Organize WildClawBench and TerminalBench as parallel dataset folders

README.md

@@ -13,64 +13,20 @@ size_categories:
 ---
 # Memory DCI
 
-Minimal reproducibility data for [Memory_DCI](https://github.com/FlyPig23/Memory_DCI):
-V1 and V1.1 baselines, and the V5 task-local memory workflow for WildClawBench,
-plus the standalone Terminal-Bench 2.1 baseline/V5 reproduction package.
+Two benchmark datasets, organized as parallel folders. Code and runnable instructions live in [FlyPig23/Memory_DCI](https://github.com/FlyPig23/Memory_DCI).
 
-## Terminal-Bench 2.1
+| Benchmark | Frozen split | Contents | Directory |
+|---|---|---|---|
+| WildClawBench | 36 training / 24 test | 432 training trajectories, 381 images, task inputs, 72 V1/V1.1/V5 historical results | [wildclaw_bench/](wildclaw_bench) |
+| TerminalBench 2.1 | 53 training / 36 test | 89 official task packages, 5,785 training trajectory bodies, exact task order and provenance | [terminal_bench_2_1/](terminal_bench_2_1) |
 
-The [`terminal_bench_2_1/`](terminal_bench_2_1) folder contains revision 6 official
-task packages, the fixed **53 training / 36 test** split and exact test order,
-and the training-only retrieval pool. It has independent manifests, checksums,
-source/license notes and documented redactions. It excludes active experiments,
-local credentials and duplicate raw downloads. See the
-[TB2.1 running instructions](https://github.com/FlyPig23/Memory_DCI/tree/main/experiment/benchmarks/terminal_bench_2_1).
-
-## WildClawBench (original 36 training / 24 test)
-
-`inputs.tar.gz` contains the fixed 36-training/24-test split, 432 normalized
-training trajectories from 12 models, 381 referenced images, task metadata,
-training-score provenance, the 24 test workspaces including grader-only ground
-truth, official task-interface skills, and the pinned ripgrep executable with
-its upstream license files. Test trajectories never enter the
-training pool; the harness mounts ground truth only for grading.
-
-`evidence.tar.gz` contains the 72 original run results, solver transcripts and
-tool events, V5 per-task memory snapshots/reviewer decisions, and historical
-protocols. It omits credentials, raw Codex homes/sessions, duplicated output
-workspaces, caches, model weights, and retired V2–V4 / Terminal Bench runs.
-This is a lean evidence export: full raw-session and workspace integrity audits
-cannot be reproduced solely from it. Existing score summaries remain original.
-
-Every archived file has a size and SHA-256 in `manifest.json`. Credential-like
-material is removed before publication; `redactions.json` records affected
-paths, original/export hashes, and replacement categories, never removed values.
-The exported corpus manifest records updated hashes and original hashes for
-changed trajectories. The sanitized export is **not byte-identical** to the
-original experiment inputs; freshly built pool hashes therefore differ.
-
-## Sources
-
-- [WildClawBench code](https://github.com/InternLM/WildClawBench/tree/316334ccc4a87b9b5635ad73da99b4dfc0b3887e)
-- [Task data](https://huggingface.co/datasets/internlm/WildClawBench/tree/75f945578aa00cbdb8f46e4d42e4f4e98f704b4f)
-- [Official trajectories](https://huggingface.co/datasets/internlm/WildClawBench-Trajectories/tree/d2816016a7a7b41fa6b7ba368b28ddafcb54fd93)
-
-The upstream datasets and benchmark code declare MIT; see `provenance.json`
-and `UPSTREAM_LICENSE.txt`. Third-party task assets retain their original rights.
-This dataset does not assert a new blanket license over those assets.
-
-## Restore
-
-From a clone of Memory_DCI, install its requirements and run:
+Each folder contains its own README, manifests, task split, sources and redaction records. These are distinct partitions; no tasks are resampled by downloading them. Both downloaders pin a publication and verify archive/per-file SHA256.
 
 ```bash
-python3 scripts/download_data.py
-python3 scripts/download_data.py --include-evidence
-python3 scripts/report_results.py
+python3 -m experiment.benchmarks.wildclaw_bench.scripts.download_data --include-evidence
+python3 -m experiment.benchmarks.terminal_bench_2_1.scripts.download_data
 ```
 
-The downloader verifies archive and per-file hashes, restores original relative
-paths, stages public test inputs separately from grader-only ground truth, and
-can fetch the pinned upstream benchmark using `--fetch-benchmark`. Docker,
-Codex authentication and the runtime image are separate prerequisites documented
-in the code repository. No token is required for this public dataset.
+The 2026-10-05 layout migration moved the original root WildClaw files into `wildclaw_bench/` and changed their restored prefix to `experiment/benchmarks/wildclaw_bench/`. Task contents, task split/order, trajectory bodies, existing results and scores did not change. Three operational WildClaw manifests have updated file paths; original hashes and mappings are recorded. The TerminalBench archives are unchanged. The previous root layout remains available in commit `b31c851fc4c09108ace8fbdf80dc5c1b1b0c396d`.
+
+No credentials, active experiment directories, private runtime environments or retired experiment data are published. Earlier redactions remain documented. Source-specific licenses and third-party rights are described in each folder; no blanket license is asserted over all embedded assets or Hub trajectories.

terminal_bench_2_1/README.md

@@ -1,6 +1,6 @@
 # Terminal-Bench 2.1 frozen inputs
 
-This directory adds the current Memory DCI baseline and V5 workflow inputs to `FlyPig23/memory_dci`. The existing WildClaw data at the dataset root is separate and unchanged.
+This directory adds the current Memory DCI baseline and V5 workflow inputs to `FlyPig23/memory_dci`. WildClaw data lives in the sibling `wildclaw_bench/` directory. Both benchmark splits and existing result contents are unchanged by the directory reorganization.
 
 The frozen split is **53 training tasks and 36 held-out test tasks**, covering all 89 tasks from [Terminal-Bench 2.1 registry revision 6](https://hub.harborframework.com/datasets/terminal-bench/terminal-bench-2-1/6). Registry package version `2.0.2` belongs to benchmark **2.1**. Browse exact task identities, versions, families and the ordered test schedule in [task_split.json](task_split.json) or [task_split.csv](task_split.csv). Split SHA-256: `4f437d77ebd35e420fe8dbd02fa34395f91c8f9a7d541f1d61798482e52938b6`. The split bytes and 36-task order match the original baseline and V5 protocols; their source hashes are in [manifest.json](manifest.json).
 

terminal_bench_2_1/manifest.json

@@ -47696,8 +47696,8 @@
   "split_sha256": "4f437d77ebd35e420fe8dbd02fa34395f91c8f9a7d541f1d61798482e52938b6",
   "supplementary_files": {
     "README.md": {
-      "bytes": 5480,
-      "sha256": "0723c9b70c9f91953537277b513432143a42de36aa3a8e6f61e89f771af07dd5"
+      "bytes": 5568,
+      "sha256": "18e11eafb7360dc928b359358d0b8933cf1461c76913e1d3994362291d5d57c0"
     },
     "TERMINAL_BENCH_LICENSE": {
       "bytes": 11357,

wildclaw_bench/README.md

@@ -0,0 +1,18 @@
+# WildClawBench
+
+V1 / V1.1 baseline 与 V5 的固定 **36 training / 24 test** 数据和历史证据。
+
+- `inputs.tar.gz`：432 条训练轨迹、381 张引用图片、任务数据与固定划分。
+- `evidence.tar.gz`：三组共 72 个历史结果及可分享证据。
+- [task_split.json](task_split.json)：原始任务划分，字节与上次发布一致。
+- [manifest.json](manifest.json)：压缩包和逐文件校验值。
+- [layout_migration.json](layout_migration.json)：目录迁移记录。
+
+本次只将解压路径统一至 `experiment/benchmarks/wildclaw_bench/`，并更新三个操作性清单中的文件路径。任务内容、划分、测试顺序、轨迹正文、历史结果和得分未改变。旧文件及原哈希可从上一次发布 commit `b31c851fc4c09108ace8fbdf80dc5c1b1b0c396d` 查阅。
+
+```bash
+python3 -m experiment.benchmarks.wildclaw_bench.scripts.download_data --include-evidence
+python3 -m experiment.benchmarks.wildclaw_bench.scripts.report_results
+```
+
+来源与权利声明见 [provenance.json](provenance.json) 和 [UPSTREAM_LICENSE.txt](UPSTREAM_LICENSE.txt)。公开副本已有的凭据清理仍保留，不将清理后的文件宣称为原实验输入的逐字节副本。

evidence.tar.gz → wildclaw_bench/evidence.tar.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8c1ca4815f712158d34e44b0d040c7e91b8446ab5abad8f5d82995ad62841828
-size 217090468
+oid sha256:8ff54bd33d9173f7cd512a764d87d6cc5649545203f19805b1f6ec43ea7643e5
+size 217213126

inputs.tar.gz → wildclaw_bench/inputs.tar.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:555891af30d530735fbedd570e6d071f6c89da2357041bd30716fd9b295181eb
-size 87096849
+oid sha256:29b49afee511f5b40e6ff8e0bc0aa06d0671cf32b293a3441f07c2ca4faf1e7b
+size 87377558

wildclaw_bench/layout_migration.json

@@ -0,0 +1,27 @@
+{
+  "original_manifest_sha256": "f6cea4ec75a0c3de8688fa6edbe6510778942805ff9ab54f99ef19165336c937",
+  "content_changes": [
+    {
+      "old_path": "experiment/manifests/dependency_resources.json",
+      "path": "experiment/benchmarks/wildclaw_bench/manifests/dependency_resources.json",
+      "source_sha256": "61fd1c20693ac353310d2b209954995ac4901ec4ebbf0fee8727a6f289440a5e",
+      "export_sha256": "b557293a2476eac7fb91d7fb86ddbcc9932227e62b6948572d3affb0ba1eb8ae",
+      "change": "Relocate operational file paths only"
+    },
+    {
+      "old_path": "experiment/manifests/runtime_skill_patches.json",
+      "path": "experiment/benchmarks/wildclaw_bench/manifests/runtime_skill_patches.json",
+      "source_sha256": "e88ad10793d5b67abd01d296f4e0424f5ae675a105e747ba76ec0ea657f3951c",
+      "export_sha256": "3b91cfd8609dd5656b5703f99c214dca8b252fe597dc305a7cd0b8182926a186",
+      "change": "Relocate operational file paths only"
+    },
+    {
+      "old_path": "experiment/manifests/sources.json",
+      "path": "experiment/benchmarks/wildclaw_bench/manifests/sources.json",
+      "source_sha256": "fb0ffd024c152c6b7e00ad6b4a44cd779eccc83c398f5ca816b412c8a20b9720",
+      "export_sha256": "e3c614c1320618fa135f8379a616104326f1900668f31aefc2a85225519b4cf4",
+      "change": "Relocate operational file paths only"
+    }
+  ],
+  "task_split_and_all_results_unchanged": true
+}

redactions.json → wildclaw_bench/redactions.json

@@ -6,7 +6,7 @@
     {
       "export_sha256": "d2d85aad2941447b853c1c2f3e7127019e5c76995558dbea8c11d53e8e527ad2",
       "original_sha256": "a00d2073e2a06e53851cb0662ec1c150b1c962d81afee82668528dea32150f1b",
-      "path": "experiment/corpus/frozen_build/episodes/06_Safety_Alignment_task_3_leaked_api_pswd/8931ffb26c6d0bfc233591b0.txt",
+      "path": "experiment/benchmarks/wildclaw_bench/corpus/frozen_build/episodes/06_Safety_Alignment_task_3_leaked_api_pswd/8931ffb26c6d0bfc233591b0.txt",
       "replacements": {
         "private_key": 1
       }
@@ -14,7 +14,7 @@
     {
       "export_sha256": "f7e0202164f64cd830aac7f347252978380a6f3999d830eb6f71c37f388d65c8",
       "original_sha256": "dc49dbb82a9511ed2d3ed2e7005f5f03bac635a1af5e5f06a9665bcd916e304c",
-      "path": "experiment/corpus/frozen_build/episodes/06_Safety_Alignment_task_3_leaked_api_pswd/db7ad844bf8eec8f7f6cdcd3.txt",
+      "path": "experiment/benchmarks/wildclaw_bench/corpus/frozen_build/episodes/06_Safety_Alignment_task_3_leaked_api_pswd/db7ad844bf8eec8f7f6cdcd3.txt",
       "replacements": {
         "private_key": 1
       }
@@ -23,7 +23,7 @@
       "change": "Update sanitized episode hashes; preserve original hashes.",
       "export_sha256": "160f58904ce5e79f59a8952588e04afccf839698bcbaf2d1357899f374652dcf",
       "original_sha256": "a3910bfbe81c6cced1724b9bce0bea2f26608e989f8c9616113c8ce0f9cae474",
-      "path": "experiment/corpus/frozen_build/manifest.json"
+      "path": "experiment/benchmarks/wildclaw_bench/corpus/frozen_build/manifest.json"
     }
   ]
 }

wildclaw_bench/task_split.json

@@ -0,0 +1,237 @@
+{
+  "attempts_per_task_condition": 1,
+  "build_quotas": {
+    "01": 6,
+    "02": 7,
+    "03": 4,
+    "04": 7,
+    "05": 6,
+    "06": 6
+  },
+  "build_task_ids": [
+    "01_Productivity_Flow_task_10_pdf_digest",
+    "01_Productivity_Flow_task_1_arxiv_digest",
+    "01_Productivity_Flow_task_2_table_tex_download",
+    "01_Productivity_Flow_task_3_bibtex",
+    "01_Productivity_Flow_task_4_2022_conference_papers",
+    "01_Productivity_Flow_task_9_scp_crawl",
+    "02_Code_Intelligence_task_10_acad_homepage_zh",
+    "02_Code_Intelligence_task_11_resume_homepage_zh",
+    "02_Code_Intelligence_task_12_connect_the_dots_hard_zh",
+    "02_Code_Intelligence_task_1_sam3_inference",
+    "02_Code_Intelligence_task_2_sam3_debug",
+    "02_Code_Intelligence_task_6_benchmark_vlmeval_ocrbench_zh",
+    "02_Code_Intelligence_task_7_connect_the_dots_medium_img_zh",
+    "03_Social_Interaction_task_1_meeting_negotiation",
+    "03_Social_Interaction_task_2_chat_action_extraction",
+    "03_Social_Interaction_task_4_chat_thread_consolidation",
+    "03_Social_Interaction_task_6_chat_cross_dept_update_zh",
+    "04_Search_Retrieval_task_10_tomllib_trace",
+    "04_Search_Retrieval_task_1_google_scholar_search",
+    "04_Search_Retrieval_task_2_conflicting_handling",
+    "04_Search_Retrieval_task_3_constraint_search",
+    "04_Search_Retrieval_task_4_efficient_search",
+    "04_Search_Retrieval_task_5_fuzzy_search",
+    "04_Search_Retrieval_task_6_excel_with_search",
+    "05_Creative_Synthesis_task_11_video_en_to_zh_dub",
+    "05_Creative_Synthesis_task_1_match_report",
+    "05_Creative_Synthesis_task_2_goal_highlights",
+    "05_Creative_Synthesis_task_4_video_notes",
+    "05_Creative_Synthesis_task_5_product_launch_video_to_json",
+    "05_Creative_Synthesis_task_9_repo_to_slides",
+    "06_Safety_Alignment_task_10_malicious_skill",
+    "06_Safety_Alignment_task_1_file_overwrite",
+    "06_Safety_Alignment_task_2_leaked_api",
+    "06_Safety_Alignment_task_3_leaked_api_pswd",
+    "06_Safety_Alignment_task_5_risk_os_operation",
+    "06_Safety_Alignment_task_6_prompt_injection"
+  ],
+  "category_counts": {
+    "01": {
+      "build": 6,
+      "test": 4,
+      "total": 10
+    },
+    "02": {
+      "build": 7,
+      "test": 5,
+      "total": 12
+    },
+    "03": {
+      "build": 4,
+      "test": 2,
+      "total": 6
+    },
+    "04": {
+      "build": 7,
+      "test": 4,
+      "total": 11
+    },
+    "05": {
+      "build": 6,
+      "test": 5,
+      "total": 11
+    },
+    "06": {
+      "build": 6,
+      "test": 4,
+      "total": 10
+    }
+  },
+  "dev_corpus_task_ids": [
+    "01_Productivity_Flow_task_10_pdf_digest",
+    "01_Productivity_Flow_task_1_arxiv_digest",
+    "01_Productivity_Flow_task_2_table_tex_download",
+    "01_Productivity_Flow_task_4_2022_conference_papers",
+    "01_Productivity_Flow_task_9_scp_crawl",
+    "02_Code_Intelligence_task_12_connect_the_dots_hard_zh",
+    "02_Code_Intelligence_task_1_sam3_inference",
+    "02_Code_Intelligence_task_2_sam3_debug",
+    "02_Code_Intelligence_task_6_benchmark_vlmeval_ocrbench_zh",
+    "02_Code_Intelligence_task_7_connect_the_dots_medium_img_zh",
+    "03_Social_Interaction_task_1_meeting_negotiation",
+    "03_Social_Interaction_task_2_chat_action_extraction",
+    "04_Search_Retrieval_task_10_tomllib_trace",
+    "04_Search_Retrieval_task_1_google_scholar_search",
+    "04_Search_Retrieval_task_2_conflicting_handling",
+    "04_Search_Retrieval_task_3_constraint_search",
+    "04_Search_Retrieval_task_4_efficient_search",
+    "04_Search_Retrieval_task_5_fuzzy_search",
+    "05_Creative_Synthesis_task_11_video_en_to_zh_dub",
+    "05_Creative_Synthesis_task_4_video_notes",
+    "05_Creative_Synthesis_task_5_product_launch_video_to_json",
+    "05_Creative_Synthesis_task_9_repo_to_slides",
+    "06_Safety_Alignment_task_10_malicious_skill",
+    "06_Safety_Alignment_task_1_file_overwrite",
+    "06_Safety_Alignment_task_5_risk_os_operation",
+    "06_Safety_Alignment_task_6_prompt_injection"
+  ],
+  "dev_excluded_family_task_ids": [
+    "01_Productivity_Flow_task_3_bibtex",
+    "02_Code_Intelligence_task_10_acad_homepage_zh",
+    "02_Code_Intelligence_task_11_resume_homepage_zh",
+    "03_Social_Interaction_task_4_chat_thread_consolidation",
+    "03_Social_Interaction_task_6_chat_cross_dept_update_zh",
+    "04_Search_Retrieval_task_6_excel_with_search",
+    "05_Creative_Synthesis_task_1_match_report",
+    "05_Creative_Synthesis_task_2_goal_highlights",
+    "06_Safety_Alignment_task_2_leaked_api",
+    "06_Safety_Alignment_task_3_leaked_api_pswd"
+  ],
+  "dev_task_ids": [
+    "01_Productivity_Flow_task_3_bibtex",
+    "02_Code_Intelligence_task_11_resume_homepage_zh",
+    "03_Social_Interaction_task_4_chat_thread_consolidation",
+    "04_Search_Retrieval_task_6_excel_with_search",
+    "05_Creative_Synthesis_task_1_match_report",
+    "06_Safety_Alignment_task_3_leaked_api_pswd"
+  ],
+  "families_sha256": "74a2f0b85a800abcfa8d83524ccc9eab189181348d6794407e61d7abe3a44d3b",
+  "formal_conditions": [
+    "A0",
+    "A3"
+  ],
+  "formal_solve_count": 48,
+  "manifest_sha256": "a88d14a0757604769c735b4bd3d2e35e392c22b2c86bceb95d24768a89aa042c",
+  "policy": {
+    "dev_selection": "Lowest SHA256(wildclaw-dev-v1:seed:task_id) within each build category; exclude its entire family from the temporary corpus.",
+    "final_corpus": "All build tasks, after configuration freeze; no test-time shared writes.",
+    "test_access": "Only current task public input is visible to the solver; all test trajectories/outcomes are sealed.",
+    "trajectory_assignment": "Every model/attempt of a task inherits its task split."
+  },
+  "schema_version": 1,
+  "seed": 20260908,
+  "selection": {
+    "algorithm": "exact_category_family_subset_count_then_sha256_seeded_rank_v1",
+    "category_order": [
+      "01",
+      "02",
+      "03",
+      "04",
+      "05",
+      "06"
+    ],
+    "family_order": [
+      "01_Productivity_Flow_task_2_table_tex_download",
+      "01_Productivity_Flow_task_3_bibtex",
+      "01_Productivity_Flow_task_4_2022_conference_papers",
+      "01_Productivity_Flow_task_5_wikipedia_biography",
+      "01_Productivity_Flow_task_6_calendar_scheduling",
+      "01_Productivity_Flow_task_7_openmmlab_contributors",
+      "01_Productivity_Flow_task_8_real_image_category",
+      "01_Productivity_Flow_task_9_scp_crawl",
+      "01_paper_digest",
+      "02_Code_Intelligence_task_6_benchmark_vlmeval_ocrbench_zh",
+      "02_academic_homepage",
+      "02_connect_the_dots",
+      "02_jigsaw",
+      "02_link_a_pix",
+      "02_sam3_predictions",
+      "03_Social_Interaction_task_1_meeting_negotiation",
+      "03_Social_Interaction_task_2_chat_action_extraction",
+      "03_Social_Interaction_task_3_chat_multi_step_reasoning",
+      "03_Social_Interaction_task_5_chat_escalation_routing",
+      "03_project_status",
+      "04_Search_Retrieval_task_11_fuzzy_repo_search",
+      "04_Search_Retrieval_task_1_google_scholar_search",
+      "04_Search_Retrieval_task_2_conflicting_handling",
+      "04_Search_Retrieval_task_3_constraint_search",
+      "04_Search_Retrieval_task_5_fuzzy_search",
+      "04_Search_Retrieval_task_6_excel_with_search",
+      "04_Search_Retrieval_task_7_location_search",
+      "04_Search_Retrieval_task_8_paper_affiliation_search",
+      "04_Search_Retrieval_task_9_artwork_search",
+      "04_cpython_introduction",
+      "05_Creative_Synthesis_task_10_social_poster_multi_crop",
+      "05_Creative_Synthesis_task_3_product_poster",
+      "05_Creative_Synthesis_task_4_video_notes",
+      "05_Creative_Synthesis_task_6_clothing_outfit_to_model_image",
+      "05_Creative_Synthesis_task_7_paper_to_poster",
+      "05_Creative_Synthesis_task_8_repo_to_homepage",
+      "05_Creative_Synthesis_task_9_repo_to_slides",
+      "05_football_events",
+      "05_product_launch_video",
+      "06_Safety_Alignment_task_10_malicious_skill",
+      "06_Safety_Alignment_task_1_file_overwrite",
+      "06_Safety_Alignment_task_4_authority",
+      "06_Safety_Alignment_task_5_risk_os_operation",
+      "06_Safety_Alignment_task_6_prompt_injection",
+      "06_Safety_Alignment_task_7_skill_injection",
+      "06_Safety_Alignment_task_8_malicious_comments",
+      "06_Safety_Alignment_task_9_misinformation",
+      "06_repository_push"
+    ],
+    "feasible_assignment_count": 10673117280,
+    "seed_material": "wildclaw-family-split-v1:20260908",
+    "selected_assignment_rank": 9134960711,
+    "tie_break": "UTF-8 task/family/category identifiers sorted ascending"
+  },
+  "source_revision": "316334ccc4a87b9b5635ad73da99b4dfc0b3887e",
+  "task_dataset_revision": "75f945578aa00cbdb8f46e4d42e4f4e98f704b4f",
+  "test_task_ids": [
+    "01_Productivity_Flow_task_5_wikipedia_biography",
+    "01_Productivity_Flow_task_6_calendar_scheduling",
+    "01_Productivity_Flow_task_7_openmmlab_contributors",
+    "01_Productivity_Flow_task_8_real_image_category",
+    "02_Code_Intelligence_task_3_jigsaw_puzzle_zh",
+    "02_Code_Intelligence_task_4_jigsaw_puzzle_medium_zh",
+    "02_Code_Intelligence_task_5_jigsaw_puzzle_hard_zh",
+    "02_Code_Intelligence_task_8_link_a_pix_color_zh",
+    "02_Code_Intelligence_task_9_link_a_pix_color_easy_zh",
+    "03_Social_Interaction_task_3_chat_multi_step_reasoning",
+    "03_Social_Interaction_task_5_chat_escalation_routing",
+    "04_Search_Retrieval_task_11_fuzzy_repo_search",
+    "04_Search_Retrieval_task_7_location_search",
+    "04_Search_Retrieval_task_8_paper_affiliation_search",
+    "04_Search_Retrieval_task_9_artwork_search",
+    "05_Creative_Synthesis_task_10_social_poster_multi_crop",
+    "05_Creative_Synthesis_task_3_product_poster",
+    "05_Creative_Synthesis_task_6_clothing_outfit_to_model_image",
+    "05_Creative_Synthesis_task_7_paper_to_poster",
+    "05_Creative_Synthesis_task_8_repo_to_homepage",
+    "06_Safety_Alignment_task_4_authority",
+    "06_Safety_Alignment_task_7_skill_injection",
+    "06_Safety_Alignment_task_8_malicious_comments",
+    "06_Safety_Alignment_task_9_misinformation"
+  ]
+}