Add original FACTPROP graph checkpoint and verified loader

GRAPH.md

@@ -0,0 +1,39 @@
+# FACTPROP graph checkpoint
+
+`final.pkl` is the unchanged original checkpoint, 125,274,720 bytes.
+
+SHA-256: `437a434260edbb019c85575d65b4775cb2461f145e17966ee3acc8e4625ce7c8`
+
+## Structure
+
+The pickle contains a dictionary. `payload["graph"]` is a NetworkX **DiGraph**, with 100,015 nodes and 432,562 stored directed edges. Additional keys are `state`, `seed_entities`, `validator_state`, and `scheduler_stats`, retained as original construction metadata.
+
+Node identifiers are entity-label strings. Node attributes are `qid` and `qid_status`; QID mappings may be missing or shared by multiple nodes.
+
+| Edge field | Meaning |
+| --- | --- |
+| `relation` | Relation identifier |
+| `question` | Associated natural-language question |
+| `surface` | Natural-language statement |
+| `evidence` | Stored supporting text |
+| `confidence` | Construction-time confidence value |
+| `group` | Construction category |
+| `is_inverse` | Whether the edge is an inverse traversal edge |
+
+There are 357,205 forward edges and 75,357 inverse edges. Popularity counts incoming edges with `is_inverse is not True`. Do not use total in-degree without filtering inverse edges when comparing to the browser index. Field presence does not guarantee that all values are non-empty or independently correct.
+
+## Load and inspect
+
+Install `networkx` and `huggingface_hub`, then run the accompanying `load_graph.py`. It downloads the graph, checks its checksum, loads the dictionary, and prints graph counts. Python pickle loading can execute code, so only load a checkpoint you trust.
+
+```python
+from load_graph import load_graph
+
+graph = load_graph()
+entity = "Apple Inc."
+score = sum(1 for _, _, attrs in graph.in_edges(entity, data=True)
+            if attrs.get("is_inverse") is not True)
+assert score == 467
+```
+
+No graph nodes, mappings, edges, or checkpoint metadata were changed for this release. Mapping revisions must be maintained separately.

README.md

@@ -1,5 +1,5 @@
 ---
-pretty_name: FACTPROP — Entity Popularity Index
+pretty_name: FACTPROP — Factual Graph and Entity Index
 language:
   - en
 size_categories:
@@ -25,14 +25,17 @@ configs:
 
 ## Overview
 
-FACTPROP is a factual graph for studying how knowledge updates affect connected facts in large language models. This initial release provides the **entity popularity index** used by the project’s browser explorer: 100,015 graph-node records with entity labels, available Wikidata QIDs, and forward-edge object in-degree.
+FACTPROP is a factual graph for studying how knowledge updates affect connected facts in large language models. This release includes the original `final.pkl` graph checkpoint and the **entity popularity index** used by the project’s browser explorer: 100,015 graph-node records with entity labels, available Wikidata QIDs, and forward-edge object in-degree.
 
-**This is an index release, not the full factual graph.** Relation triples, natural-language QA, evidence passages, and experiment outputs are not included. The full graph package is being prepared separately.
+The graph checkpoint includes directed relations and edge attributes for questions, surface statements, and evidence. Experiment outputs and model checkpoints are not included. The dataset viewer displays the entity index; download `final.pkl` to access the complete graph.
 
 ## Contents
 
 | File | Description |
 | --- | --- |
+| `final.pkl` | Original graph checkpoint (125,274,720 bytes), unchanged |
+| `GRAPH.md` | Checkpoint structure and graph loading guide |
+| `load_graph.py` | Download, verify, and load the graph |
 | `data/entities.jsonl` | One object per graph node, suitable for the dataset viewer |
 | `browser-index.json` | Unchanged index used by the project website |
 | `metadata/graph.json` | Graph counts, thresholds, export date, and source-graph checksum |
@@ -60,6 +63,16 @@ This score measures connectivity in this graph version. It does not measure sear
 
 For example, `Apple Inc.` / `Q312` has degree **467**, while `Apple` / `Q89` has degree **4**.
 
+## Load the graph
+
+See [GRAPH.md](GRAPH.md) for the checkpoint schema. After installing `networkx` and `huggingface_hub`, run:
+
+```bash
+python load_graph.py
+```
+
+The loader verifies the original file checksum before loading it. Only load pickle files from a trusted source.
+
 ## Load the index
 
 ```python
@@ -88,7 +101,7 @@ with open(path, encoding="utf-8") as stream:
 - Some expected mappings, including the Java and Python programming languages, are absent.
 - A missing mapping does not imply low popularity or a score of zero. A known graph node may independently have a true degree of zero.
 - Multiple nodes with one QID must be reviewed separately. Do not automatically sum their degrees or take the maximum.
-- This index supports node-level lookup. It cannot reconstruct the full graph or reproduce all experiments in the paper.
+- The index supports node-level lookup; use `final.pkl` for graph traversal. The graph alone does not include every experiment input or output needed to reproduce all paper results.
 
 ## Provenance and versions
 
@@ -98,7 +111,7 @@ Entity labels, QIDs, and scores are preserved without remapping or aggregation.
 
 ## License
 
-A dataset-specific license has not yet been specified by the authors. This index release does not assign a new license to its upstream sources. Licensing and attribution for the full graph package remain under review.
+A dataset-specific license has not yet been specified by the authors. This index release does not assign a new license to its upstream sources. Licensing and upstream attribution remain under review; no new license is assigned by this upload.
 
 ## Citation
 

SHA256SUMS

@@ -1,5 +1,8 @@
-c73ffb327e412b6be4a46db449abcd81c763a883784aa79bd9c74a10b63721c2  README.md
+edb72ce528ffc4495ab97a80767eaa81a08a57d18f4c25d09f47c516b3e414f0  GRAPH.md
+9206dcac8a3fb4ebf99a8d4d1f6e35b4276dbf0c2e4d71572f259c5b53be5966  README.md
 fa7d9aec01b9c98c09cebc941265e4351693b362ffb3048d2c86585e84d7800d  browser-index.json
 2c21f2aefb9cbcc749cd813412dd90fea440050eeb8efb3356de786ebeb501e6  data/entities.jsonl
+437a434260edbb019c85575d65b4775cb2461f145e17966ee3acc8e4625ce7c8  final.pkl
+dc3b8e818d864a3966ee1c0bba19e788b2648e90100dfc84cdb1f427bae9e389  load_graph.py
 29e825c9838eb795a18672eb6a569cc2b666cf6971b5bd911655113a138e94ad  metadata/graph.json
-2af1ec77114c1dfe86fc6f55f6e2b7f491f04dc6cdf755d2eab4986fbdd5f626  metadata/release.json
+8fb3ae92e171613be24a99da45559538da97b81dba174ca70b2b1b366e634acb  metadata/release.json

final.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:437a434260edbb019c85575d65b4775cb2461f145e17966ee3acc8e4625ce7c8
+size 125274720

load_graph.py

@@ -0,0 +1,18 @@
+"""Download, verify, and load the trusted original FACTPROP checkpoint."""
+import hashlib
+import pickle
+from huggingface_hub import hf_hub_download
+
+SHA256 = "437a434260edbb019c85575d65b4775cb2461f145e17966ee3acc8e4625ce7c8"
+
+def load_graph():
+    path = hf_hub_download("factprop/FACTPROP", "final.pkl", repo_type="dataset")
+    with open(path, "rb") as stream:
+        content = stream.read()
+    if hashlib.sha256(content).hexdigest() != SHA256:
+        raise ValueError("Graph checksum mismatch")
+    return pickle.loads(content)["graph"]
+
+if __name__ == "__main__":
+    graph = load_graph()
+    print(type(graph).__name__, graph.number_of_nodes(), graph.number_of_edges())

metadata/release.json

@@ -1,10 +1,12 @@
 {
-  "release": "entity-index-v1",
+  "release": "graph-and-index-v1",
   "source_repository": "https://github.com/factprop/FACTPROP",
   "source_commit": "738ab3471f259340db6cacf404f1c0cf8063cfc3",
   "source_file": "website/data/entities.json",
   "records": 100015,
   "missing_qid_records": 33901,
   "qids_with_multiple_nodes": 5037,
-  "scope": "Entity popularity index; full graph and QA are not included."
+  "scope": "Original graph checkpoint and entity popularity index.",
+  "graph_file": "final.pkl",
+  "graph_sha256": "437a434260edbb019c85575d65b4775cb2461f145e17966ee3acc8e4625ce7c8"
 }