it is V-1 with source_patch applied.
 SCHEMA = pa.schema([
     ("instance_id", pa.string()),
     ("project", pa.string()),
     ("repo", pa.string()),
-    ("base_commit", pa.string()),       # V-1 (parent)
-    ("commit", pa.string()),            # V0  (ground-truth commit)
-    ("commit_date", pa.string()),
-    ("commit_message", pa.string()),
-    ("type", pa.string()),
-    ("difficulty", pa.string()),
-    ("labels", pa.list_(pa.string())),  # breaking / stale / missing
-    ("test_change_type", pa.string()),
+    ("v_minus_1_commit", pa.string()),   # V-1: parent state, before the change
+    ("v0_commit", pa.string()),          # V0: full commit incl. GT test changes
+    ("v0_date", pa.string()),
+    ("v0_message", pa.string()),
+    ("labels", pa.list_(pa.string())),   # breaking / stale / missing
     ("changed_test_methods", pa.list_(pa.string())),
     ("new_test_files", pa.list_(pa.string())),
-    ("new_test_methods", pa.string()),  # JSON blob from source table
+    ("new_test_methods", pa.string()),   # JSON map {TestClass: [methods]}
     ("num_changed_files", pa.int64()),
     ("src_changed_lines", pa.int64()),
     ("test_changed_lines", pa.int64()),
-    ("source_patch", pa.string()),      # V-1 -> V-0.5 (non-test, given to agent)
-    ("test_patch", pa.string()),        # V-0.5 -> V0  (gold test changes)
-    ("label_reasoning", pa.string()),
+    ("source_patch", pa.string()),       # V-1 -> V-0.5 (non-test; the system input)
+    ("test_patch", pa.string()),         # V-0.5 -> V0  (gold test changes)
 ])
 
 
@@ -160,20 +157,17 @@ def main():
         lab = labels.get((project, short))
         if lab is None:
             missing_labels += 1
-            lab = {"labels": [], "label_reasoning": "", "new_test_methods": "", "new_test_files": []}
+            lab = {"labels": [], "new_test_methods": "", "new_test_files": []}
 
         records.append({
             "instance_id": f"{project}__{short}",
             "project": project,
             "repo": REPO_SLUG.get(project, project),
-            "base_commit": base,
-            "commit": full,
-            "commit_date": str(row.Date),
-            "commit_message": msg,
-            "type": str(row.Type),
-            "difficulty": str(row.Difficulty),
+            "v_minus_1_commit": base,
+            "v0_commit": full,
+            "v0_date": str(row.Date),
+            "v0_message": msg,
             "labels": lab["labels"],
-            "test_change_type": str(row.TestChangeType),
             "changed_test_methods": to_list(row.ChangedTestMethods),
             "new_test_files": lab["new_test_files"],
             "new_test_methods": lab["new_test_methods"],
@@ -182,7 +176,6 @@ def main():
             "test_changed_lines": int(row.TestChangedLines),
             "source_patch": src_patch,
             "test_patch": test_patch,
-            "label_reasoning": lab["label_reasoning"],
         })
         if i % 50 == 0:
             print(f"  ...{i}/{len(spine)}")

data/test.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bfd6fbe724bd7a56f6c32fe7f8c1f91603528e6d9dc4d075657f62833a190434
-size 5204033
+oid sha256:64eff60fb61779fac101d7063d0970dd4690bb583c7faa0fb3e673e539ddba85
+size 5149425

validate_patches.py

@@ -1,7 +1,7 @@
 #!/usr/bin/env python3
 """Check that every instance's source_patch + test_patch apply cleanly.
 
-For each task we create an isolated detached worktree at ``base_commit``, apply
+For each task we create an isolated detached worktree at ``v_minus_1_commit``, apply
 ``source_patch`` (recreating V-0.5), then apply ``test_patch`` (the gold V0 test
 changes). Failures are reported per instance. Requires the local repos under
 ``defects4j-projects``; only used for verification, not for dataset consumption.
@@ -48,7 +48,7 @@ def main():
         wt = tempfile.mkdtemp(prefix="teb_wt_")
         try:
             subprocess.run(["git", "-C", str(repo), "worktree", "add", "-q",
-                            "--detach", wt, r["base_commit"]], capture_output=True)
+                            "--detach", wt, r["v_minus_1_commit"]], capture_output=True)
             s_ok, s_err = apply_check(wt, r["source_patch"])
             if s_ok:
                 ok_src += 1