Update Croissant metadata to 3-task benchmark_v2 structure

etg_croissant.json

@@ -11,7 +11,7 @@
   "@type": "sc:Dataset",
   "dct:conformsTo": "http://mlcommons.org/croissant/RAI/1.0",
   "name": "HackerSignal",
-  "description": "HackerSignal is a large-scale, multi-source dataset linking hacker community discourse, exploit databases, vulnerability advisories, and fix commits through a shared CVE identifier space. It spans 64 public forum/source identifiers across eight source layers, covers 1988\u20132026, and supports five benchmark tasks: exploit relevance classification, CVE linkage retrieval, leakage-sanitized severity prediction, hacker exploit labeling, and hacker exploit signal detection.",
+  "description": "HackerSignal is a large-scale, multi-source dataset linking hacker community discourse, exploit databases, vulnerability advisories, and fix commits through a shared CVE identifier space. It spans 64 public forum/source identifiers across eight source layers, covers 1988\u20132026, and supports three benchmark tasks: CVE linkage retrieval, hacker signal detection with CVE grounding, and CVE-disjoint temporal generalization.",
   "url": "https://huggingface.co/datasets/DatasetSubmission/HackerSignal",
   "version": "1.0.0",
   "license": "https://creativecommons.org/licenses/by/4.0/",
@@ -23,11 +23,13 @@
     "zh",
     "ru"
   ],
-  "citation": "Anonymous (2026). HackerSignal \u2014 A Multi-Source Dataset for Hacker Community Analysis and CVE Linkage. Submitted to NeurIPS Evaluations & Datasets Track.",
+  "citation": "Ampel, B. (2026). HackerSignal \u2014 A Multi-Source Dataset for Hacker Community Analysis and CVE Linkage. Proceedings of NeurIPS Evaluations & Datasets Track.",
   "creator": [
     {
       "@type": "sc:Person",
-      "name": "Anonymous"
+      "name": "Benjamin Ampel",
+      "email": "ben.ampel@gmail.com",
+      "affiliation": "University of Arizona"
     }
   ],
   "keywords": [
@@ -89,7 +91,7 @@
   "prov:wasGeneratedBy": {
     "@type": "prov:Activity",
     "name": "HackerSignal ETL Pipeline",
-    "description": "Collection scripts executed by the dataset authors between January and April 2026 using public APIs, public dataset ingestion, web scraping of publicly accessible pages, and pre-processed dataset imports. The pipeline performs HTML stripping, encoding normalization, text truncation (8000 chars), exact deduplication by SHA-256 fingerprint, author pseudonymization, and source-layer normalization.",
+    "description": "Collection scripts executed by Benjamin Ampel between January and April 2026 using public APIs, public dataset ingestion, web scraping of publicly accessible pages, and pre-processed dataset imports. The pipeline performs HTML stripping, encoding normalization, text truncation (8000 chars), exact deduplication by SHA-256 fingerprint, author pseudonymization, and source-layer normalization.",
     "prov:startedAtTime": "2026-01-01",
     "prov:endedAtTime": "2026-04-30"
   },
@@ -108,7 +110,7 @@
   "rai:personalSensitiveInformation": "No private communications, medical records, financial records, or government classified information is intentionally included. All text was collected from publicly accessible sources. Author strings are pseudonymized via SHA-256 hashing of source-namespaced author identifiers; original author strings are not recoverable from the released dataset. Public forum usernames are not distributed. The release supports takedown and correction requests through the repository issue tracker.",
   "rai:dataUseCases": "Intended uses: defensive cybersecurity research, temporal retrieval, provenance-aware benchmark design, vulnerability lifecycle analysis, exploit prediction and early warning systems, cross-lingual threat intelligence, graph neural network research on exploit\u2013CVE\u2013fix commit knowledge graphs, and pre-training domain-adapted language models for cybersecurity. Prohibited uses: training systems for automated exploit code generation or malware; de-anonymizing forum participants or linking pseudonymous identities outside approved research protocols; operational blocking, law-enforcement, or employment decisions based solely on model outputs; republishing source text from metadata_or_pointer_only sources.",
   "rai:dataSocialImpact": "HackerSignal enables earlier detection of vulnerability exploitation activity by connecting hacker community discourse to the CVE lifecycle. Positive impacts include faster defensive response times and better understanding of exploit emergence patterns. Potential negative impacts include misuse for automated attack generation or targeting of individuals; these are mitigated through responsible-use restrictions, author pseudonymization, source-level release governance, and explicit prohibition of offensive applications in the dataset license.",
-  "rai:dataReleaseMaintenancePlan": "Maintained by the dataset maintainer (contact via repository issue tracker). Annual updates planned to incorporate new CVE records, CISA KEV entries, and newly published advisories. Versioning follows semantic versioning (MAJOR.MINOR.PATCH). All tagged versions remain available on HuggingFace Hub. Takedown and correction requests acknowledged within 10 business days; disputed text removed in next patch release. Sources that change terms post-release are downgraded to metadata_or_pointer_only until permissions are clarified.",
+  "rai:dataReleaseMaintenancePlan": "Maintained by Benjamin Ampel (ben.ampel@gmail.com). Annual updates planned to incorporate new CVE records, CISA KEV entries, and newly published advisories. Versioning follows semantic versioning (MAJOR.MINOR.PATCH). All tagged versions remain available on HuggingFace Hub. Takedown and correction requests acknowledged within 10 business days; disputed text removed in next patch release. Sources that change terms post-release are downgraded to metadata_or_pointer_only until permissions are clarified.",
   "rai:hasSyntheticData": "none",
   "distribution": [
     {
@@ -281,9 +283,9 @@
     },
     {
       "@type": "cr:RecordSet",
-      "@id": "task1_exploit_clf",
-      "name": "task1_exploit_clf",
-      "description": "Task 1 \u2014 Exploit Relevance Classification. Binary classification: label=1 if exploit-relevant, label=0 otherwise. Temporal split: train (pre-2022), val (2022\u20132023), test (2024+). Balanced within source/split cells to remove source-membership confounding.",
+      "@id": "task1_cve_linkage",
+      "name": "task1_cve_linkage",
+      "description": "Task 1 \u2014 CVE Linkage Retrieval (CVE-R). Cross-source temporally OOD entity grounding: given exploit/advisory evidence text, retrieve the correct NVD CVE entry from a corpus of 340,536 descriptions. Quality-controlled to remove unretrievable rows. 40,698 train / 2,551 val / 550 test.",
       "field": [
         {
           "@type": "cr:Field",
@@ -299,10 +301,10 @@
         },
         {
           "@type": "cr:Field",
-          "@id": "task1/label",
-          "name": "label",
-          "dataType": "sc:Integer",
-          "description": "0 = not exploit-relevant, 1 = exploit-relevant"
+          "@id": "task1/cve_id",
+          "name": "cve_id",
+          "dataType": "sc:Text",
+          "description": "Target CVE identifier in the corpus."
         },
         {
           "@type": "cr:Field",
@@ -327,9 +329,9 @@
     },
     {
       "@type": "cr:RecordSet",
-      "@id": "task2_cve_linkage",
-      "name": "task2_cve_linkage",
-      "description": "Task 2 \u2014 Quality-Controlled CVE Linkage Retrieval. Given source evidence text, retrieve the metadata-linked NVD CVE entry from a corpus of ~340K CVE descriptions. Quality-controlled to remove unretrievable empty/high-risk rows. Evaluated as a ranking problem (Recall@K, MRR).",
+      "@id": "task2_signal_detection",
+      "name": "task2_signal_detection",
+      "description": "Task 2 \u2014 Hacker Signal Detection (HSD). Two-stage: detect actionable exploit signals from hacker community noise (10:1 negative ratio), then ground positives to NVD CVE. 4,202 train / 1,606 val / 407 test.",
       "field": [
         {
           "@type": "cr:Field",
@@ -343,11 +345,19 @@
           "name": "text",
           "dataType": "sc:Text"
         },
+        {
+          "@type": "cr:Field",
+          "@id": "task2/actionable_label",
+          "name": "actionable_label",
+          "dataType": "sc:Integer",
+          "description": "0=not_actionable, 1=actionable_exploit_signal"
+        },
         {
           "@type": "cr:Field",
           "@id": "task2/cve_id",
           "name": "cve_id",
-          "dataType": "sc:Text"
+          "dataType": "sc:Text",
+          "description": "Primary CVE target for actionable records; empty for non-actionable."
         },
         {
           "@type": "cr:Field",
@@ -372,9 +382,9 @@
     },
     {
       "@type": "cr:RecordSet",
-      "@id": "task3_severity",
-      "name": "task3_severity",
-      "description": "Task 3 \u2014 Severity Prediction. 4-class classification: predict CVSS severity bucket (0=low, 1=medium, 2=high, 3=critical) from advisory text. Explicit severity/CVSS strings are stripped from model input after label extraction. Automated leakage audit finds zero strict sanitized hits.",
+      "@id": "task3_temporal_generalization",
+      "name": "task3_temporal_generalization",
+      "description": "Task 3 \u2014 Temporal Generalization (TG). Identical retrieval formulation to Task 1 but with strict CVE-disjoint constraint: C_train \u2229 C_test = \u2205. Split by CVE publication year (train <2022, test 2024+). 41,630 train / 1,776 val / 393 test.",
       "field": [
         {
           "@type": "cr:Field",
@@ -386,29 +396,14 @@
           "@type": "cr:Field",
           "@id": "task3/text",
           "name": "text",
-          "dataType": "sc:Text",
-          "description": "Advisory text after severity/CVSS leakage strings have been stripped."
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task3/text_raw",
-          "name": "text_raw",
-          "dataType": "sc:Text",
-          "description": "Original text before explicit severity/CVSS leakage strings were stripped."
+          "dataType": "sc:Text"
         },
         {
           "@type": "cr:Field",
-          "@id": "task3/severity",
-          "name": "severity",
+          "@id": "task3/cve_id",
+          "name": "cve_id",
           "dataType": "sc:Text",
-          "description": "Severity label string: low / medium / high / critical"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task3/severity_label",
-          "name": "severity_label",
-          "dataType": "sc:Integer",
-          "description": "Severity as integer: 0=low, 1=medium, 2=high, 3=critical"
+          "description": "Target CVE identifier (from 2024+ CVE set, disjoint from training CVEs)."
         },
         {
           "@type": "cr:Field",
@@ -430,126 +425,6 @@
           "description": "train / val / test"
         }
       ]
-    },
-    {
-      "@type": "cr:RecordSet",
-      "@id": "task4_hacker_exploit_labeling",
-      "name": "task4_hacker_exploit_labeling",
-      "description": "Task 4 \u2014 Hacker Exploit Labeling. Ternary weak-supervision task over hacker-community messages: 0=non-exploit/noise, 1=vulnerability discussion, 2=actionable exploit intelligence. Labels derived from CVE metadata and actionability cues.",
-      "field": [
-        {
-          "@type": "cr:Field",
-          "@id": "task4/id",
-          "name": "id",
-          "dataType": "sc:Text"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task4/text",
-          "name": "text",
-          "dataType": "sc:Text"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task4/label",
-          "name": "label",
-          "dataType": "sc:Integer",
-          "description": "0=non_exploit_noise, 1=vulnerability_discussion, 2=actionable_exploit"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task4/label_name",
-          "name": "label_name",
-          "dataType": "sc:Text",
-          "description": "Human-readable label: non_exploit_noise, vulnerability_discussion, or actionable_exploit"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task4/cve_ids",
-          "name": "cve_ids",
-          "dataType": "sc:Text",
-          "description": "Comma-separated CVE identifiers when applicable."
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task4/source",
-          "name": "source",
-          "dataType": "sc:Text"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task4/timestamp",
-          "name": "timestamp",
-          "dataType": "sc:DateTime"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task4/split",
-          "name": "split",
-          "dataType": "sc:Text",
-          "description": "train / val / test"
-        }
-      ]
-    },
-    {
-      "@type": "cr:RecordSet",
-      "@id": "task5_hacker_signal_detection",
-      "name": "task5_hacker_signal_detection",
-      "description": "Task 5 \u2014 Hacker Exploit Signal Detection. Joint task: detect actionable exploit intelligence in hacker-community messages and retrieve the associated NVD CVE context when applicable. Positives are Task 4 actionable records with CVE metadata.",
-      "field": [
-        {
-          "@type": "cr:Field",
-          "@id": "task5/id",
-          "name": "id",
-          "dataType": "sc:Text"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task5/text",
-          "name": "text",
-          "dataType": "sc:Text"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task5/actionable_label",
-          "name": "actionable_label",
-          "dataType": "sc:Integer",
-          "description": "0=not_actionable_exploit_signal, 1=actionable_exploit_signal"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task5/cve_id",
-          "name": "cve_id",
-          "dataType": "sc:Text",
-          "description": "Primary CVE target for actionable records; empty for non-actionable records."
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task5/cve_ids",
-          "name": "cve_ids",
-          "dataType": "sc:Text",
-          "description": "Comma-separated CVE identifiers when applicable."
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task5/source",
-          "name": "source",
-          "dataType": "sc:Text"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task5/timestamp",
-          "name": "timestamp",
-          "dataType": "sc:DateTime"
-        },
-        {
-          "@type": "cr:Field",
-          "@id": "task5/split",
-          "name": "split",
-          "dataType": "sc:Text",
-          "description": "train / val / test"
-        }
-      ]
     }
   ]
 }