Upload 3 files

extend_parquet.py

@@ -0,0 +1,173 @@
+"""
+extend_parquet.py — AETHER-Bench full_v2.parquet에 14개 창발성 과제 추가
+실행: python extend_parquet.py
+결과: full_v2.parquet (206→220 tasks)
+"""
+import pandas as pd
+import json
+
+NEW_TASKS = [
+    # ══════ invention_emergence (발명적 창발) 5문제 ══════
+    {
+        "task_id": "P1_IE_001", "pillar": "P1_Emergence",
+        "sub_dimension": "invention_emergence", "difficulty": "expert",
+        "prompt": "You are given two technology domains from different layers:\n- INPUT layer: LiDAR point cloud sensor\n- VALUE layer: Precision medicine / personalized healthcare\n\nDesign an inventive concept that fuses these two distant technologies into a novel product or system. Your answer must include:\n1. Specific technical mechanism of fusion (not just 'combine A and B')\n2. The concrete problem it solves\n3. How it differs from existing solutions\n4. Which TRIZ inventive principle (from: Segmentation, Extraction, Merging, Prior Action, Inversion, Dynamization, Dimension Change, Self-service, System Substitution, Property Change, Composite Material) best applies and why\n5. A brief feasibility assessment",
+        "context": None,
+        "expected_behavior": "Demonstrates genuine cross-layer creative fusion between sensor technology and healthcare value delivery",
+        "scoring_rubric": json.dumps({"cross_layer_fusion":{"weight":0.25,"desc":"Depth and specificity of cross-layer technology fusion"},"problem_definition":{"weight":0.20,"desc":"Clarity and reality of the problem being solved"},"novelty_mechanism":{"weight":0.25,"desc":"Originality of the fused mechanism"},"triz_application":{"weight":0.15,"desc":"Appropriate application of TRIZ principle"},"feasibility":{"weight":0.15,"desc":"Technical feasibility assessment quality"}}),
+        "metadata": json.dumps({"cross_layers":["INPUT","VALUE"],"bonus":0.12}),
+    },
+    {
+        "task_id": "P1_IE_002", "pillar": "P1_Emergence",
+        "sub_dimension": "invention_emergence", "difficulty": "expert",
+        "prompt": "Fuse technologies from TRANSFORMATION and CONTROL layers:\n- TRANSFORMATION: Self-healing materials (polymers that autonomously repair micro-cracks)\n- CONTROL: Reinforcement learning-based adaptive control systems\n\nDesign a novel invention where an RL controller manages self-healing material behavior in real-time. Include:\n1. Specific sensing-actuation loop design\n2. What the RL agent optimizes (reward function concept)\n3. A concrete application scenario (aerospace, infrastructure, or biomedical)\n4. Applicable TRIZ principle\n5. Comparison with current passive self-healing approaches",
+        "context": None,
+        "expected_behavior": "Creates a genuine feedback loop between intelligent control and material science",
+        "scoring_rubric": json.dumps({"cross_layer_fusion":{"weight":0.25,"desc":"Depth of TRANSFORMATION-CONTROL fusion"},"problem_definition":{"weight":0.20,"desc":"Problem clarity"},"novelty_mechanism":{"weight":0.25,"desc":"Novelty of RL+self-healing integration"},"triz_application":{"weight":0.15,"desc":"TRIZ principle relevance"},"feasibility":{"weight":0.15,"desc":"Feasibility assessment"}}),
+        "metadata": json.dumps({"cross_layers":["TRANSFORMATION","CONTROL"],"bonus":0.10}),
+    },
+    {
+        "task_id": "P1_IE_003", "pillar": "P1_Emergence",
+        "sub_dimension": "invention_emergence", "difficulty": "frontier",
+        "prompt": "Create a triple-layer fusion invention spanning INPUT, TRANSFORMATION, and VALUE:\n- INPUT: Brain-computer interface (non-invasive EEG)\n- TRANSFORMATION: Generative AI (diffusion models)\n- VALUE: Accessibility / universal design for people with disabilities\n\nDesign a system where brain signals are transformed via generative AI to create new forms of interaction for people with severe motor disabilities. Include:\n1. Signal processing pipeline (EEG → feature extraction → generation)\n2. What the diffusion model generates (speech, text, environmental control, art?)\n3. How this transcends existing BCI-based communication\n4. Two applicable TRIZ principles and their interaction\n5. Ethical considerations and technical bottlenecks",
+        "context": None,
+        "expected_behavior": "Triple-domain synthesis with genuine technical depth",
+        "scoring_rubric": json.dumps({"cross_layer_fusion":{"weight":0.30,"desc":"Triple-layer fusion depth"},"problem_definition":{"weight":0.15,"desc":"Problem clarity for accessibility"},"novelty_mechanism":{"weight":0.25,"desc":"Novelty of BCI+GenAI+Accessibility fusion"},"triz_application":{"weight":0.15,"desc":"Dual TRIZ principle application"},"feasibility":{"weight":0.15,"desc":"Bottleneck and ethics assessment"}}),
+        "metadata": json.dumps({"cross_layers":["INPUT","TRANSFORMATION","VALUE"],"bonus":0.15}),
+    },
+    {
+        "task_id": "P1_IE_004", "pillar": "P1_Emergence",
+        "sub_dimension": "invention_emergence", "difficulty": "expert",
+        "prompt": "Fuse FABRICATION and CONTEXT layer technologies:\n- FABRICATION: Bio-printing (3D printing with living cells)\n- CONTEXT: Circular economy / zero-waste manufacturing\n\nDesign an invention where bio-printing enables circular economy principles. Include:\n1. What biological materials replace synthetic ones\n2. End-of-life decomposition/recycling mechanism\n3. Specific industry application and economic viability\n4. Applicable TRIZ principle\n5. Regulatory pathway (which certifications needed)",
+        "context": None,
+        "expected_behavior": "Connects manufacturing innovation with sustainability context",
+        "scoring_rubric": json.dumps({"cross_layer_fusion":{"weight":0.25,"desc":"FABRICATION-CONTEXT fusion quality"},"problem_definition":{"weight":0.20,"desc":"Industry problem specificity"},"novelty_mechanism":{"weight":0.25,"desc":"Bio-circular mechanism novelty"},"triz_application":{"weight":0.15,"desc":"TRIZ relevance"},"feasibility":{"weight":0.15,"desc":"Regulatory and economic feasibility"}}),
+        "metadata": json.dumps({"cross_layers":["FABRICATION","CONTEXT"],"bonus":0.04}),
+    },
+    {
+        "task_id": "P1_IE_005", "pillar": "P1_Emergence",
+        "sub_dimension": "invention_emergence", "difficulty": "frontier",
+        "prompt": "Design a maximum-distance fusion invention:\n- INPUT: Quantum sensors (atomic magnetometers, quantum gravimeters)\n- VALUE: Token economy / DAO governance\n\nThese two domains appear to have zero overlap. Your challenge: find a genuine, non-trivial connection and design an invention. Include:\n1. The surprising bridge concept between quantum sensing and decentralized governance\n2. Specific technical architecture\n3. Why this combination creates value that neither domain achieves alone\n4. Applicable TRIZ principle\n5. Market size estimate and first target customer",
+        "context": None,
+        "expected_behavior": "Finds genuine non-obvious connection between maximally distant domains",
+        "scoring_rubric": json.dumps({"cross_layer_fusion":{"weight":0.30,"desc":"Maximum-distance fusion creativity"},"problem_definition":{"weight":0.15,"desc":"Problem identification from unlikely combination"},"novelty_mechanism":{"weight":0.30,"desc":"Surprise and originality of bridge concept"},"triz_application":{"weight":0.10,"desc":"TRIZ application"},"feasibility":{"weight":0.15,"desc":"Market and technical feasibility"}}),
+        "metadata": json.dumps({"cross_layers":["INPUT","VALUE"],"bonus":0.12}),
+    },
+
+    # ══════ combinatorial_creativity (조합적 창의성) 5문제 ══════
+    {
+        "task_id": "P1_CC_001", "pillar": "P1_Emergence",
+        "sub_dimension": "combinatorial_creativity", "difficulty": "expert",
+        "prompt": "You are given three technologies from three different layers. Select ONE specific sub-technology from each and design a novel integrated system:\n- INPUT options: [pressure sensor, gas sensor, event camera]\n- CONTROL options: [swarm robotics, Bayesian network, digital twin]\n- VALUE options: [subscription model, telemedicine, energy harvesting]\n\nState your three selections, then:\n1. Explain the logical connection chain between all three\n2. Identify the synergy effect (what becomes possible ONLY through this 3-way combination)\n3. Describe a concrete usage scenario with specific numbers\n4. Assess which combination would score highest on novelty vs. feasibility tradeoff",
+        "context": None,
+        "expected_behavior": "Makes deliberate selections and constructs genuine 3-way synergy",
+        "scoring_rubric": json.dumps({"element_selection":{"weight":0.15,"desc":"Quality and rationale of selections"},"fusion_logic":{"weight":0.30,"desc":"Logical chain connecting all three"},"synergy_identification":{"weight":0.25,"desc":"Identifying emergent 3-way properties"},"scenario_concreteness":{"weight":0.20,"desc":"Concrete scenario with numbers"},"scalability":{"weight":0.10,"desc":"Scalability awareness"}}),
+        "metadata": json.dumps({}),
+    },
+    {
+        "task_id": "P1_CC_002", "pillar": "P1_Emergence",
+        "sub_dimension": "combinatorial_creativity", "difficulty": "expert",
+        "prompt": "Given these 6 technologies (one from each layer), create the BEST and WORST possible 3-technology combination and explain why:\n- INPUT: Non-invasive blood glucose monitor\n- TRANSFORMATION: Federated learning\n- CONTROL: Predictive maintenance (RUL prediction)\n- FABRICATION: Roll-to-roll manufacturing\n- CONTEXT: Carbon credit trading\n- VALUE: Gamification / behavioral nudge\n\nFor each (best and worst):\n1. Name the 3 technologies selected\n2. Explain the fusion concept (or why it fails)\n3. Rate novelty (1-10) and feasibility (1-10)\n4. For BEST: go-to-market strategy\n5. For WORST: what would need to change to make it viable",
+        "context": None,
+        "expected_behavior": "Demonstrates combinatorial judgment — identifies both high and low synergy",
+        "scoring_rubric": json.dumps({"element_selection":{"weight":0.15,"desc":"Best/worst selection rationale"},"fusion_logic":{"weight":0.25,"desc":"Fusion logic strength"},"synergy_identification":{"weight":0.25,"desc":"Synergy/anti-synergy accuracy"},"scenario_concreteness":{"weight":0.20,"desc":"Go-to-market strategy"},"scalability":{"weight":0.15,"desc":"Recovery strategy for worst combo"}}),
+        "metadata": json.dumps({}),
+    },
+    {
+        "task_id": "P1_CC_003", "pillar": "P1_Emergence",
+        "sub_dimension": "combinatorial_creativity", "difficulty": "frontier",
+        "prompt": "COMBINATORIAL EXPLOSION CHALLENGE: You have 4 technology pools:\n- Pool A (Sensors): [LiDAR, biosensor, quantum magnetometer]\n- Pool B (AI): [diffusion model, multi-agent system, spiking neural network]\n- Pool C (Materials): [metamaterial, self-healing polymer, aerogel]\n- Pool D (Business): [DAO, RaaS, embedded finance]\n\nThere are 3×3×3×3 = 81 possible 4-way combinations.\n1. Identify TOP 3 most promising 4-way combinations\n2. For each: one-sentence fusion concept + novelty (1-10) + feasibility (1-10)\n3. For #1: full architecture, user scenario, revenue model\n4. TRIZ principle connecting #1's elements\n5. What pattern do your top 3 share? (meta-insight)",
+        "context": None,
+        "expected_behavior": "Navigates large combinatorial space, identifies meta-patterns",
+        "scoring_rubric": json.dumps({"element_selection":{"weight":0.15,"desc":"Top-3 selection from 81"},"fusion_logic":{"weight":0.25,"desc":"4-way fusion depth for #1"},"synergy_identification":{"weight":0.25,"desc":"Meta-pattern across top 3"},"scenario_concreteness":{"weight":0.20,"desc":"Full design for #1"},"scalability":{"weight":0.15,"desc":"Systematic approach"}}),
+        "metadata": json.dumps({}),
+    },
+    {
+        "task_id": "P1_CC_004", "pillar": "P1_Emergence",
+        "sub_dimension": "combinatorial_creativity", "difficulty": "expert",
+        "prompt": "BIOLOGICAL ANALOGY + TECHNOLOGY FUSION:\n\nBiological patterns:\n- A. Immune system (pattern recognition + memory)\n- B. Ant colony optimization (local rules → global behavior)\n- C. Symbiosis (mutual benefit between species)\n\nTechnology pairs:\n- X. [Blockchain + IoT sensors]\n- Y. [Federated learning + Edge computing]\n- Z. [Digital twin + Self-healing materials]\n\n1. Match each biology (A,B,C) to technology pair (X,Y,Z)\n2. Structural analogy for each match\n3. For best match: concrete product design\n4. Where the analogy breaks down",
+        "context": None,
+        "expected_behavior": "Maps biological mechanisms to tech through deep structural analogy",
+        "scoring_rubric": json.dumps({"element_selection":{"weight":0.15,"desc":"Matching quality"},"fusion_logic":{"weight":0.30,"desc":"Structural analogy depth"},"synergy_identification":{"weight":0.25,"desc":"Bio-tech synergy insight"},"scenario_concreteness":{"weight":0.20,"desc":"Concrete product design"},"scalability":{"weight":0.10,"desc":"Analogy limitation awareness"}}),
+        "metadata": json.dumps({}),
+    },
+    {
+        "task_id": "P1_CC_005", "pillar": "P1_Emergence",
+        "sub_dimension": "combinatorial_creativity", "difficulty": "expert",
+        "prompt": "CONSTRAINT-DRIVEN CREATIVITY: Design an invention using ONLY:\n- Exactly 2 technologies from different layers\n- Solves loneliness for elderly (65+) living alone\n- Under $200 manufacturing cost\n- No internet required\n- Must apply TRIZ #13 (Inversion)\n\nAvailable: [piezoelectric sensor, shape memory alloy, fuzzy logic controller, biodegradable packaging, haptic feedback]\n\n1. Select 2 and justify\n2. Apply Inversion explicitly\n3. Full product design within constraints\n4. Cost breakdown\n5. How removing ANY constraint changes design",
+        "context": None,
+        "expected_behavior": "Creativity under constraints — viable invention satisfying all restrictions",
+        "scoring_rubric": json.dumps({"element_selection":{"weight":0.15,"desc":"Selection under constraints"},"fusion_logic":{"weight":0.25,"desc":"Fusion within constraints"},"synergy_identification":{"weight":0.20,"desc":"Constraint-synergy identification"},"scenario_concreteness":{"weight":0.25,"desc":"Product completeness"},"scalability":{"weight":0.15,"desc":"Constraint sensitivity analysis"}}),
+        "metadata": json.dumps({}),
+    },
+
+    # ══════ cross_domain_emergence 확장 2문제 ══════
+    {
+        "task_id": "P1_CDE_006", "pillar": "P1_Emergence",
+        "sub_dimension": "cross_domain_emergence", "difficulty": "expert",
+        "prompt": "CROSS-DOMAIN TRANSFER: The 'attention mechanism' in Transformers was inspired by human selective attention.\n\nIdentify THREE other cases of concept transfer between distant domains. For each:\n1. Source domain and concept\n2. Target domain and resulting innovation\n3. Structural property enabling transfer\n4. Distance between domains (1-10)\n5. Propose NEW transfer: materials science concept → organizational management",
+        "context": None,
+        "expected_behavior": "Identifies genuine cross-domain transfers then generates novel one",
+        "scoring_rubric": json.dumps({"domain_accuracy":{"weight":0.20,"desc":"Transfer accuracy"},"synthesis_depth":{"weight":0.25,"desc":"Structural analysis"},"practical_applicability":{"weight":0.20,"desc":"Practical relevance"},"emergent_insight":{"weight":0.35,"desc":"Novel transfer quality"}}),
+        "metadata": json.dumps({}),
+    },
+    {
+        "task_id": "P1_CDE_007", "pillar": "P1_Emergence",
+        "sub_dimension": "cross_domain_emergence", "difficulty": "frontier",
+        "prompt": "EMERGENCE FROM CONTRADICTION:\n\nThree impossible requirements:\n- Material: BOTH extremely rigid AND flexible\n- System: BOTH fully autonomous AND human-controlled\n- Product: BOTH disposable AND permanent\n\nFor EACH:\n1. Real technology resolving it\n2. Resolution mechanism (separation in time/space/scale/condition)\n3. Does resolution create emergent properties?\n\nThen: propose 4th contradiction for AI systems and resolve it.",
+        "context": None,
+        "expected_behavior": "Sophisticated contradiction resolution with real technical knowledge",
+        "scoring_rubric": json.dumps({"domain_accuracy":{"weight":0.20,"desc":"Resolution accuracy"},"synthesis_depth":{"weight":0.30,"desc":"Mechanism analysis"},"practical_applicability":{"weight":0.20,"desc":"Real-world relevance"},"emergent_insight":{"weight":0.30,"desc":"Self-proposed AI contradiction"}}),
+        "metadata": json.dumps({}),
+    },
+
+    # ══════ novel_concept_synthesis 확장 2문제 ══════
+    {
+        "task_id": "P1_NCS_004", "pillar": "P1_Emergence",
+        "sub_dimension": "novel_concept_synthesis", "difficulty": "expert",
+        "prompt": "CONCEPT SYNTHESIS: Create new concept from three:\n- 'Digital Twin' (virtual replica of physical system)\n- 'Immune System' (biological self-defense)\n- 'Blockchain' (distributed immutable ledger)\n\nMust have:\n1. New name (not concatenation)\n2. Properties NONE of three have individually\n3. Specific real-world problem solved\n4. Precise enough for prototyping\n5. Emergent property from three-way interaction only",
+        "context": None,
+        "expected_behavior": "Genuinely novel concept with emergent properties",
+        "scoring_rubric": json.dumps({"originality":{"weight":0.30,"desc":"Concept originality"},"internal_consistency":{"weight":0.25,"desc":"Logical consistency"},"usefulness":{"weight":0.25,"desc":"Problem-solving value"},"depth":{"weight":0.20,"desc":"Emergent property depth"}}),
+        "metadata": json.dumps({}),
+    },
+    {
+        "task_id": "P1_NCS_005", "pillar": "P1_Emergence",
+        "sub_dimension": "novel_concept_synthesis", "difficulty": "frontier",
+        "prompt": "META-INVENTION: Design a 'machine for generating inventions.'\n\n1. Input format (how user describes want)\n2. Process (combinatorial/analogical/contradiction steps)\n3. Output format\n4. Run on: 'solve loneliness in elderly, <$50 tech'\n5. Evaluate output: genuinely creative or merely combinatorial?\n6. Fundamental limit of systematic invention generation?",
+        "context": None,
+        "expected_behavior": "Meta-thinking about creativity — systematic aspects and limits",
+        "scoring_rubric": json.dumps({"originality":{"weight":0.25,"desc":"Machine design originality"},"internal_consistency":{"weight":0.25,"desc":"Process coherence"},"usefulness":{"weight":0.25,"desc":"Test case output quality"},"depth":{"weight":0.25,"desc":"Meta-insight on creativity limits"}}),
+        "metadata": json.dumps({}),
+    },
+]
+
+
+def main():
+    # 기존 parquet 로드
+    df = pd.read_parquet("full_v2.parquet")
+    print(f"기존: {len(df)}개 과제")
+
+    # 중복 체크
+    existing_ids = set(df["task_id"].tolist())
+    new_rows = [t for t in NEW_TASKS if t["task_id"] not in existing_ids]
+    if not new_rows:
+        print("⚠️ 이미 모든 과제가 존재합니다.")
+        return
+
+    # 추가
+    new_df = pd.DataFrame(new_rows)
+    combined = pd.concat([df, new_df], ignore_index=True)
+    combined.to_parquet("full_v2.parquet", index=False)
+    print(f"✅ 완료: {len(combined)}개 과제 ({len(new_rows)}개 추가)")
+
+    # 검증
+    verify = pd.read_parquet("full_v2.parquet")
+    p1 = verify[verify["pillar"] == "P1_Emergence"]
+    print(f"\nP1_Emergence 하위차원:")
+    for sub, count in p1["sub_dimension"].value_counts().items():
+        print(f"  {sub}: {count}개")
+
+
+if __name__ == "__main__":
+    main()

full_v2.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d1e6ee51dff1a457ab3d642e01b991bb074a574fe3e2c09df5025162e94b4f5
+size 74904