SupingSun commited on
Commit
d2873b3
·
verified ·
1 Parent(s): d22ec12

Add verified partial paper bundle

Browse files
.gitattributes CHANGED
@@ -1 +1,17 @@
1
  *.gz filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  *.gz filter=lfs diff=lfs merge=lfs -text
2
+ task_pool/013-2026.iclr.10007159.pdf filter=lfs diff=lfs merge=lfs -text
3
+ task_pool/017-2026.acl-long.154.pdf filter=lfs diff=lfs merge=lfs -text
4
+ task_pool/019-2026.iclr.10010470.pdf filter=lfs diff=lfs merge=lfs -text
5
+ task_pool/038-2026.iclr.10010317.pdf filter=lfs diff=lfs merge=lfs -text
6
+ task_pool/044-2026.acl-long.704.pdf filter=lfs diff=lfs merge=lfs -text
7
+ task_pool/047-2026.acl-long.1062.pdf filter=lfs diff=lfs merge=lfs -text
8
+ task_pool/052-2026.acl-long.2151.pdf filter=lfs diff=lfs merge=lfs -text
9
+ task_pool/069-2026.iclr.10010430.pdf filter=lfs diff=lfs merge=lfs -text
10
+ task_pool/078-2025.iccv.Zhu_Training-free_Geometric_Image_Editing_on_Diffusion_Models_ICCV_2025.pdf filter=lfs diff=lfs merge=lfs -text
11
+ task_pool/084-2026.acl-long.2059.pdf filter=lfs diff=lfs merge=lfs -text
12
+ task_pool/090-2026.iclr.10009775.pdf filter=lfs diff=lfs merge=lfs -text
13
+ task_pool/099-2026.iclr.10010023.pdf filter=lfs diff=lfs merge=lfs -text
14
+ task_pool/118-2025.emnlp.1496.pdf filter=lfs diff=lfs merge=lfs -text
15
+ task_pool/119-2026.acl-long.358.pdf filter=lfs diff=lfs merge=lfs -text
16
+ task_pool/132-2026.acl-long.1010.pdf filter=lfs diff=lfs merge=lfs -text
17
+ task_pool/138-2026.iclr.10008475.pdf filter=lfs diff=lfs merge=lfs -text
DATASET_MANIFEST.json CHANGED
@@ -10,9 +10,10 @@
10
  "trajectory_file_count": 423,
11
  "trajectory_compressed_bytes": 398450656,
12
  "trajectory_file_format": "redacted agent.log.gz",
13
- "paper_pdf_count": 0,
 
14
  "paper_pdf_layout": "task_pool/<project_id>.pdf",
15
- "paper_pdf_status": "not_released_pending_source_and_redistribution_review",
16
  "trajectory_artifact_counts": {
17
  "complete_evidence_bundle": 418,
18
  "raw_unresolved_trajectory": 4,
@@ -23,7 +24,7 @@
23
  "leaderboard and ranking results",
24
  "datasets and model weights",
25
  "complete third-party repositories",
26
- "source-paper PDFs without verified redistribution terms",
27
  "raw security telemetry",
28
  "provider API payloads and credentials",
29
  "private machine records"
 
10
  "trajectory_file_count": 423,
11
  "trajectory_compressed_bytes": 398450656,
12
  "trajectory_file_format": "redacted agent.log.gz",
13
+ "paper_pdf_count": 16,
14
+ "paper_pdf_bytes": 74441999,
15
  "paper_pdf_layout": "task_pool/<project_id>.pdf",
16
+ "paper_pdf_status": "partial_16_of_47_redistribution_unverified",
17
  "trajectory_artifact_counts": {
18
  "complete_evidence_bundle": 418,
19
  "raw_unresolved_trajectory": 4,
 
24
  "leaderboard and ranking results",
25
  "datasets and model weights",
26
  "complete third-party repositories",
27
+ "31 source-paper PDFs not yet mapped to verified files",
28
  "raw security telemetry",
29
  "provider API payloads and credentials",
30
  "private machine records"
DATASET_SHA256SUMS CHANGED
@@ -1,9 +1,26 @@
1
  11c164d4ef2972467c5903e69b83383fed0bdbb77bf9aedeeba2820026abf0e9 .gitattributes
2
- bbdb49ecd9b72c7ef39e67516429acfd27f905e0225551e723af20b71c03cc53 DATASET_MANIFEST.json
3
- 5872fad71a3848fde032bd0adcb743c0c4b0c6c40adef09fe39cf68422d2e226 README.md
 
4
  f7e1ea72e5e6ba74052abf5a569e99ebf34df8f1eac28953e98cd5bd8fab1d23 assets/construction-pipeline.svg
5
  c328f27f2949a4c7fd74432dbf0baa62151d09706dd23eefd6ff7bc4fc047774 docs/REDACTION_POLICY.md
6
- 6f08fdc0672ca7f80cccf408737eefabebe03c9cadeea4d5ca5d0305ec363982 scripts/verify_dataset.py
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7
  77305a5202366f9a8a564c66f93c80ebe3e875b86155d08fc1bc0c3e65b423e5 trajectories/TRAJECTORY_INDEX.csv
8
  0e0c43e8759f91fb0333af76cbaaea710569791a1c7a055cf51f690e3c7be080 trajectories/runs/claude-opus-4-8-cc/claude/013-2026.iclr.10007159/agent.log.gz
9
  f1d1e4a759fb54ddff1f4b66f61c600fe0f12252682c4f89e5406448c4854874 trajectories/runs/claude-opus-4-8-cc/claude/016-2025.emnlp.109/agent.log.gz
 
1
  11c164d4ef2972467c5903e69b83383fed0bdbb77bf9aedeeba2820026abf0e9 .gitattributes
2
+ f2e6f8e0c4fb7d8caefaa031707e0ba1dc1d71829aec860cee22326dae14eb81 DATASET_MANIFEST.json
3
+ f9770872b575fdb01b56ff09c1eb46be372b48a75418739e888f78feea6bc652 PAPER_INDEX.csv
4
+ a2c2a5468b6a72d9cd3e0a2212949128ecefc50a91a3ad4c6ebb70fb993b8e4d README.md
5
  f7e1ea72e5e6ba74052abf5a569e99ebf34df8f1eac28953e98cd5bd8fab1d23 assets/construction-pipeline.svg
6
  c328f27f2949a4c7fd74432dbf0baa62151d09706dd23eefd6ff7bc4fc047774 docs/REDACTION_POLICY.md
7
+ 17f70ee9122e21e8b29af555d091db9763aac99e749c7295f7332ca214aacf9c scripts/verify_dataset.py
8
+ 7ae4d3e65301d258ffb4cf0429f18a80d4e0340f2a7b18c0f20b6c0280c0bc5f task_pool/013-2026.iclr.10007159.pdf
9
+ 430be4e1ddb509d5cb0f3b5d081efce1d86f528ff714bc5480b90cd8ad7df7a6 task_pool/017-2026.acl-long.154.pdf
10
+ 1d43629bb77077d60e484b6b868454be1f317ba031a3ae175d8bb121af9fedb6 task_pool/019-2026.iclr.10010470.pdf
11
+ e26c460dbae97c3c62a166a43740645632c33a0c1d4eba11ae843e7e68dc6381 task_pool/038-2026.iclr.10010317.pdf
12
+ e50d243b1f547e9a11e3756e25aab7608b2a06894f9525a1050a74485f8dff3a task_pool/044-2026.acl-long.704.pdf
13
+ 0b92c0e253dcdfb83408e97466bfa21838f112494dcbf77ec0c5031a2e259c5a task_pool/047-2026.acl-long.1062.pdf
14
+ 4aa76850f46d9dac947297e70330b70e4a0726932fcb62cc4f1dacc7e13e3bdf task_pool/052-2026.acl-long.2151.pdf
15
+ a5496c4ffb59f8defc0683e9ee0b3f0e3e43b5c1f0a4853202f9feba455a98e8 task_pool/069-2026.iclr.10010430.pdf
16
+ 46b56fbb28d9ce9279a7a346c0549de74eb32a0e2fec1d71bebb15a761c69d8f task_pool/078-2025.iccv.Zhu_Training-free_Geometric_Image_Editing_on_Diffusion_Models_ICCV_2025.pdf
17
+ 13a3d0a065948d3ceb509218c05bc519be75c524796bc2638da82ba2f9414d19 task_pool/084-2026.acl-long.2059.pdf
18
+ 7493ef5a96ba07e13e55b1a11d9c435d2942313efa136d3aeb5a990259ffa64a task_pool/090-2026.iclr.10009775.pdf
19
+ 910af2b7672c5a277cba1967688438525f3c334605301b1bdffa0441ee04cc17 task_pool/099-2026.iclr.10010023.pdf
20
+ c30f00aa35b6e80e32a528c3da5c0c84b0b61f5d333ad1b23cdf8a06e2d4b78f task_pool/118-2025.emnlp.1496.pdf
21
+ b0031d04af12d9df2358513e085c15828a7f0424daca931a0792fad28bb9c593 task_pool/119-2026.acl-long.358.pdf
22
+ fd1b4145b556ee4b100e72f93de82abe19851e4c50a49c289dfde8eb59db259d task_pool/132-2026.acl-long.1010.pdf
23
+ d8761fba88724b6239c384e17cec0bf6c01e9dba0f032ea66f54950dfd0f107a task_pool/138-2026.iclr.10008475.pdf
24
  77305a5202366f9a8a564c66f93c80ebe3e875b86155d08fc1bc0c3e65b423e5 trajectories/TRAJECTORY_INDEX.csv
25
  0e0c43e8759f91fb0333af76cbaaea710569791a1c7a055cf51f690e3c7be080 trajectories/runs/claude-opus-4-8-cc/claude/013-2026.iclr.10007159/agent.log.gz
26
  f1d1e4a759fb54ddff1f4b66f61c600fe0f12252682c4f89e5406448c4854874 trajectories/runs/claude-opus-4-8-cc/claude/016-2025.emnlp.109/agent.log.gz
PAPER_INDEX.csv ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ project_id,paper_title,pdf_path,source_url,sha256,bytes,page_count,redistribution_status
2
+ 013-2026.iclr.10007159,The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs,task_pool/013-2026.iclr.10007159.pdf,https://arxiv.org/pdf/2507.11097,7ae4d3e65301d258ffb4cf0429f18a80d4e0340f2a7b18c0f20b6c0280c0bc5f,1961703,30,unverified
3
+ 017-2026.acl-long.154,Multi-component Causal Tracing in Large Language Models,task_pool/017-2026.acl-long.154.pdf,https://aclanthology.org/2026.acl-long.154.pdf,430be4e1ddb509d5cb0f3b5d081efce1d86f528ff714bc5480b90cd8ad7df7a6,1469425,21,unverified
4
+ 019-2026.iclr.10010470,SinkTrack: Attention Sink based Context Anchoring for Large Language Models,task_pool/019-2026.iclr.10010470.pdf,https://arxiv.org/pdf/2604.10027,1d43629bb77077d60e484b6b868454be1f317ba031a3ae175d8bb121af9fedb6,1718027,17,unverified
5
+ 038-2026.iclr.10010317,Prosperity before Collapse: How Far Can Off-Policy RL Reach with Stale Data on LLMs?,task_pool/038-2026.iclr.10010317.pdf,https://arxiv.org/pdf/2510.01161,e26c460dbae97c3c62a166a43740645632c33a0c1d4eba11ae843e7e68dc6381,1115995,16,unverified
6
+ 044-2026.acl-long.704,Hallucination Detection in LLMs with Topological Divergence on Attention Graphs,task_pool/044-2026.acl-long.704.pdf,https://aclanthology.org/2026.acl-long.704.pdf,e50d243b1f547e9a11e3756e25aab7608b2a06894f9525a1050a74485f8dff3a,1217786,22,unverified
7
+ 047-2026.acl-long.1062,Aligning Agents via Planning: A Benchmark for Trajectory-Level Reward Modeling,task_pool/047-2026.acl-long.1062.pdf,https://aclanthology.org/2026.acl-long.1062.pdf,0b92c0e253dcdfb83408e97466bfa21838f112494dcbf77ec0c5031a2e259c5a,1014378,27,unverified
8
+ 052-2026.acl-long.2151,DPEPO: Diverse Parallel Exploration Policy Optimization for LLM-based Agents,task_pool/052-2026.acl-long.2151.pdf,https://aclanthology.org/2026.acl-long.2151.pdf,4aa76850f46d9dac947297e70330b70e4a0726932fcb62cc4f1dacc7e13e3bdf,649673,23,unverified
9
+ 069-2026.iclr.10010430,FlashVID: Efficient Video Large Language Models via Training-free Tree-based Spatiotemporal Token Merging,task_pool/069-2026.iclr.10010430.pdf,https://arxiv.org/pdf/2602.08024,a5496c4ffb59f8defc0683e9ee0b3f0e3e43b5c1f0a4853202f9feba455a98e8,13500070,30,unverified
10
+ 078-2025.iccv.Zhu_Training-free_Geometric_Image_Editing_on_Diffusion_Models_ICCV_2025,Training-free Geometric Image Editing on Diffusion Models,task_pool/078-2025.iccv.Zhu_Training-free_Geometric_Image_Editing_on_Diffusion_Models_ICCV_2025.pdf,https://openaccess.thecvf.com/content/ICCV2025/papers/Zhu_Training-free_Geometric_Image_Editing_on_Diffusion_Models_ICCV_2025_paper.pdf,46b56fbb28d9ce9279a7a346c0549de74eb32a0e2fec1d71bebb15a761c69d8f,7327301,11,unverified
11
+ 084-2026.acl-long.2059,Through the Magnifying Glass: Adaptive Perception Magnification for Hallucination-Free VLM Decoding,task_pool/084-2026.acl-long.2059.pdf,https://aclanthology.org/2026.acl-long.2059.pdf,13a3d0a065948d3ceb509218c05bc519be75c524796bc2638da82ba2f9414d19,13486798,22,unverified
12
+ 090-2026.iclr.10009775,PPE: Positional Preservation Embedding for Token Compression in Multimodal Large Language Models,task_pool/090-2026.iclr.10009775.pdf,https://arxiv.org/pdf/2510.22936,7493ef5a96ba07e13e55b1a11d9c435d2942313efa136d3aeb5a990259ffa64a,8065419,21,unverified
13
+ 099-2026.iclr.10010023,Overcoming Joint Intractability with Lossless Hierarchical Speculative Decoding,task_pool/099-2026.iclr.10010023.pdf,https://arxiv.org/pdf/2601.05724,910af2b7672c5a277cba1967688438525f3c334605301b1bdffa0441ee04cc17,855438,34,unverified
14
+ 118-2025.emnlp.1496,A Causal Lens for Evaluating Faithfulness Metrics,task_pool/118-2025.emnlp.1496.pdf,https://aclanthology.org/2025.emnlp-main.1496.pdf,c30f00aa35b6e80e32a528c3da5c0c84b0b61f5d333ad1b23cdf8a06e2d4b78f,1037284,25,unverified
15
+ 119-2026.acl-long.358,GIFT: Guided Fine-Tuning and Transfer for Enhancing Instruction-Tuned Language Models,task_pool/119-2026.acl-long.358.pdf,https://aclanthology.org/2026.acl-long.358.pdf,b0031d04af12d9df2358513e085c15828a7f0424daca931a0792fad28bb9c593,2876442,13,unverified
16
+ 132-2026.acl-long.1010,Counteracting the Matthew Effect in Self-Improvement of LVLMs through Head-Tail Re-balancing,task_pool/132-2026.acl-long.1010.pdf,https://aclanthology.org/2026.acl-long.1010.pdf,fd1b4145b556ee4b100e72f93de82abe19851e4c50a49c289dfde8eb59db259d,8566460,18,unverified
17
+ 138-2026.iclr.10008475,SoFlow: Solution Flow Models for One-Step Generative Modeling,task_pool/138-2026.iclr.10008475.pdf,https://arxiv.org/pdf/2512.15657,d8761fba88724b6239c384e17cec0bf6c01e9dba0f032ea66f54950dfd0f107a,9579800,19,unverified
README.md CHANGED
@@ -47,13 +47,20 @@ redacted `agent.log.gz` files. The authoritative task catalog remains on
47
  [GitHub](https://github.com/ReproBoundary/ReproBoundary/blob/main/tasks/TASK_INVENTORY_47.csv)
48
  instead of being duplicated here.
49
 
 
 
 
 
50
  ```text
51
  ReproBoundary-Execution-Trajectories/
52
  ├── README.md
53
  ├── DATASET_MANIFEST.json
54
  ├── DATASET_SHA256SUMS
 
55
  ├── assets/
56
  │ └── construction-pipeline.svg
 
 
57
  ├── trajectories/
58
  │ ├── TRAJECTORY_INDEX.csv
59
  │ └── runs/
@@ -64,9 +71,9 @@ ReproBoundary-Execution-Trajectories/
64
  └── verify_dataset.py
65
  ```
66
 
67
- Source-paper PDFs are not part of the current snapshot. If their official
68
- sources and redistribution terms are verified, they can be added separately as
69
- `task_pool/<project_id>.pdf`; an unverified local PDF will not be published.
70
 
71
  Hugging Face does not render the gzip logs as table rows. Start with
72
  `trajectories/TRAJECTORY_INDEX.csv`, then open only the logs needed for an
@@ -74,8 +81,9 @@ analysis.
74
 
75
  ## Download and Verify
76
 
77
- The public trajectory snapshot is fixed at revision
78
- `57e17e11f795c6df9f7806ad2262fd32897ec2e6`. No account or token is required.
 
79
 
80
  ```bash
81
  python3 -m venv .hf-venv
@@ -84,7 +92,7 @@ python -m pip install "huggingface_hub>=1.0,<2"
84
 
85
  hf download SupingSun/ReproBoundary-Execution-Trajectories \
86
  --repo-type dataset \
87
- --revision 57e17e11f795c6df9f7806ad2262fd32897ec2e6 \
88
  --local-dir ReproBoundary-Execution-Trajectories
89
 
90
  cd ReproBoundary-Execution-Trajectories
@@ -94,10 +102,10 @@ python3 scripts/verify_dataset.py
94
  Expected output:
95
 
96
  ```text
97
- verification OK: projects=47 settings=9 trajectories=423
98
  ```
99
 
100
- The download is about 399 MB; allow at least 1 GB of free space. The verifier
101
  checks all indexed paths and hashes, gzip readability, release checksums, and
102
  configured sensitive-text patterns.
103
 
 
47
  [GitHub](https://github.com/ReproBoundary/ReproBoundary/blob/main/tasks/TASK_INVENTORY_47.csv)
48
  instead of being duplicated here.
49
 
50
+ `task_pool/` currently contains 16 source-paper PDFs with exact project IDs.
51
+ `PAPER_INDEX.csv` records their titles, official source URLs, sizes, page counts,
52
+ and SHA-256 fingerprints. The remaining 31 papers are not yet included.
53
+
54
  ```text
55
  ReproBoundary-Execution-Trajectories/
56
  ├── README.md
57
  ├── DATASET_MANIFEST.json
58
  ├── DATASET_SHA256SUMS
59
+ ├── PAPER_INDEX.csv
60
  ├── assets/
61
  │ └── construction-pipeline.svg
62
+ ├── task_pool/
63
+ │ └── <project_id>.pdf
64
  ├── trajectories/
65
  │ ├── TRAJECTORY_INDEX.csv
66
  │ └── runs/
 
71
  └── verify_dataset.py
72
  ```
73
 
74
+ The 16 PDFs were matched to the frozen project IDs and checked for readable PDF
75
+ structure and title agreement. Their redistribution terms remain unverified;
76
+ consult each official source before reusing or redistributing a paper.
77
 
78
  Hugging Face does not render the gzip logs as table rows. Start with
79
  `trajectories/TRAJECTORY_INDEX.csv`, then open only the logs needed for an
 
81
 
82
  ## Download and Verify
83
 
84
+ The 423-log payload is inherited from fixed revision
85
+ `57e17e11f795c6df9f7806ad2262fd32897ec2e6`. Download `main` for the current
86
+ log index, paper bundle, and verifier. No account or token is required.
87
 
88
  ```bash
89
  python3 -m venv .hf-venv
 
92
 
93
  hf download SupingSun/ReproBoundary-Execution-Trajectories \
94
  --repo-type dataset \
95
+ --revision main \
96
  --local-dir ReproBoundary-Execution-Trajectories
97
 
98
  cd ReproBoundary-Execution-Trajectories
 
102
  Expected output:
103
 
104
  ```text
105
+ verification OK: projects=47 settings=9 trajectories=423 papers=16
106
  ```
107
 
108
+ The download is about 470 MB; allow at least 1 GB of free space. The verifier
109
  checks all indexed paths and hashes, gzip readability, release checksums, and
110
  configured sensitive-text patterns.
111
 
scripts/verify_dataset.py CHANGED
@@ -27,6 +27,9 @@ runs = list(
27
  (ROOT / "trajectories/TRAJECTORY_INDEX.csv").open(encoding="utf-8")
28
  )
29
  )
 
 
 
30
 
31
  project_ids = {row["project"] for row in runs}
32
  task_ids = {row["task_id"] for row in runs}
@@ -38,6 +41,31 @@ if len(runs) != 423 or len(task_ids) != 423:
38
  errors.append("trajectory index is not 423 unique assignments")
39
  if len(settings) != 9:
40
  errors.append("trajectory index is not nine unique model-agent settings")
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
41
  strong_patterns = {
42
  "local_path": re.compile(r"/Users/ssp|/lustre/prod_glm_volumes"),
43
  "api_key": re.compile(r"(?<![A-Za-z0-9])sk-(?:ant-)?[A-Za-z0-9_-]{12,}"),
@@ -86,6 +114,8 @@ if manifest.get("model_agent_setting_count") != 9:
86
  errors.append("manifest setting count is not nine")
87
  if manifest.get("trajectory_file_count") != 423:
88
  errors.append("manifest trajectory count is not 423")
 
 
89
 
90
  checksum_file = ROOT / "DATASET_SHA256SUMS"
91
  for raw in checksum_file.read_text(encoding="utf-8").splitlines():
@@ -102,5 +132,6 @@ if errors:
102
 
103
  print(
104
  "verification OK: "
105
- f"projects={len(project_ids)} settings={len(settings)} trajectories={len(runs)}"
 
106
  )
 
27
  (ROOT / "trajectories/TRAJECTORY_INDEX.csv").open(encoding="utf-8")
28
  )
29
  )
30
+ papers = list(
31
+ csv.DictReader((ROOT / "PAPER_INDEX.csv").open(encoding="utf-8-sig"))
32
+ )
33
 
34
  project_ids = {row["project"] for row in runs}
35
  task_ids = {row["task_id"] for row in runs}
 
41
  errors.append("trajectory index is not 423 unique assignments")
42
  if len(settings) != 9:
43
  errors.append("trajectory index is not nine unique model-agent settings")
44
+
45
+ paper_ids = {row["project_id"] for row in papers}
46
+ paper_paths = {row["pdf_path"] for row in papers}
47
+ if len(papers) != 16 or len(paper_ids) != 16:
48
+ errors.append("paper index is not 16 unique projects")
49
+ if not paper_ids.issubset(project_ids):
50
+ errors.append("paper index contains projects outside the trajectory index")
51
+ actual_papers = {
52
+ str(path.relative_to(ROOT)) for path in (ROOT / "task_pool").glob("*.pdf")
53
+ }
54
+ if actual_papers != paper_paths:
55
+ errors.append("paper files do not match the paper index")
56
+ for row in papers:
57
+ paper = ROOT / row["pdf_path"]
58
+ if not paper.is_file():
59
+ errors.append(f"missing paper: {row['project_id']}")
60
+ continue
61
+ if paper.read_bytes()[:5] != b"%PDF-":
62
+ errors.append(f"invalid PDF header: {row['project_id']}")
63
+ if sha256(paper) != row["sha256"]:
64
+ errors.append(f"paper hash mismatch: {row['project_id']}")
65
+ if paper.stat().st_size != int(row["bytes"]):
66
+ errors.append(f"paper size mismatch: {row['project_id']}")
67
+ if not row["source_url"].startswith("https://"):
68
+ errors.append(f"paper source is not HTTPS: {row['project_id']}")
69
  strong_patterns = {
70
  "local_path": re.compile(r"/Users/ssp|/lustre/prod_glm_volumes"),
71
  "api_key": re.compile(r"(?<![A-Za-z0-9])sk-(?:ant-)?[A-Za-z0-9_-]{12,}"),
 
114
  errors.append("manifest setting count is not nine")
115
  if manifest.get("trajectory_file_count") != 423:
116
  errors.append("manifest trajectory count is not 423")
117
+ if manifest.get("paper_pdf_count") != len(papers):
118
+ errors.append("manifest paper count does not match the paper index")
119
 
120
  checksum_file = ROOT / "DATASET_SHA256SUMS"
121
  for raw in checksum_file.read_text(encoding="utf-8").splitlines():
 
132
 
133
  print(
134
  "verification OK: "
135
+ f"projects={len(project_ids)} settings={len(settings)} "
136
+ f"trajectories={len(runs)} papers={len(papers)}"
137
  )
task_pool/013-2026.iclr.10007159.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ae4d3e65301d258ffb4cf0429f18a80d4e0340f2a7b18c0f20b6c0280c0bc5f
3
+ size 1961703
task_pool/017-2026.acl-long.154.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:430be4e1ddb509d5cb0f3b5d081efce1d86f528ff714bc5480b90cd8ad7df7a6
3
+ size 1469425
task_pool/019-2026.iclr.10010470.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d43629bb77077d60e484b6b868454be1f317ba031a3ae175d8bb121af9fedb6
3
+ size 1718027
task_pool/038-2026.iclr.10010317.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e26c460dbae97c3c62a166a43740645632c33a0c1d4eba11ae843e7e68dc6381
3
+ size 1115995
task_pool/044-2026.acl-long.704.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e50d243b1f547e9a11e3756e25aab7608b2a06894f9525a1050a74485f8dff3a
3
+ size 1217786
task_pool/047-2026.acl-long.1062.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b92c0e253dcdfb83408e97466bfa21838f112494dcbf77ec0c5031a2e259c5a
3
+ size 1014378
task_pool/052-2026.acl-long.2151.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4aa76850f46d9dac947297e70330b70e4a0726932fcb62cc4f1dacc7e13e3bdf
3
+ size 649673
task_pool/069-2026.iclr.10010430.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5496c4ffb59f8defc0683e9ee0b3f0e3e43b5c1f0a4853202f9feba455a98e8
3
+ size 13500070
task_pool/078-2025.iccv.Zhu_Training-free_Geometric_Image_Editing_on_Diffusion_Models_ICCV_2025.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46b56fbb28d9ce9279a7a346c0549de74eb32a0e2fec1d71bebb15a761c69d8f
3
+ size 7327301
task_pool/084-2026.acl-long.2059.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13a3d0a065948d3ceb509218c05bc519be75c524796bc2638da82ba2f9414d19
3
+ size 13486798
task_pool/090-2026.iclr.10009775.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7493ef5a96ba07e13e55b1a11d9c435d2942313efa136d3aeb5a990259ffa64a
3
+ size 8065419
task_pool/099-2026.iclr.10010023.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:910af2b7672c5a277cba1967688438525f3c334605301b1bdffa0441ee04cc17
3
+ size 855438
task_pool/118-2025.emnlp.1496.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c30f00aa35b6e80e32a528c3da5c0c84b0b61f5d333ad1b23cdf8a06e2d4b78f
3
+ size 1037284
task_pool/119-2026.acl-long.358.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0031d04af12d9df2358513e085c15828a7f0424daca931a0792fad28bb9c593
3
+ size 2876442
task_pool/132-2026.acl-long.1010.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd1b4145b556ee4b100e72f93de82abe19851e4c50a49c289dfde8eb59db259d
3
+ size 8566460
task_pool/138-2026.iclr.10008475.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8761fba88724b6239c384e17cec0bf6c01e9dba0f032ea66f54950dfd0f107a
3
+ size 9579800