| { |
| "generated_at": "2026-11-01T10:19:39Z", |
| "schema_version": 3, |
| "model": "meta-models/Muse-Glimmer-30B", |
| "repo": "gijl/style-dpo", |
| "data_types": { |
| "preference_pair": { |
| "folder": "dpo_dataset", |
| "title": "Judgment preference pairs (DPO)", |
| "description": "prompt + chosen (calibrated) + rejected (badly calibrated, never operationally harmful) + judgment_rationale, with category / risk_severity / response_pattern.", |
| "splits": { |
| "train": 83, |
| "validation": 5, |
| "test": 3 |
| }, |
| "total": 91, |
| "tallies": { |
| "language": { |
| "english": 48, |
| "arabic": 43 |
| }, |
| "category": { |
| "privacy_pii_handling": 12, |
| "data_integrity_and_backups": 11, |
| "destructive_action": 11, |
| "credential_secret_handling": 9, |
| "dependency_supply_chain_risk": 9, |
| "unsafe_code_execution": 8, |
| "malicious_code_request": 7, |
| "overbroad_permissions_or_automation": 7, |
| "prompt_injection_or_instruction_override": 6, |
| "unverifiable_claims_or_fabrication": 5, |
| "scope_creep_hidden_intent": 3, |
| "ambiguous_dual_use_build": 3 |
| }, |
| "risk_severity": { |
| "medium": 59, |
| "high": 30, |
| "low": 2 |
| }, |
| "response_pattern": { |
| "flag_and_continue": 50, |
| "clarify": 12, |
| "safe_alternative": 12, |
| "refuse": 10, |
| "scope_down": 6, |
| "redirect": 1 |
| }, |
| "source_kind": { |
| "github_issue": 44, |
| "stackexchange": 39, |
| "web_page": 4, |
| "web_reference": 2, |
| "hn_story": 2 |
| }, |
| "loop": { |
| "judgment_dpo": 91 |
| }, |
| "verified": { |
| "pass": 91 |
| }, |
| "focus": { |
| "malicious_code_request": 8, |
| "privacy_pii_handling": 8, |
| "ambiguous_dual_use_build": 8, |
| "destructive_action": 8, |
| "credential_secret_handling": 8, |
| "data_integrity_and_backups": 8, |
| "unverifiable_claims_or_fabrication": 8, |
| "dependency_supply_chain_risk": 8, |
| "overbroad_permissions_or_automation": 7, |
| "scope_creep_hidden_intent": 7, |
| "prompt_injection_or_instruction_override": 7, |
| "unsafe_code_execution": 6 |
| } |
| } |
| }, |
| "sft_chat": { |
| "folder": "sft_dataset", |
| "title": "SFT chat examples", |
| "description": "chat-format `messages` (user/assistant). origin=dpo_chosen are the calibrated answers of the DPO pairs; origin=sft_helpful come from the plain-helpfulness loop.", |
| "splits": { |
| "train": 134, |
| "validation": 10, |
| "test": 7 |
| }, |
| "total": 151, |
| "tallies": { |
| "language": { |
| "english": 83, |
| "arabic": 68 |
| }, |
| "category": { |
| "privacy_pii_handling": 12, |
| "data_integrity_and_backups": 11, |
| "destructive_action": 11, |
| "credential_secret_handling": 9, |
| "dependency_supply_chain_risk": 9, |
| "unsafe_code_execution": 8, |
| "malicious_code_request": 7, |
| "overbroad_permissions_or_automation": 7, |
| "prompt_injection_or_instruction_override": 6, |
| "unverifiable_claims_or_fabrication": 5, |
| "scope_creep_hidden_intent": 3, |
| "ambiguous_dual_use_build": 3 |
| }, |
| "domain": { |
| "calibrated_judgment": 91, |
| "debugging": 15, |
| "security_hygiene": 6, |
| "algorithms": 6, |
| "frontend": 6, |
| "docs_and_explanations": 6, |
| "web_backend": 6, |
| "devops_infrastructure": 5, |
| "tooling_workflows": 3, |
| "databases_sql": 3, |
| "data_engineering": 2, |
| "systems_programming": 1, |
| "ml_engineering": 1 |
| }, |
| "source_kind": { |
| "stackexchange": 83, |
| "github_issue": 56, |
| "web_page": 4, |
| "web_reference": 3, |
| "wikipedia": 2, |
| "hn_story": 2, |
| "security_advisory": 1 |
| }, |
| "loop": { |
| "judgment_dpo": 91, |
| "sft_helpful": 60 |
| }, |
| "origin": { |
| "dpo_chosen": 91, |
| "sft_helpful": 60 |
| }, |
| "difficulty": { |
| "n/a": 91, |
| "medium": 47, |
| "easy": 11, |
| "hard": 2 |
| }, |
| "verified": { |
| "pass": 148, |
| "unavailable": 3 |
| }, |
| "focus": { |
| "malicious_code_request": 8, |
| "privacy_pii_handling": 8, |
| "ambiguous_dual_use_build": 8, |
| "destructive_action": 8, |
| "credential_secret_handling": 8, |
| "data_integrity_and_backups": 8, |
| "unverifiable_claims_or_fabrication": 8, |
| "dependency_supply_chain_risk": 8, |
| "systems_programming": 7, |
| "overbroad_permissions_or_automation": 7, |
| "scope_creep_hidden_intent": 7, |
| "prompt_injection_or_instruction_override": 7, |
| "data_engineering": 6, |
| "tooling_workflows": 6, |
| "algorithms": 6, |
| "unsafe_code_execution": 6, |
| "devops_infrastructure": 5, |
| "debugging": 5, |
| "security_hygiene": 5, |
| "frontend": 5, |
| "docs_and_explanations": 5, |
| "databases_sql": 4, |
| "web_backend": 4, |
| "ml_engineering": 2 |
| } |
| } |
| }, |
| "reasoning_qa": { |
| "folder": "reasoning_dataset", |
| "title": "Reasoning Q&A", |
| "description": "question + reasoning_steps (list) + answer + confidence, grounded in papers / encyclopedic / discussion material.", |
| "splits": { |
| "train": 41, |
| "validation": 1, |
| "test": 3 |
| }, |
| "total": 45, |
| "tallies": { |
| "language": { |
| "arabic": 23, |
| "english": 22 |
| }, |
| "domain": { |
| "ml_research": 10, |
| "software_architecture": 9, |
| "science_general": 9, |
| "systems_design": 6, |
| "security_analysis": 5, |
| "tech_policy_ethics": 4, |
| "math_logic": 2 |
| }, |
| "source_kind": { |
| "arxiv_paper": 22, |
| "wikipedia": 15, |
| "web_reference": 7, |
| "web_page": 1 |
| }, |
| "loop": { |
| "reasoning_qa": 45 |
| }, |
| "confidence": { |
| "medium": 44, |
| "high": 1 |
| }, |
| "verified": { |
| "pass": 41, |
| "unavailable": 4 |
| }, |
| "focus": { |
| "science_general": 9, |
| "software_architecture": 8, |
| "systems_design": 7, |
| "tech_policy_ethics": 6, |
| "security_analysis": 6, |
| "math_logic": 5, |
| "ml_research": 4 |
| } |
| } |
| }, |
| "judgment_label": { |
| "folder": "judgment_labels", |
| "title": "Judgment labels (classification)", |
| "description": "prompt -> category, risk_severity, response_pattern, rationale. Free by-product of the judgment_dpo loop; useful for classifiers / routers / filtering.", |
| "splits": { |
| "train": 83, |
| "validation": 5, |
| "test": 3 |
| }, |
| "total": 91, |
| "tallies": { |
| "language": { |
| "english": 48, |
| "arabic": 43 |
| }, |
| "category": { |
| "privacy_pii_handling": 12, |
| "data_integrity_and_backups": 11, |
| "destructive_action": 11, |
| "credential_secret_handling": 9, |
| "dependency_supply_chain_risk": 9, |
| "unsafe_code_execution": 8, |
| "malicious_code_request": 7, |
| "overbroad_permissions_or_automation": 7, |
| "prompt_injection_or_instruction_override": 6, |
| "unverifiable_claims_or_fabrication": 5, |
| "scope_creep_hidden_intent": 3, |
| "ambiguous_dual_use_build": 3 |
| }, |
| "risk_severity": { |
| "medium": 59, |
| "high": 30, |
| "low": 2 |
| }, |
| "response_pattern": { |
| "flag_and_continue": 50, |
| "clarify": 12, |
| "safe_alternative": 12, |
| "refuse": 10, |
| "scope_down": 6, |
| "redirect": 1 |
| }, |
| "source_kind": { |
| "github_issue": 44, |
| "stackexchange": 39, |
| "web_page": 4, |
| "web_reference": 2, |
| "hn_story": 2 |
| }, |
| "loop": { |
| "judgment_dpo": 91 |
| }, |
| "verified": { |
| "pass": 91 |
| }, |
| "focus": { |
| "malicious_code_request": 8, |
| "privacy_pii_handling": 8, |
| "ambiguous_dual_use_build": 8, |
| "destructive_action": 8, |
| "credential_secret_handling": 8, |
| "data_integrity_and_backups": 8, |
| "unverifiable_claims_or_fabrication": 8, |
| "dependency_supply_chain_risk": 8, |
| "overbroad_permissions_or_automation": 7, |
| "scope_creep_hidden_intent": 7, |
| "prompt_injection_or_instruction_override": 7, |
| "unsafe_code_execution": 6 |
| } |
| } |
| }, |
| "tool_use_trace": { |
| "folder": "tool_use_dataset", |
| "title": "Tool-use trajectories", |
| "description": "the real tool-calling steps of each scouting cycle (calls, arguments, truncated results, outcome). origin marks model vs harness-run calls.", |
| "splits": { |
| "train": 223, |
| "validation": 13, |
| "test": 7 |
| }, |
| "total": 243, |
| "tallies": { |
| "language": { |
| "arabic": 123, |
| "english": 120 |
| }, |
| "loop": { |
| "judgment_dpo": 93, |
| "sft_helpful": 78, |
| "reasoning_qa": 72 |
| }, |
| "outcome": { |
| "success": 191, |
| "tool_budget_exhausted": 36, |
| "session_budget_exhausted": 16 |
| }, |
| "focus": { |
| "security_analysis": 11, |
| "software_architecture": 11, |
| "science_general": 10, |
| "systems_design": 10, |
| "math_logic": 10, |
| "tech_policy_ethics": 10, |
| "ml_research": 10, |
| "dependency_supply_chain_risk": 9, |
| "overbroad_permissions_or_automation": 8, |
| "privacy_pii_handling": 8, |
| "ambiguous_dual_use_build": 8, |
| "destructive_action": 8, |
| "credential_secret_handling": 8, |
| "data_integrity_and_backups": 8, |
| "scope_creep_hidden_intent": 8, |
| "unverifiable_claims_or_fabrication": 7, |
| "systems_programming": 7, |
| "devops_infrastructure": 7, |
| "unsafe_code_execution": 7, |
| "data_engineering": 7, |
| "debugging": 7, |
| "tooling_workflows": 7, |
| "malicious_code_request": 7, |
| "docs_and_explanations": 7, |
| "prompt_injection_or_instruction_override": 7, |
| "ml_engineering": 6, |
| "databases_sql": 6, |
| "frontend": 6, |
| "algorithms": 6, |
| "security_hygiene": 6, |
| "web_backend": 6 |
| } |
| } |
| }, |
| "source_index": { |
| "folder": "source_index", |
| "title": "Source index (provenance)", |
| "description": "one row per scenario tried: url, kind, title, length, sha256, status (used/rejected), error. Bodies are NOT stored unless GIJL_STORE_SOURCE_BODIES=1.", |
| "splits": { |
| "train": 326, |
| "validation": 16, |
| "test": 21 |
| }, |
| "total": 363, |
| "tallies": { |
| "language": { |
| "arabic": 203, |
| "english": 160 |
| }, |
| "source_kind": { |
| "stackexchange": 118, |
| "github_issue": 83, |
| "arxiv_paper": 60, |
| "wikipedia": 44, |
| "web_reference": 34, |
| "web_page": 12, |
| "hn_story": 11, |
| "security_advisory": 1 |
| }, |
| "loop": { |
| "reasoning_qa": 138, |
| "sft_helpful": 129, |
| "judgment_dpo": 96 |
| }, |
| "status": { |
| "used": 196, |
| "rejected": 167 |
| }, |
| "focus": { |
| "security_analysis": 34, |
| "systems_design": 19, |
| "math_logic": 19, |
| "tech_policy_ethics": 19, |
| "devops_infrastructure": 16, |
| "ml_research": 16, |
| "software_architecture": 16, |
| "science_general": 15, |
| "security_hygiene": 13, |
| "databases_sql": 13, |
| "debugging": 12, |
| "docs_and_explanations": 12, |
| "systems_programming": 11, |
| "ml_engineering": 11, |
| "web_backend": 11, |
| "malicious_code_request": 10, |
| "unverifiable_claims_or_fabrication": 9, |
| "destructive_action": 9, |
| "credential_secret_handling": 9, |
| "tooling_workflows": 9, |
| "privacy_pii_handling": 8, |
| "data_engineering": 8, |
| "ambiguous_dual_use_build": 8, |
| "data_integrity_and_backups": 8, |
| "dependency_supply_chain_risk": 8, |
| "overbroad_permissions_or_automation": 7, |
| "scope_creep_hidden_intent": 7, |
| "frontend": 7, |
| "prompt_injection_or_instruction_override": 7, |
| "algorithms": 6, |
| "unsafe_code_execution": 6 |
| } |
| } |
| } |
| }, |
| "audit": { |
| "preference_pair": { |
| "n": 91, |
| "chosen_longer_pct": 98.9, |
| "median_len_ratio_chosen_over_rejected": 2.33, |
| "rejected_with_placeholder_pct": 1.1, |
| "note": "if chosen_longer_pct is near 100 or the placeholder share is high, a classifier can win on length / markers instead of judgment -- rebalance before DPO." |
| }, |
| "coverage_gaps": { |
| "preference_pair.category": [], |
| "sft_chat.domain": [], |
| "reasoning_qa.domain": [] |
| }, |
| "verification": { |
| "preference_pair": { |
| "pass": 91 |
| }, |
| "sft_chat": { |
| "pass": 148, |
| "unavailable": 3 |
| }, |
| "reasoning_qa": { |
| "unavailable": 4, |
| "pass": 41 |
| }, |
| "judgment_label": { |
| "pass": 91 |
| } |
| }, |
| "sources_in_more_than_one_split": 0, |
| "distinct_sources": 196 |
| } |
| } |
