--- license: cc-by-nc-4.0 language: - en tags: - synthetic - finance - options - derivatives - knowledge-base - instruction-tuning - rag - adversarial-robustness pretty_name: "Options Strategies Knowledge Base (FIN-QA-002) — Sample" size_categories: - 1K **Positioning note.** Answers are *templated, structurally-controlled prose* > rendered from a concept/strategy ontology — not human-verified factual ground > truth. This corpus is built for **structural / retrieval / reranker / > adversarial-robustness / agent-evaluation** work, not for teaching factual > options knowledge via supervised fine-tuning. See **Limitations**. Each item is > **educational** and **not investment advice**. ## Depth tiers, not a gold/distractor scheme Every question in `multi_depth_answers` has **exactly four answers**, one per audience depth: `beginner`, `intermediate`, `advanced`, `institutional`. All four are legitimate renderings at different sophistication levels; there is intentionally **no single "correct" answer**. The structure supports depth-conditioned generation and depth-ranking tasks. ## Calibration anchors | Metric | Observed (seed 42) | Target | Anchor | |---|---|---|---| | Difficulty mean (1–5) | 3.06 | 2.85–3.25 | Bloom's Taxonomy difficulty centering | | Mid-difficulty (level 3) share | 0.31 | 0.26–0.42 | Bloom's mid-level concentration | | Question-type spread | 5 types, even | exactly 5 even | OCC question-type taxonomy | | Persona spread | 4 personas, even | exactly 4 even | retail/trader/quant/PM breadth | | Reasoning-required answer share | 0.80 | 0.78–0.82 | FinQA reasoning fraction | | Adversarial attack-type spread | 4 types, even | exactly 4 even | OWASP LLM Top-10 | | Formula-coverage share | 0.45 | 0.30–0.60 | options-formula coverage | Heavily-weighted **structural integrity floors** (all exact, all pass): exactly 4 distinct depth answers per question; full referential integrity across all FKs; exactly one misconception per concept (all concepts covered); exactly 5 Greeks scenarios per strategy; per-table column-count contract; no relation self-loops; complete adversarial behavior coverage. ## Tables (schema highlights) | Table | Rows (sample) | Key columns | |---|---|---| | `finqa002_qa_pairs` | 500 | `qa_id`, `concept_id` (concept **or** strategy id), `question_text`, `question_type`, `persona_type`, `difficulty` | | `finqa002_multi_depth_answers` | 2,000 | `answer_id`, `qa_id`, `depth_level`, `answer_text`, `contains_formula_flag`, `requires_reasoning_flag` | | `finqa002_misconceptions` | 120 | `misconception_id`, `concept_id`, `incorrect_statement`, `why_wrong`, `correct_explanation`, `error_type` | | `finqa002_adversarial_queries` | 120 | `adv_id`, `qa_id`, `attack_type`, `adversarial_question`, `expected_behavior` | | `finqa002_conversations` | 80 | `conv_id`, `persona_type`, `turn_sequence` (JSON), `topic_drift_flag`, `resolution_flag` | | `finqa002_options_concepts` | 120 | `concept_id`, `concept_name`, `category_l1`, `category_l2`, `difficulty_level`, `institutional_relevance_score`, `description_short`, `description_long` | | `finqa002_options_strategies` | 30 | `strategy_id`, `strategy_name`, `legs_description`, `market_outlook`, `max_profit`, `max_loss`, `breakeven_points`, `risk_profile`, `margin_requirement_estimate` | | `finqa002_ontology` | 117 | `node_id`, `node_type`, `name`, `parent_node`, `depth_level` | | `finqa002_relations` | 75 | `relation_id`, `source_node`, `target_node`, `relation_type`, `strength_score` | | `finqa002_payoff_formulas` | 30 | `formula_id`, `strategy_id`, `formula_latex`, `variable_definitions`, `interpretation` | | `finqa002_greeks_scenarios` | 150 | `scenario_id`, `strategy_id`, `underlying_move_pct`, `iv_change_pct`, `time_decay_days`, `delta_impact`, `gamma_impact`, `theta_impact`, `vega_impact`, `pnl_estimate` | | `finqa002_examples` | 30 | `example_id`, `strategy_id`, `example_type`, `example_description`, `solution_steps` | `qa_pairs.concept_id` references **either** a concept (`CON_*`) **or** a strategy (`STR_*`) — the QA layer is built over a merged entity set. `relations.source_node` / `target_node` reference **ontology `node_id`** values (not concept ids). `conversations.turn_sequence` is JSON-encoded. ## Loading ```python import pandas as pd qa = pd.read_csv("finqa002_qa_pairs.csv") answers = pd.read_csv("finqa002_multi_depth_answers.csv") merged = qa.merge(answers, on="qa_id") print(merged.groupby("qa_id").size().value_counts()) # all == 4 ``` ```python from datasets import load_dataset qa = load_dataset("xpertsystems/fin-qa-002-sample", "qa_pairs")["train"] greeks = load_dataset("xpertsystems/fin-qa-002-sample", "greeks_scenarios")["train"] ``` ## Use cases - **SFT (style/format/depth):** depth-conditioned options-answer generation (retail-plain vs institutional-desk voice). - **Preference / ranking data:** depth-preference pairs encoding *audience fit* (not factual correctness) for reranker / RLHF-style signals. - **RAG & reranker evaluation:** topic-calibrated `(query, answer-shape)` pairs over an 11-domain options ontology for MRR/NDCG-style metrics. - **Adversarial robustness:** `adversarial_queries` provides mispricing, volatility-trap, oversimplification, and hallucination-bait probes with `expected_behavior` labels (clarify / correct / reject). - **Scenario reasoning:** `greeks_scenarios` gives a deterministic underlying-move × IV-change × time-decay P&L grid for sensitivity tasks. ## Limitations (full disclosure) The build process inspected the engine line-by-line. Disclosed observations: 1. **Answers are templated prose, not verified facts.** Answer text is rendered from concept/strategy metadata and depth templates; it is plausible and structurally complete but **not** human-verified options truth. Do not use `(question, answer)` pairs as factual SFT ground truth. 2. **No gold/preferred tier.** This is a 4-level depth corpus by design. 3. **Question-type and persona mixes are deterministic, not random.** They are produced by index/modulo cycling, so they are exactly uniform by construction (5 question types, 4 personas, 4 adversarial attack types). Only `difficulty` is sampled (via an isolated `random.Random(seed)`). 4. **Misconceptions are one-per-concept, drawn from 5 templates.** Each concept gets exactly one misconception cycled from a 5-entry template pool, anchored to that concept's name. At full scale (800 concepts) the same 5 templates recur; treat misconceptions as concept-tagged exemplars, not 800 unique items. 5. **`greeks_scenarios` is a closed-form heuristic grid**, not a Black-Scholes pricer. Delta/gamma/theta/vega impacts and `pnl_estimate` are deterministic linear approximations for scenario-shape training, not accurate option P&L. 6. **Strategy variants are parameterized labels.** Beyond the 18 base strategies, additional rows are suffix-labeled variants ("for Earnings", "with Wider Wings", …) sharing the base payoff metadata. 7. **`relations` reference ontology node ids**, and the relation set is partly deterministic (curated pairs) plus a small per-strategy random edge; FK integrity to `ontology.node_id` is verified. 8. **Manifest embeds the run's `output_dir` path** and a `seed`, but **no wall-clock timestamp** — so the manifest is reproducible up to the output path. Data files are byte-identical per seed. No benchmark-theater was found: no hardcoded validation values, no `max(actual, target)` floors, no always-true passes, no referential-integrity leaks. Scorecard ranges were calibrated to observed 6-seed behavior; deterministic distributions are scored as exact-target floors and the heavy weight sits on structural integrity. ## Sample vs. full product | Dimension | Sample (this repo) | Full product | |---|---|---| | Questions | 500 | 20,000 | | Concepts | 120 | 800 | | Strategies | 30 | 400 | | Multi-depth answers | 2,000 | 80,000 | | Misconceptions | 120 | 800 | | Adversarial queries | 120 | 3,000 | | Greeks scenarios | 150 | 2,000 | | Conversations | 80 | 3,000 | | License | CC-BY-NC-4.0 | Commercial | | Validation | 6/6 seeds Grade A+ (10.0/10) | Full-scale QA suite | ## Determinism Re-running the engine with the same seed produces **byte-identical data files** (verified across all 12 CSVs) and **identical scored metrics**. The manifest is reproducible up to the embedded output-directory path; there is no wall-clock timestamp. ## Citation ```bibtex @misc{xpertsystems_finqa002_2026, title = {Options Strategies Knowledge Base (FIN-QA-002): A Synthetic, Ontology-Driven Multi-Depth Options & Derivatives Q&A Corpus}, author = {XpertSystems.ai}, year = {2026}, howpublished = {Hugging Face Datasets}, note = {Sample (500 questions) of a 20,000-question commercial product. Difficulty mix calibrated to Bloom's Taxonomy; strategy and question-type taxonomy to OCC options curricula; reasoning-required fraction to FinQA-style financial-reasoning corpora; adversarial attack-type taxonomy to the OWASP LLM Top-10. License CC-BY-NC-4.0.}, url = {https://xpertsystems.ai} } ``` Anchored benchmarks referenced for calibration: Bloom's Taxonomy of educational objectives; OCC / standard options-strategy curricula (strategy & question-type taxonomy); FinQA (Chen et al., financial numerical-reasoning QA); OWASP Top-10 for LLM Applications (adversarial attack taxonomy). ## Disclaimer This dataset is **synthetic** and provided for AI/ML research and engineering. Its content — including option strategies, payoff descriptions, and Greeks scenarios — is **educational and illustrative only**, is **not investment advice**, is not a recommendation to trade any option or security, and is not a substitute for professional financial, legal, or compliance guidance. Options trading involves substantial risk of loss.