Expose all nine DSEvaluation benchmarks

README.md

@@ -5,23 +5,29 @@ pretty_name: DSEvaluation Data
 
 # DSEvaluation Data
 
-Versioned data mirror and access catalog for
+Versioned data mirror and access catalog for all **9** benchmarks registered in
 [DSEvaluation](https://github.com/yangzhr1/DSEvaluation).
 
-Mirrored assets preserve their upstream licenses:
+| Directory | Benchmark | Tasks | Public content in this repository | Raw data |
+| --- | --- | ---: | --- | --- |
+| `agenticdatabench/` | AgenticDataBench | 246 | Tasks, datasets, gold data, evaluator | Mirrored |
+| `codabench/` | CoDA-Bench | 1,009 | Task metadata, license, archive checksums | Official Kaggle-derived archives |
+| `dabench/` | DABench | 257 | Questions, tables, labels | Mirrored |
+| `dabstep/` | DABstep | 450 | Context, tasks, license | Mirrored; historical submissions excluded |
+| `dacode/` | DA-Code | 500 | Source, gold data, evaluation configs | Mirrored |
+| `longds/` | LongDS | 68 | Pinned source manifest | Official dataset |
+| `mledojo/` | MLE-Dojo | 10 | Selected task catalog and access instructions | Official Kaggle competitions |
+| `moscibench/` | MoSciBench | 88 | Official archives, manifest, license | Mirrored with upstream provenance |
+| `sab/` | ScienceAgentBench | 102 | Verified task metadata and access instructions | Official password-protected archive |
 
-- `agenticdatabench/`: Apache-2.0
-- `dabench/`: Apache-2.0
-- `dabstep/`: CC-BY-4.0 context and task files with upstream attribution
-- `dacode/`: MIT
-- `moscibench/`: official MoSciBench distribution (MIT benchmark license;
-  source datasets retain their upstream provenance)
-- `codabench/`: MIT-licensed task metadata and archive checksum catalog only
+The top-level directory set intentionally matches the 9 benchmark plugins.
+`metadata-mirrored` entries are runnable through DSEvaluation, but their raw
+assets remain on official providers because upstream redistribution terms do
+not permit a public copy here. In particular, ScienceAgentBench asks users not
+to redistribute unzipped data, and MLE-Dojo requires accepting each Kaggle
+competition's terms.
 
-CoDA-Bench community archives, LongDS raw task data, ScienceAgentBench, and
-MLE-Dojo are cataloged but not mirrored. Their upstream terms require users to
-download through the official providers. The DSEvaluation CLI handles both
-mirrored and official-download-only assets:
+Prepare any benchmark through the same CLI:
 
 ```bash
 uv run dseval prepare BENCHMARK

catalog.json

@@ -4,6 +4,7 @@
       "availability": "mirrored",
       "license": "Apache-2.0",
       "path": "agenticdatabench",
+      "tasks": 246,
       "upstream": "https://github.com/AgenticDataBench/AgenticDataBench"
     },
     "codabench": {
@@ -11,12 +12,14 @@
       "license": "MIT task metadata; constituent Kaggle datasets retain independent terms",
       "path": "codabench",
       "raw_data": "official-download-only",
+      "tasks": 1009,
       "upstream": "https://huggingface.co/datasets/RUC-DataLab/CoDA-Bench"
     },
     "dabench": {
       "availability": "mirrored",
       "license": "Apache-2.0",
       "path": "dabench",
+      "tasks": 257,
       "upstream": "https://huggingface.co/datasets/infiagent/DABench"
     },
     "dabstep": {
@@ -24,34 +27,44 @@
       "license": "CC-BY-4.0",
       "path": "dabstep",
       "scope": "context and task files only; historical submissions excluded",
+      "tasks": 450,
       "upstream": "https://huggingface.co/datasets/adyen/DABstep"
     },
     "dacode": {
       "availability": "mirrored",
       "license": "MIT",
       "path": "dacode",
+      "tasks": 500,
       "upstream": "https://github.com/yiyihum/da-code"
     },
     "longds": {
       "availability": "official-download-only",
       "path": "longds/source.json",
       "reason": "The official dataset declares license: other and contains Kaggle-derived data.",
+      "tasks": 68,
       "upstream": "https://huggingface.co/datasets/zjunlp/LongDS"
     },
     "mledojo": {
-      "availability": "official-download-only",
+      "availability": "metadata-mirrored",
+      "path": "mledojo",
+      "raw_data": "official-download-only",
       "reason": "Each Kaggle competition has independent access terms and licenses.",
+      "tasks": 10,
       "upstream": "https://github.com/MLE-Dojo/MLE-Dojo"
     },
     "moscibench": {
       "availability": "mirrored",
       "license": "MIT benchmark distribution; source datasets retain upstream provenance",
       "path": "moscibench",
+      "tasks": 88,
       "upstream": "https://github.com/usail-hkust/MoSciBench"
     },
     "sab": {
-      "availability": "official-download-only",
+      "availability": "metadata-mirrored",
+      "path": "sab",
+      "raw_data": "official-download-only",
       "reason": "Upstream asks users not to redistribute unzipped benchmark data.",
+      "tasks": 102,
       "upstream": "https://github.com/OSU-NLP-Group/ScienceAgentBench"
     }
   },

manifest.json

@@ -1,11 +1,11 @@
 {
-  "bytes": 39942052023,
-  "file_count": 1270,
+  "bytes": 39942183334,
+  "file_count": 1274,
   "files": [
     {
-      "bytes": 1039,
+      "bytes": 1846,
       "path": "README.md",
-      "sha256": "f9e4ac6ccef8b582d9c9679bab492a70aec023ef22df1a06de5b9e89f672be0a"
+      "sha256": "85cee23f79929fcb619d504476da9f3f04395d55c7487098673b18ffa31062e2"
     },
     {
       "bytes": 0,
@@ -3943,9 +3943,9 @@
       "sha256": "1e1a6ba11322ff239927cf46c0acabcd8a81bc83371167deb1192c4d9ae844aa"
     },
     {
-      "bytes": 2109,
+      "bytes": 2411,
       "path": "catalog.json",
-      "sha256": "357a9d915a28df592b398ac77b416a437021a66dfbf596cd50160a91bcffad51"
+      "sha256": "4336b64c3ddd2d2ebe2153557bb5ac6b21586f448fa232d4db2c48e9c3919fa5"
     },
     {
       "bytes": 1068,
@@ -3953,9 +3953,9 @@
       "sha256": "623dc8646f6d15298e4f52b1192a0bb450b47e8a8b279dfb5f200f56276e766b"
     },
     {
-      "bytes": 3653,
+      "bytes": 3656,
       "path": "codabench/archive-manifest.json",
-      "sha256": "18091288aca7197e38cd562d02297166f8e98613c2b355bdb10e1e8610d394dc"
+      "sha256": "b8dd0f08f1067af2f7c19b5a53f0ff4343cea63e645ca7c2a2bd25f6a82dc6c1"
     },
     {
       "bytes": 3234796,
@@ -6312,6 +6312,16 @@
       "path": "longds/source.json",
       "sha256": "6b68e9759d892ae88c6736dc2124b121c150adcde0a94ec7fae830d7289c5796"
     },
+    {
+      "bytes": 336,
+      "path": "mledojo/source.json",
+      "sha256": "3f15f334657c4b97d2af2c15bcfa866f7ec02a35f0f4c73719877b0838a3faf7"
+    },
+    {
+      "bytes": 413,
+      "path": "mledojo/tasks.jsonl",
+      "sha256": "7ca785d44e3852b85518f2d56c8d0c8f2054601ab3b898d189e38bad2d4ff4f8"
+    },
     {
       "bytes": 1073,
       "path": "moscibench/LICENSE",
@@ -6351,6 +6361,16 @@
       "bytes": 868,
       "path": "moscibench/manifest.json",
       "sha256": "49d50d3e38780b49684ca0c57fc0206a21ed7778b65d2be55565e44ecb74c854"
+    },
+    {
+      "bytes": 364,
+      "path": "sab/source.json",
+      "sha256": "32fd83199aaa4cc8daf3eb3bfc4e6ee248a18b73d6197faee191de88b992bdd7"
+    },
+    {
+      "bytes": 129086,
+      "path": "sab/verified-00000-of-00001.parquet",
+      "sha256": "c6f937863a220bd1762a00c20a0f79cc8dfca900b819bdb552150310731ae147"
     }
   ],
   "version": "v1"

mledojo/source.json

@@ -0,0 +1,9 @@
+{
+  "benchmark": "MLE-Dojo",
+  "prepare": "KAGGLE_API_TOKEN=... uv run dseval prepare mledojo",
+  "raw_data": "official-download-only",
+  "reason": "Each selected Kaggle competition has independent access and redistribution terms.",
+  "task_catalog": "tasks.jsonl",
+  "tasks": 10,
+  "upstream": "https://github.com/MLE-Dojo/MLE-Dojo"
+}

mledojo/tasks.jsonl

@@ -0,0 +1,10 @@
+{"task_id": "dont-call-me-turkey"}
+{"task_id": "GiveMeSomeCredit"}
+{"task_id": "forest-cover-type-kernels-only"}
+{"task_id": "novozymes-enzyme-stability-prediction"}
+{"task_id": "integer-sequence-learning"}
+{"task_id": "random-acts-of-pizza"}
+{"task_id": "afsis-soil-properties"}
+{"task_id": "sberbank-russian-housing-market"}
+{"task_id": "aerial-cactus-identification"}
+{"task_id": "crowdflower-weather-twitter"}

sab/source.json

@@ -0,0 +1,9 @@
+{
+  "benchmark": "ScienceAgentBench",
+  "metadata": "verified task metadata is mirrored when present",
+  "prepare": "uv run dseval prepare sab",
+  "raw_data": "official-download-only",
+  "reason": "Upstream explicitly asks users not to redistribute unzipped data files online.",
+  "tasks": 102,
+  "upstream": "https://github.com/OSU-NLP-Group/ScienceAgentBench"
+}

sab/verified-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6f937863a220bd1762a00c20a0f79cc8dfca900b819bdb552150310731ae147
+size 129086