{ "version": "0.2.0", "updated": "2026-09-24", "axes": { "ability": [ "perception", "inference", "design" ], "domains": [ "quantum", "chemistry", "materials", "biology" ], "scale_cells_nm": [ "0.1-1", "1-10", "10-100", "100-1000" ], "status": { "implemented": "数值模板 + 四选一均已接通并有真实试跑", "pilot": "新族函数已实现，2–4 份真实输入试跑，待人工审核", "existing_legacy": "旧题与计算器存在，尚未迁入统一族接口", "rebuild_needed": "旧题几何非必要或需重建，暂不扩量", "planned": "已定义路线，尚无实现", "blocked": "缺评价器或证据，暂不开放" } }, "summary": { "types_defined": 54, "types_with_running_code": 48, "status": { "implemented": 2, "pilot": 46, "existing_legacy": 3, "planned": 1, "rebuild_needed": 1, "blocked": 1 }, "by_ability": { "perception": { "implemented": 2, "pilot": 12, "existing_legacy": 2 }, "inference": { "pilot": 19, "planned": 1, "existing_legacy": 1, "rebuild_needed": 1 }, "design": { "pilot": 15, "blocked": 1 } }, "development_instances": 316, "instance_failures": 1, "independent_pass": 316, "independent_checked": 316, "legacy_candidates": 80, "legacy_decisions": { "retain": 42, "revise": 32, "rebuild": 6 }, "note": "“支持的题型”与“已生成实例”分开计数；全部实例待人工审核，未计入题库配额。" }, "family_run": "family-pilot-v37", "rows": [ { "id": "global_extent", "name": "全局最大间距", "ability": "perception", "status": "implemented", "priority": 9, "concept": "有限点集的整体尺度：所有点对中的最大欧氏距离。", "domains": [ "quantum", "materials", "chemistry", "biology" ], "input_scale": "XYZ 点集的最大间距（即答案本身）", "reasoning_scale": "同输入尺度", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "只适用于有限点集；周期晶胞内的值依赖表示方式。", "inputs": "一个已确认 Å 单位的 XYZ，明确点集与结构选择范围。", "physical_conditions": "无；纯几何。", "output_forms": [ "numeric", "choice" ], "source_basis": "真实结构文件（QM7-X/OE62、COD、PDB）", "answer_method": "逐对距离取最大；Decimal 独立循环复核。", "distractor_mechanisms": [ "heavy_atoms_only", "bounding_box_diagonal", "centroid_diameter", "two_largest_radii", "with_vdw_radii", "largest_axis_span", "radius_as_extent" ], "four_choice_validation": "extent_choice_v2：正确值秩按批次平衡，数值按升序排列；与正确值间隔 > 0.05 Å。 v0.3.0：radius_as_extent 迷惑程度调为 2（混淆两种尺寸量），否则答案永远不能在 D（探针 runs/enum-v20）。", "legacy_families": [ "global_extent", "periodic_global_extent", "protein_global_extent" ], "legacy_qa": [ "QNP007", "QNP008", "QNP009", "MNP006", "MNP007", "MNP010", "CNP007", "CNP008", "CNP009", "BNP001" ], "code": [ "template_engine.py", "choice_engine.py", "task_families/extent_choice.py", "checkers/families.py" ], "tests": [ "tests/test_template_engine.py", "tests/test_choice_engine.py", "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "专用试跑覆盖 M0/M1/M4/M5/M6 的等价检查；未走旧 pipeline。", "gap": "已足量；本轮不再扩量，只改进干扰项。", "engine_families": [ "extent_choice_v2" ], "independent_checker": "checkers/families.py:extent", "instance_ids": [ "FC-QNP007", "FC-QNP008", "FC-QNP009", "FC-MNP006", "FC-MNP007", "FC-MNP010", "FC-CNP007", "FC-CNP008", "FC-CNP009", "FC-BNP001-1CRN" ], "instance_count": 10, "independent_pass": 10, "instance_failures": 0, "instance_reasoning_cells": [ "1-10" ], "legacy_decisions": { "retain": 10 }, "legacy_choice_v01": 9 }, { "id": "equal_weight_rg", "name": "等权回转半径", "ability": "perception", "status": "implemented", "priority": 9, "concept": "点集相对质心的均方根距离，衡量整体紧凑度。", "domains": [ "quantum", "chemistry", "biology" ], "input_scale": "XYZ 点集最大间距", "reasoning_scale": "全局点集", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "不是质量加权或实测 SAXS Rg。", "inputs": "一个已确认 Å 单位的 XYZ；明确所有点等权。", "physical_conditions": "等权；不是质量加权或 SAXS Rg。", "output_forms": [ "numeric", "choice" ], "source_basis": "真实结构文件", "answer_method": "质心公式；成对距离恒等式 Decimal 复核。", "distractor_mechanisms": [ "mass_weighted", "heavy_atoms_only", "mean_radius", "n_minus_one", "maximum_radius", "diameter_for_radius" ], "four_choice_validation": "extent_choice_v2：正确值秩按批次平衡，数值按升序排列；与正确值间隔 > 0.05 Å（与 global_extent 相同）。 v0.3.0：点对公式漏 1/2（约 √2·Rg）与“最大中心距当半径”迷惑程度调为 2，否则答案不能在 A/B。", "legacy_families": [ "radius_of_gyration", "protein_geometric_gyration" ], "legacy_qa": [ "QNP010", "QNP011", "CNP011", "BNP002" ], "code": [ "template_engine.py", "choice_engine.py", "task_families/extent_choice.py", "checkers/families.py" ], "tests": [ "tests/test_template_engine.py", "tests/test_choice_engine.py", "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "专用试跑覆盖 M0/M1/M4/M5/M6 的等价检查；未走旧 pipeline（与 global_extent 相同）。", "gap": "已足量；本轮不再扩量。", "engine_families": [ "extent_choice_v2" ], "independent_checker": "checkers/families.py:extent", "instance_ids": [ "FC-QNP010", "FC-QNP011", "FC-CNP011", "FC-BNP002-1UBQ" ], "instance_count": 4, "independent_pass": 4, "instance_failures": 0, "instance_reasoning_cells": [ "1-10" ], "legacy_decisions": { "retain": 4 }, "legacy_choice_v01": 3 }, { "id": "named_bond_angle", "name": "命名化学键角", "ability": "perception", "status": "pilot", "priority": 1, "concept": "指定化学连接（如糖苷桥氧 C–O–C、芳环内角、甲醚键角）在给定构象中的键角；原子身份由连接关系确认，不是任意行号。", "domains": [ "chemistry" ], "input_scale": "整个分子最大间距（1–2 nm 宿主）", "reasoning_scale": "三原子两两最大间距（约 0.24 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "局部几何；结论只对该构象成立。", "inputs": "完整 XYZ（Å）、三个命名原子及行号、连接描述。", "physical_conditions": "单一给定构象，不优化。", "output_forms": [ "numeric", "choice" ], "source_basis": "SAMPL9 宿主结构（PDB CONECT 确认连接）", "answer_method": "acos 向量夹角；atan2 实现复核；共价半径规则确认顶点与两端成键。", "distractor_mechanisms": [ "idealized_value", "other_neighbor", "angle_at_adjacent_atom", "supplement", "triangle_angle_at_first_atom" ], "four_choice_validation": "与正确值 > 1.5°，干扰项之间 > 0.5°；升序排列，秩按批次平衡；易排除项每个计 4 个秩差惩罚。 2026-10-06：枚举实例的题干改用化学命名（task_families/chem_names.py）——按价态与登记的总电荷求键级（单 / 双 / 芳香 / 离域），按键级、邻接与氢数给出原子的官能团角色（如“硫脲碳 (C=S)”“磺酰氧 (S=O)”“铵氮 (NH3+)”），认不出的原子不出题；36 个分子的全部重原子都能命名。", "legacy_families": [ "specified_local_angle", "ring_internal_angle", "methyl_ether_local_geometry", "nearest_neighbor_angle" ], "legacy_qa": [ "CNP001", "CNP002", "CNP005", "CNP006", "QNP005" ], "code": [ "task_families/local_geometry.py", "task_families/kit.py", "checkers/families.py", "task_families/chem_names.py", "enumerators.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py", "tests/test_chem_names.py" ], "builder_modules": "M0 范围、M1 输入哈希、M3 族契约、M4 构造、M5 四项核验、M6 分包（族引擎内等价实现）。", "gap": "局部距离（二硫键、氢键给体–受体）尚未迁入同一接口；下一步复用同一 kit。", "engine_families": [ "named_bond_angle" ], "independent_checker": "checkers/families.py:named_bond_angle", "instance_ids": [ "FA-ANG-BCD", "FA-ANG-WP6", "FA-ANG-DMBCD", "FA-ANG-7040-OSO" ], "instance_count": 4, "independent_pass": 4, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": { "retain": 4, "revise": 1 }, "legacy_choice_v01": 0 }, { "id": "backbone_torsion", "name": "蛋白主链二面角 φ/ψ", "ability": "perception", "status": "pilot", "priority": 1, "concept": "Ramachandran 主链扭转角；符号区分手性（左/右手构象），需要按 IUPAC 约定定向计算。", "domains": [ "biology" ], "input_scale": "三残基片段最大间距（约 1 nm）", "reasoning_scale": "二面角首末原子间距（约 0.3 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "接近 ±180° 的值因舍入歧义被拒绝。", "inputs": "PDB 连续三残基 ATOM 片段（坐标不改）、残基编号与扭转角定义。", "physical_conditions": "晶体结构单一模型，无替代构象。", "output_forms": [ "numeric", "choice" ], "source_basis": "RCSB PDB 晶体结构（1CRN、1UBQ、6LYZ）", "answer_method": "法向量 atan2 与投影 atan2 两种实现；检查链连续（C–N ≤ 1.5 Å）与替代构象。", "distractor_mechanisms": [ "opposite_sign", "other_backbone_torsion", "neighbor_residue_same_torsion", "peptide_omega", "unsigned_plane_angle", "backbone_bond_angle" ], "four_choice_validation": "按 360° 周期计算间隔；与正确值 > 5°；升序排列。 v0.2.0：角度是周期量，选项按圆周递增排列，起点由批次目标位置决定（题干说明）；固定从 −180° 排序时，大多数主链二面角为负，正确项集中在 A/B（枚举探针 runs/enum-v12）。“法向量夹角丢失符号”迷惑程度调为 2。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/local_geometry.py", "task_families/kit.py", "checkers/families.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "同 named_bond_angle。", "gap": "尚无立体构型（R/S、顺反）题型；建议按会议 2 的感知→推断→设计梯子继续。", "engine_families": [ "backbone_torsion" ], "independent_checker": "checkers/families.py:backbone_torsion", "instance_ids": [ "FA-TOR-1CRN-G20", "FA-TOR-1UBQ-V26", "FA-TOR-6LYZ-T43", "EN-1A6M-TOR-2-PHI", "EN-1FHA-TOR-6-PHI", "EN-12E8-TOR-2-PHI", "EN-1HHK-TOR-2-PHI" ], "instance_count": 7, "independent_pass": 7, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "named_bond_distance", "name": "命名局部距离", "ability": "perception", "status": "pilot", "priority": 1, "concept": "指定化学实体之间的键长：磺酰 S=O、芳环 C–Cl、末端 N–N、C–NH2、二硫键 SG–SG；原子由行号或 PDB 原子名加化学描述指定，并由成键规则确认。", "domains": [ "quantum", "biology", "chemistry" ], "input_scale": "整个结构", "reasoning_scale": "两原子距离（0.1–0.5 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "单一距离不能独立判定成键状态。", "inputs": "XYZ（按行号命名）或 PDB 片段（含 SSBOND 记录与两个残基的 ATOM 行）。", "physical_conditions": "单一构象：QM7-X 的 DFTB3+MBD 优化几何，或晶体结构中的主构象；不考虑热振动与零点运动。", "output_forms": [ "numeric", "choice" ], "source_basis": "PDB、SAMPL9、QM7-X", "answer_method": "欧氏距离（浮点与 Decimal 两种实现）；共价半径规则确认成键。", "distractor_mechanisms": [ "other_bond_same_atom", "typical_value", "covalent_radii_sum", "nearest_nonbonded_same_element", "bohr_units", "one_bond_too_far" ], "four_choice_validation": "与正确值 > 0.02 Å，干扰项之间 > 0.01 Å；升序排列，秩按批次轮换。 2026-10-04 枚举探针：新增“多数了一根键”（1–3 距离）；枚举实例设 strict_rank，目标答案位置不可达时拒绝出题而不是退到最近位置。 2026-10-06：枚举实例的题干改用化学命名（task_families/chem_names.py）——按价态与登记的总电荷求键级（单 / 双 / 芳香 / 离域），按键级、邻接与氢数给出原子的官能团角色（如“硫脲碳 (C=S)”“磺酰氧 (S=O)”“铵氮 (NH3+)”），认不出的原子不出题；36 个分子的全部重原子都能命名。", "legacy_families": [ "disulfide_bond_length", "carboxylate_bond_metric", "donor_acceptor_distance", "shortest_element_pair_distance", "closest_interchain_contact", "watson_crick_hbond_geometry" ], "legacy_qa": [ "BNP007", "BNP008", "BNP010", "CNP003", "HBP001", "QNP002", "QNP003", "QNP004", "QNP006", "CNP010", "BNP012" ], "code": [ "task_families/local_geometry.py", "checkers/migrated.py", "task_families/chem_names.py", "enumerators.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py", "tests/test_chem_names.py" ], "builder_modules": "族引擎内等价实现；旧 QNP002–004/006、BNP007 由新版取代（supersede）。", "gap": "氢键给体–受体、链间接触等非键距离尚未迁移。", "engine_families": [ "named_bond_distance" ], "independent_checker": "checkers/migrated.py:named_bond_distance", "instance_ids": [ "FA-DIST-7035-SO", "FA-DIST-7070-CCl", "FA-DIST-7002-NN", "FA-DIST-7122-CN", "FA-DIST-6LYZ-SS", "FA-DIST-6LYZ-SS30", "FA-DIST-1CRN-SS16", "FA-DIST-WP6-C20O22" ], "instance_count": 8, "independent_pass": 8, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": { "revise": 6, "retain": 5 }, "legacy_choice_v01": 0 }, { "id": "coordination_shell", "name": "配位壳层", "ability": "perception", "status": "pilot", "priority": 1, "concept": "无限周期晶体中，指定中心周围某一物种的最近壳层（个数与距离）；包括网络节点层面的壳层（SALEM-2 的 Zn–Zn 四连接、MOF-5 μ4-O 节点的六连接）。", "domains": [ "materials" ], "input_scale": "晶胞边长", "reasoning_scale": "第一壳层半径", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "只读完全占位的 CIF；非立方晶胞按 CIF 全部对称操作展开（task_families/cell.py）。", "inputs": "常规晶胞全部原子分数坐标表 + 晶胞参数（立方只给 a，其他晶系给 a、b、c、α、β、γ）；中心原子的分数坐标与伙伴物种描述。", "physical_conditions": "周期边界；完全占位。", "output_forms": [ "choice" ], "source_basis": "COD", "answer_method": "立方：27 个相邻晶胞映像的近邻搜索，按 0.001 Å 归组壳层。其他晶系：125 个映像，问最短距离及该距离上的原子数（低对称把配位多面体拆成几种距离，“整个多面体”是登记的错误机制）。", "distractor_mechanisms": [ "no_periodic_images", "second_shell", "same_species_shell", "mixed_shells", "merged_first_two_shells", "whole_coordination_polyhedron" ], "four_choice_validation": "选项为“N at D Å”，恰好一项同时匹配个数与距离。", "legacy_families": [ "periodic_tetrahedral_coordination", "primitive_cubic_coordination_shell", "species_resolved_coordination_shell" ], "legacy_qa": [ "MNP001", "MNP002", "MNP004", "MNP005" ], "code": [ "task_families/crystal.py", "checkers/migrated.py", "tools/recompute_materials_batch2.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；旧 MNP004/005 由新版取代。", "gap": "已支持任意晶系（2026-10-03：金红石、ZnO、α-石英）；部分占位与无序结构仍不支持。", "engine_families": [ "coordination_shell" ], "independent_checker": "checkers/migrated.py:coordination_shell", "instance_ids": [ "FA-COORD-MGO", "FA-COORD-CSCL", "FA-COORD-SALEM2-ZN", "FA-COORD-MOF5-NODE", "FA-COORD-RUTILE", "FA-COORD-ZNO", "FA-COORD-QUARTZ", "FA-COORD-SI", "FA-COORD-NACL-CL", "FA-COORD-NACL-NA", "FA-COORD-CU", "FA-COORD-FE", "FA-COORD-MG", "FA-COORD-ALNI", "FA-COORD-B5H9", "FA-COORD-K2CUF4", "FA-COORD-MN2CO10", "FA-COORD-PARA-I", "FA-COORD-PARA-II", "FA-COORD-PENTACENE", "FA-COORD-HAP", "FA-COORD-ARAGONITE" ], "instance_count": 22, "independent_pass": 22, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1", "1-10" ], "legacy_decisions": { "retain": 2, "revise": 2 }, "legacy_choice_v01": 0 }, { "id": "assembly_size", "name": "大型组装体尺寸（最大间距 / 回转半径）", "ability": "perception", "status": "pilot", "priority": 1, "concept": "整个颗粒尺度的几何量：分子伴侣、核小体、病毒衣壳、DNA 折纸螺旋束的最大间距与等权回转半径（10–1000 nm）。", "domains": [ "chemistry", "biology" ], "input_scale": "整个组装体", "reasoning_scale": "所选点集的最大间距（整个颗粒）", "scale_cells": [ "10-100", "100-1000" ], "scale_limits": "约简表示上的几何量，不是溶液中的流体力学尺寸。暂无数据源的格：chemistry 100-1000（规划器不据此给出路线）。", "inputs": "每残基一点的 XYZ，或按组分标记的链质心表（可只选某一组分，如 VP5 外壳）。", "physical_conditions": "约简表示；几何量在所给点集上定义。", "output_forms": [ "numeric", "choice" ], "source_basis": "RCSB PDB 大型组装体（1AON、1KX5、1A34、6CGR、6NCL、7ARQ），经沉积的对称操作展开并约简为每残基/每核苷酸一点或每链质心", "answer_method": "numpy 分块搜索最大点对 + Decimal 复核；回转半径用质心公式并以 Decimal 复核。", "distractor_mechanisms": [ "bounding_box_diagonal", "centroid_diameter", "largest_axis_span", "sphere_radius_from_rg", "single_axis_rms", "wrong_selection", "sphere_diameter_from_rg", "rg_from_dmax_solid_sphere" ], "four_choice_validation": "与正确值 > 0.3 nm；升序排列，秩按批次轮换（加入偏小的错误机制，避免正确项总是最小）。 v0.3.0：加入两个方向的实心球换算；空心外壳上其他尺寸估计常与 Dmax 重合而被拒（探针 runs/enum-v23）。枚举的回转半径题设 strict_rank。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/assembly.py", "checkers/migrated.py", "tools/build_assembly_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；原始下载保存在本地 inputs/pdb（不入库），约简资产及来源哈希入库。", "gap": "材料、量子的大尺度对象待接入（按论文参数构建的模型）。", "engine_families": [ "assembly_extent", "assembly_rg" ], "independent_checker": "checkers/migrated.py:assembly_extent, assembly_rg", "instance_ids": [ "FA-ASM-EXT-1AON", "FA-ASM-EXT-1KX5", "FA-ASM-EXT-1A34", "FA-ASM-EXT-6CGR", "FA-ASM-EXT-7ARQ", "FA-ASM-EXT-7ARQ-SCAF", "FA-ASM-EXT-6NCL", "FA-ASM-EXT-1RYP", "FA-ASM-EXT-1STM", "FA-ASM-EXT-2BUK", "FA-ASM-EXT-1CWP", "FA-ASM-EXT-1SVA", "FA-ASM-EXT-2FT1", "FA-ASM-RG-1AON", "FA-ASM-RG-1KX5", "FA-ASM-RG-1A34", "FA-ASM-RG-6CGR", "FA-ASM-RG-7ARQ", "FA-ASM-RG-7ARQ-SCAF", "FA-ASM-RG-1RYP", "FA-ASM-RG-1STM", "FA-ASM-RG-2BUK", "FA-ASM-RG-1CWP", "FA-ASM-RG-1SVA", "FA-ASM-RG-2FT1" ], "instance_count": 25, "independent_pass": 25, "instance_failures": 0, "instance_reasoning_cells": [ "10-100", "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "capsid_architecture", "name": "病毒衣壳结构（Caspar–Klug）", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由外壳蛋白拷贝数 N 推断二十面体衣壳的三角剖分数 T=N/60、壳粒总数 10T+2 与六邻体数 10(T−1)；需要从整个衣壳计数并运用结构理论。 大型 dsDNA 病毒（PBCV-1）的主衣壳蛋白形成三聚体假六聚壳粒、五邻体由另一蛋白构成：N = 30(T−1)。", "domains": [ "biology" ], "input_scale": "整个衣壳", "reasoning_scale": "整个外壳（所选组分点集的最大间距）", "scale_cells": [ "10-100", "100-1000" ], "scale_limits": "假设完整二十面体外壳，不区分门户顶点等局部不对称。", "inputs": "按组分标记的链质心表（沉积的 60 个二十面体操作展开）。", "physical_conditions": "约简表示；几何量在所给点集上定义。", "output_forms": [ "numeric", "choice" ], "source_basis": "RCSB PDB 大型组装体（1AON、1KX5、1A34、6CGR、6NCL、7ARQ），经沉积的对称操作展开并约简为每残基/每核苷酸一点或每链质心", "answer_method": "计数所选组分的拷贝数，按 Caspar–Klug 关系计算。", "distractor_mechanisms": [ "neighbouring_allowed_T", "all_hexamers", "all_pentamers", "hexons_only", "pentons_only", "hexon_subunits", "monomer_60T_rule", "trimers_as_60T", "forgot_pentons", "hexamer_rule" ], "four_choice_validation": "整数选项互不相同；独立检查器重新计数。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/assembly.py", "checkers/migrated.py", "tools/build_assembly_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；原始下载保存在本地 inputs/pdb（不入库），约简资产及来源哈希入库。", "gap": "目前 3 个衣壳（STMV T=1、HSV-1 T=16、PBCV-1 T=169）。", "engine_families": [ "capsid_architecture" ], "independent_checker": "checkers/migrated.py:capsid_architecture", "instance_ids": [ "FA-ASM-CAPS-1A34", "FA-ASM-T-6CGR", "FA-ASM-CAPS-6CGR", "FA-ASM-HEX-6CGR", "FA-ASM-T-6NCL", "FA-ASM-T-1STM", "FA-ASM-CAPS-2BUK", "FA-ASM-T-1CWP", "FA-ASM-CAPS-1CWP", "FA-ASM-HEX-1CWP", "FA-ASM-T-2FT1", "FA-ASM-HEX-2FT1" ], "instance_count": 12, "independent_pass": 12, "instance_failures": 0, "instance_reasoning_cells": [ "10-100", "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "whole_particle_scattering", "name": "整颗粒散射强度（Debye，超出 Guinier 范围）", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由整个颗粒的点集精确计算取向平均的归一化散射强度 I(q)/I(0)，q·Rg > 1.3 时 Guinier 近似与形状模型均失效，必须用全部点对。", "domains": [ "chemistry", "biology" ], "input_scale": "整个颗粒", "reasoning_scale": "所选点集的最大间距（探测长度 2π/q 同量级）", "scale_cells": [ "10-100", "100-1000" ], "scale_limits": "约简表示；非实测 SAXS。暂无数据源的格：chemistry 100-1000（规划器不据此给出路线）。", "inputs": "按组分标记的点表（DNA 折纸每核苷酸 C1′；HSV-1 衣壳每链质心），可只选某一组分。", "physical_conditions": "等权点散射体，无形状因子与溶剂；q = 4π sinθ/λ（题干写明）。", "output_forms": [ "numeric", "choice" ], "source_basis": "RCSB PDB 大型组装体（1AON、1KX5、1A34、6CGR、6NCL、7ARQ），经沉积的对称操作展开并约简为每残基/每核苷酸一点或每链质心", "answer_method": "Debye 公式 [N + 2Σ_{i 0.02；升序排列，秩按批次轮换。 v0.3.0：加入两种给出更小强度的 Guinier 误用；之前探针中正确项在 30/36 道是最小值（runs/enum-v26）。q 必须恰为题干显示的三位小数，否则拒绝。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/assembly.py", "checkers/migrated.py", "tools/build_assembly_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；原始下载保存在本地 inputs/pdb（不入库），约简资产及来源哈希入库。", "gap": "7ARQ、6CGR、6NCL。", "engine_families": [ "debye_intensity" ], "independent_checker": "checkers/migrated.py:debye_intensity", "instance_ids": [ "FA-DEB-7ARQ-018", "FA-DEB-7ARQ-030", "FA-DEB-6CGR", "FA-DEB-1RYP", "FA-DEB-2BUK", "FA-DEB-1CWP", "FA-DEB-1SVA", "FA-DEB-2FT1" ], "instance_count": 8, "independent_pass": 8, "instance_failures": 0, "instance_reasoning_cells": [ "10-100", "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "scattering_q_design", "name": "测量条件设计：使散射强度降到目标的 q", "ability": "design", "status": "pilot", "priority": 1, "concept": "推断的逆问题：给定目标 I(q)/I(0)，选择测量 q，使整个颗粒的取向平均强度首次降到该值；需要精确 Debye 计算，形状模型与 Guinier 反解都会偏。", "domains": [ "chemistry", "biology" ], "input_scale": "整个颗粒", "reasoning_scale": "所选点集的最大间距（探测长度 2π/q 同量级）", "scale_cells": [ "10-100", "100-1000" ], "scale_limits": "约简表示；非实测 SAXS。暂无数据源的格：chemistry 100-1000（规划器不据此给出路线）。", "inputs": "按组分标记的点表（DNA 折纸每核苷酸 C1′；HSV-1 衣壳每链质心），可只选某一组分。", "physical_conditions": "等权点散射体，无形状因子与溶剂；q = 4π sinθ/λ（题干写明）。", "output_forms": [ "numeric", "choice" ], "source_basis": "RCSB PDB 大型组装体（1AON、1KX5、1A34、6CGR、6NCL、7ARQ），经沉积的对称操作展开并约简为每残基/每核苷酸一点或每链质心", "answer_method": "从 q = 0 以 0.02/Rg 步进找首个穿越点，再二分 60 次；两种 Debye 实现在解处一致。", "distractor_mechanisms": [ "guinier_root", "uniform_sphere_root", "single_orientation_root", "target_as_amplitude", "reported_as_s", "dmax_sphere_root", "reported_per_angstrom", "sphere_radius_equals_rg_root", "guinier_single_axis_root", "guinier_missing_one_third_root", "guinier_half_dmax_root" ], "four_choice_validation": "与正确值相差 > 5%·q*；升序排列，秩按批次轮换。 v0.3.0：同上两种 Guinier 误用（解出更小的 q）；之前探针答案多在 B。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/assembly.py", "checkers/migrated.py", "tools/build_assembly_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；原始下载保存在本地 inputs/pdb（不入库），约简资产及来源哈希入库。", "gap": "7ARQ、6CGR、6NCL、1AON、1KX5、1A34。", "engine_families": [ "scattering_q_design" ], "independent_checker": "checkers/migrated.py:scattering_q_design", "instance_ids": [ "FA-QDES-7ARQ-050", "FA-QDES-7ARQ-030", "FA-QDES-7ARQ-SCAF-020", "FA-QDES-6CGR", "FA-QDES-6NCL", "FA-QDES-1AON", "FA-QDES-1KX5", "FA-QDES-1A34", "FA-QDES-1RYP", "FA-QDES-1CWP", "FA-QDES-1SVA", "FA-QDES-2FT1" ], "instance_count": 12, "independent_pass": 12, "instance_failures": 0, "instance_reasoning_cells": [ "10-100", "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "superlattice_geometry", "name": "纳米粒子超晶格几何（最近邻壳层 / 晶胞 / 密排面间距）", "ability": "perception", "status": "pilot", "priority": 1, "concept": "从旋转后的粒子中心坐标识别晶格：最近邻壳层（个数与距离）、常规立方晶胞边长、最密排面间距，均在晶格周期尺度（10–100 nm）。", "domains": [ "materials" ], "input_scale": "整个晶粒（约 170–340 nm）", "reasoning_scale": "晶格周期（常规晶胞边长 a，约 40–77 nm）", "scale_cells": [ "10-100" ], "scale_limits": "按论文参数构建的模型，不是实测粒子坐标；不含位置无序、峰宽与形状因子。", "inputs": "DNA 连接金纳米粒子超晶格晶粒模型的粒子中心 XYZ（nm，约 177 个，经确定性旋转）；设计题给四个候选文件。", "physical_conditions": "理想无限 FCC 晶格（按论文 SAXS 晶胞边长）；粒子视为相同点散射体；金核直径 10.3 nm 在题干给出。", "output_forms": [ "choice", "numeric" ], "source_basis": "Hill et al., Nano Lett. 8, 2341 (2008) 表 1：五种 DNA 连接子的 SAXS 晶胞边长（论文参数模型）", "answer_method": "中心粒子周围分壳（容差 1e-4 nm），核对 12/6 与 √2 比例（FCC），并与表 1 晶胞边长一致。", "distractor_mechanisms": [ "second_shell", "surface_particle", "bcc_assumption", "nearest_neighbour", "face_diagonal", "primitive_cell_root", "d200", "in_layer_row_spacing", "d220" ], "four_choice_validation": "壳层选项互不相同；数值与正确值 > 1 nm；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/superlattice.py", "checkers/superlattice.py", "tools/build_superlattice_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文表 1 参数确定性生成，sources.json 记录参数与构建方式。", "gap": "目前 1 篇论文的 5 个晶格；可加入其他超晶格（BCC、二元）。", "engine_families": [ "superlattice_shell", "superlattice_metric" ], "independent_checker": "checkers/superlattice.py:superlattice_shell, superlattice_quantity", "instance_ids": [ "FA-SL-SHELL-X", "FA-SL-EDGE-26", "FA-SL-EDGE-52", "FA-SL-D111-39" ], "instance_count": 4, "independent_pass": 4, "instance_failures": 0, "instance_reasoning_cells": [ "10-100" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "superlattice_scattering", "name": "纳米粒子超晶格 SAXS 反射与金含量", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由粒子坐标推断可观测量：第一与第二个非零 Bragg 反射的 q（q = 4π sinθ/λ），以及给定金核直径时的金体积百分数。", "domains": [ "materials" ], "input_scale": "整个晶粒（约 170–340 nm）", "reasoning_scale": "晶格周期（常规晶胞边长 a，约 40–77 nm）", "scale_cells": [ "10-100" ], "scale_limits": "按论文参数构建的模型，不是实测粒子坐标；不含位置无序、峰宽与形状因子。", "inputs": "DNA 连接金纳米粒子超晶格晶粒模型的粒子中心 XYZ（nm，约 177 个，经确定性旋转）；设计题给四个候选文件。", "physical_conditions": "理想无限 FCC 晶格（按论文 SAXS 晶胞边长）；粒子视为相同点散射体；金核直径 10.3 nm 在题干给出。", "output_forms": [ "choice", "numeric" ], "source_basis": "Hill et al., Nano Lett. 8, 2341 (2008) 表 1：五种 DNA 连接子的 SAXS 晶胞边长（论文参数模型）", "answer_method": "q* = 2π√3/a，q(200) = 4π/a；金含量 = 4·(4/3)πr³/a³。", "distractor_mechanisms": [ "q_100_forbidden", "q_from_nearest_neighbour", "q_200", "q_111_first_peak", "q_220", "missing_two_pi", "per_angstrom", "one_particle_per_cell", "two_particles_per_cell", "diameter_as_radius", "four_per_primitive_cell" ], "four_choice_validation": "与正确值 > 0.005 nm⁻¹ 或 0.05 个百分点；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/superlattice.py", "checkers/superlattice.py", "tools/build_superlattice_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文表 1 参数确定性生成，sources.json 记录参数与构建方式。", "gap": "与超晶格几何相同：目前 1 篇论文（Hill 2008）的 5 个晶格；可加入其他超晶格（BCC、二元）。", "engine_families": [ "superlattice_saxs" ], "independent_checker": "checkers/superlattice.py:superlattice_quantity", "instance_ids": [ "FA-SL-QSTAR-X", "FA-SL-Q200-13", "FA-SL-GOLD-52" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "10-100" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "superlattice_design", "name": "超晶格连接子设计选择", "ability": "design", "status": "pilot", "priority": 1, "concept": "在论文实际做出的四种 DNA 连接子中选择一种，使超晶格达到目标（首个 Bragg 反射、最近邻间距，或在金含量约束下间距最大）；四个候选各有模型晶粒。", "domains": [ "materials" ], "input_scale": "整个晶粒（约 170–340 nm）", "reasoning_scale": "晶格周期（常规晶胞边长 a，约 40–77 nm）", "scale_cells": [ "10-100" ], "scale_limits": "按论文参数构建的模型，不是实测粒子坐标；不含位置无序、峰宽与形状因子。", "inputs": "DNA 连接金纳米粒子超晶格晶粒模型的粒子中心 XYZ（nm，约 177 个，经确定性旋转）；设计题给四个候选文件。", "physical_conditions": "理想无限 FCC 晶格（按论文 SAXS 晶胞边长）；粒子视为相同点散射体；金核直径 10.3 nm 在题干给出。", "output_forms": [ "choice" ], "source_basis": "Hill et al., Nano Lett. 8, 2341 (2008) 表 1：五种 DNA 连接子的 SAXS 晶胞边长（论文参数模型）", "answer_method": "逐个候选由坐标计算性质，按目标规则取唯一最优；要求领先幅度足够。", "distractor_mechanisms": [ "q_from_nearest_neighbour", "q_200_as_first", "cell_edge_as_spacing", "plane_spacing_as_spacing", "diameter_as_radius_passes", "highest_gold_not_largest_spacing", "satisfies_but_smaller_spacing", "farther_from_target" ], "four_choice_validation": "恰好一个候选满足目标；每个错误选项注明在哪种错误下会被选中。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/superlattice.py", "checkers/superlattice.py", "tools/build_superlattice_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文表 1 参数确定性生成，sources.json 记录参数与构建方式。", "gap": "与超晶格几何相同：目前 1 篇论文（Hill 2008）的 5 个晶格；可加入其他超晶格（BCC、二元）。", "engine_families": [ "superlattice_design" ], "independent_checker": "checkers/superlattice.py:superlattice_design", "instance_ids": [ "FA-SL-DES-QSTAR", "FA-SL-DES-NN", "FA-SL-DES-GOLD" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "10-100" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "opal_geometry", "name": "胶体光子晶体几何（球径 / 密排面间距 / 晶胞）", "ability": "perception", "status": "pilot", "priority": 1, "concept": "从旋转后的球心坐标识别接触球 FCC：球径（最近邻距离）、最密排面间距、常规晶胞边长，均在 100–1000 nm。", "domains": [ "materials" ], "input_scale": "整个晶粒（约 1.3–3.3 µm）", "reasoning_scale": "晶格周期：常规晶胞边长 a（约 295–522 nm）", "scale_cells": [ "100-1000" ], "scale_limits": "按论文球径构建的模型；不含尺寸分布、堆垛缺陷与多面衍射；审核侧附实测峰值供比较。", "inputs": "胶体晶体（蛋白石）晶粒模型的球心坐标（nm，每行 \"sphere x y z\"，177 个，经确定性旋转）；设计题给四个候选文件。", "physical_conditions": "接触球理想 FCC；密排面平行于膜面；有效折射率 n_eff² = Σ f_i n_i²（或题给 n_eff）；Bragg–Snell 一级反射。", "output_forms": [ "choice", "numeric" ], "source_basis": "Gazmeh et al., Sci. Rep. 15, 38349 (2025) 表 3（聚苯乙烯球径与实测反射峰）；Fookes et al., Sensors 23, 1433 (2023) 参数表（二氧化硅 266 nm、n_eff 1.35、30°）", "answer_method": "中心球周围分壳，核对 12/6 与 √2 比例，并与论文球径一致。", "distractor_mechanisms": [ "cell_edge", "radius", "d111", "half_cell_edge", "d200", "in_layer_row_spacing", "nearest_neighbour", "third_shell_distance", "face_diagonal" ], "four_choice_validation": "与正确值 > 3 nm；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/photonic.py", "checkers/superlattice.py", "tools/build_opal_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文参数确定性生成，sources.json 记录参数、出处与实测值。", "gap": "两篇论文（聚苯乙烯 7 个样品、二氧化硅 1 个）。", "engine_families": [ "opal_geometry" ], "independent_checker": "checkers/superlattice.py:opal_quantity", "instance_ids": [ "FA-OP-DIAM-S8", "FA-OP-D111-S2", "FA-OP-EDGE-SIO2" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "opal_bragg", "name": "胶体光子晶体 Bragg 反射与结构色", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由球心坐标与题给折射率推断结构色：一级 Bragg–Snell 反射波长（正入射或斜入射），或达到目标波长所需的入射角。", "domains": [ "materials" ], "input_scale": "整个晶粒（约 1.3–3.3 µm）", "reasoning_scale": "晶格周期：常规晶胞边长 a（约 295–522 nm）", "scale_cells": [ "100-1000" ], "scale_limits": "按论文球径构建的模型；不含尺寸分布、堆垛缺陷与多面衍射；审核侧附实测峰值供比较。", "inputs": "胶体晶体（蛋白石）晶粒模型的球心坐标（nm，每行 \"sphere x y z\"，177 个，经确定性旋转）；设计题给四个候选文件。", "physical_conditions": "接触球理想 FCC；密排面平行于膜面；有效折射率 n_eff² = Σ f_i n_i²（或题给 n_eff）；Bragg–Snell 一级反射。", "output_forms": [ "choice", "numeric" ], "source_basis": "Gazmeh et al., Sci. Rep. 15, 38349 (2025) 表 3（聚苯乙烯球径与实测反射峰）；Fookes et al., Sensors 23, 1433 (2023) 参数表（二氧化硅 266 nm、n_eff 1.35、30°）", "answer_method": "λ = 2 d111 √(n_eff² − sin²θ)；n_eff² = f n_s² + (1−f) n_m²，f = π/(3√2)。", "distractor_mechanisms": [ "diameter_as_spacing", "cell_edge_as_spacing", "no_factor_two", "d200_planes", "linear_index_average", "no_index", "sphere_index", "voids_as_air", "no_refraction", "ignore_angle", "angle_from_surface", "refraction_angle_inside", "sin_not_squared" ], "four_choice_validation": "波长与正确值 > 5 nm，角度 > 1°；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/photonic.py", "checkers/superlattice.py", "tools/build_opal_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文参数确定性生成，sources.json 记录参数、出处与实测值。", "gap": "与胶体晶体几何相同：两篇论文（聚苯乙烯 7 个样品、二氧化硅 1 个）。", "engine_families": [ "opal_bragg" ], "independent_checker": "checkers/superlattice.py:opal_quantity", "instance_ids": [ "FA-OP-PEAK-S3", "FA-OP-PEAK-S7-WATER", "FA-OP-PEAK-SIO2-30", "FA-OP-ANGLE-SIO2" ], "instance_count": 4, "independent_pass": 4, "instance_failures": 0, "instance_reasoning_cells": [ "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "opal_design", "name": "结构色设计：选择球径样品", "ability": "design", "status": "pilot", "priority": 1, "concept": "在论文实际合成的四个聚苯乙烯样品中选一个，使膜的正入射反射最接近目标波长，或在波长上限内球径最大；四个候选各有模型晶粒。", "domains": [ "materials" ], "input_scale": "整个晶粒（约 1.3–3.3 µm）", "reasoning_scale": "晶格周期：常规晶胞边长 a（约 295–522 nm）", "scale_cells": [ "100-1000" ], "scale_limits": "按论文球径构建的模型；不含尺寸分布、堆垛缺陷与多面衍射；审核侧附实测峰值供比较。", "inputs": "胶体晶体（蛋白石）晶粒模型的球心坐标（nm，每行 \"sphere x y z\"，177 个，经确定性旋转）；设计题给四个候选文件。", "physical_conditions": "接触球理想 FCC；密排面平行于膜面；有效折射率 n_eff² = Σ f_i n_i²（或题给 n_eff）；Bragg–Snell 一级反射。", "output_forms": [ "choice" ], "source_basis": "Gazmeh et al., Sci. Rep. 15, 38349 (2025) 表 3（聚苯乙烯球径与实测反射峰）；Fookes et al., Sensors 23, 1433 (2023) 参数表（二氧化硅 266 nm、n_eff 1.35、30°）", "answer_method": "逐个候选由坐标计算反射波长，按目标规则取唯一最优；要求领先幅度足够。", "distractor_mechanisms": [ "diameter_as_spacing", "no_index", "linear_index_average", "sphere_index", "passes_if_no_index", "satisfies_but_smaller", "farther_from_target" ], "four_choice_validation": "恰好一个候选满足目标；每个错误选项注明在哪种错误下会被选中。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/photonic.py", "checkers/superlattice.py", "tools/build_opal_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文参数确定性生成，sources.json 记录参数、出处与实测值。", "gap": "与胶体晶体几何相同：两篇论文（聚苯乙烯 7 个样品、二氧化硅 1 个）。", "engine_families": [ "opal_design" ], "independent_checker": "checkers/superlattice.py:opal_design", "instance_ids": [ "FA-OP-DES-540", "FA-OP-DES-820", "FA-OP-DES-BELOW700" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "moire_geometry", "name": "转角石墨烯莫尔晶格几何（周期 / 晶胞面积）", "ability": "perception", "status": "pilot", "priority": 1, "concept": "从 AA 堆垛区中心图识别莫尔晶格：莫尔周期与莫尔晶胞面积（10–100 nm 尺度）。", "domains": [ "quantum" ], "input_scale": "AA 图约 70–90 nm；原子块约 4 nm", "reasoning_scale": "莫尔周期 λ（12–18 nm）", "scale_cells": [ "10-100" ], "scale_limits": "按论文转角构建的模型；实测样品常有少量异质应变，使莫尔周期偏离理想值（Kerelsky et al. 讨论）。", "inputs": "转角双层石墨烯模型：AA 堆垛区中心图（nm，每行 \"AA x y\"，相当于 STM 形貌亮点，确定性旋转）或 AA 位点附近两层全部碳原子（Å）；设计题给四个候选图。", "physical_conditions": "刚性转角（无晶格弛豫、无异质应变）；石墨烯晶格常数 0.246 nm；自旋×谷四重简并（n_s = 4/A）。", "output_forms": [ "choice", "numeric" ], "source_basis": "Cao et al., arXiv:1803.02342（Nature 2018）器件 M1/M2/D1 转角与 n_s = 4/A 定义；Kerelsky et al., arXiv:1812.08776（Nature 2019）STM 区域 1.10°、0.79°", "answer_method": "中心位点周围 6 个等距近邻，核对与论文转角给出的周期一致。", "distractor_mechanisms": [ "aa_to_ab_distance", "row_spacing", "second_neighbour", "half_period", "double_period", "square_cell", "triangle_only", "hexagon_twice", "ab_spacing_cell" ], "four_choice_validation": "与正确值 > 0.3 nm 或 3 nm²；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/moire.py", "checkers/moire.py", "tools/build_moire_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文转角确定性生成，sources.json 记录出处与模型。", "gap": "两篇论文的 5 个转角（0.79–1.16°）。", "engine_families": [ "moire_geometry" ], "independent_checker": "checkers/moire.py:moire_quantity", "instance_ids": [ "FA-MO-PERIOD-K079", "FA-MO-AREA-M2", "FA-MO-PERIOD-D1" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "10-100" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "moire_inference", "name": "转角石墨烯莫尔推断（周期、载流子密度、转角）", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由原子坐标推断两层转角并求莫尔周期；由莫尔晶胞推断完全填满与半满一组自旋×谷简并能带的载流子密度；由莫尔周期反推转角。", "domains": [ "quantum" ], "input_scale": "AA 图约 70–90 nm；原子块约 4 nm", "reasoning_scale": "莫尔周期 λ（12–18 nm）", "scale_cells": [ "10-100" ], "scale_limits": "按论文转角构建的模型；实测样品常有少量异质应变，使莫尔周期偏离理想值（Kerelsky et al. 讨论）。", "inputs": "转角双层石墨烯模型：AA 堆垛区中心图（nm，每行 \"AA x y\"，相当于 STM 形貌亮点，确定性旋转）或 AA 位点附近两层全部碳原子（Å）；设计题给四个候选图。", "physical_conditions": "刚性转角（无晶格弛豫、无异质应变）；石墨烯晶格常数 0.246 nm；自旋×谷四重简并（n_s = 4/A）。", "output_forms": [ "choice", "numeric" ], "source_basis": "Cao et al., arXiv:1803.02342（Nature 2018）器件 M1/M2/D1 转角与 n_s = 4/A 定义；Kerelsky et al., arXiv:1812.08776（Nature 2019）STM 区域 1.10°、0.79°", "answer_method": "λ = a/(2 sin(θ/2))；n_s = 4/A（A = (√3/2)λ²），半满 n_s/2；θ = 2 arcsin(a/2λ)。", "distractor_mechanisms": [ "sin_full_angle", "bond_length_as_lattice_constant", "angle_of_one_layer", "second_neighbour", "layer_spacing_as_lattice_constant", "no_degeneracy", "spin_only", "square_cell", "eight_per_cell", "ab_distance_as_period", "full_filling", "half_angle", "radians_as_degrees" ], "four_choice_validation": "与正确值 > 0.3 nm、0.05×10¹² cm⁻² 或 0.05°；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/moire.py", "checkers/moire.py", "tools/build_moire_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文转角确定性生成，sources.json 记录出处与模型。", "gap": "与莫尔几何相同：两篇论文的 5 个转角（0.79–1.16°）。", "engine_families": [ "moire_inference" ], "independent_checker": "checkers/moire.py:moire_quantity", "instance_ids": [ "FA-MO-ATOMS-K110", "FA-MO-NFULL-M1", "FA-MO-NHALF-K079", "FA-MO-TWIST-D1" ], "instance_count": 4, "independent_pass": 4, "instance_failures": 0, "instance_reasoning_cells": [ "10-100" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "moire_design", "name": "转角石墨烯样品选择设计", "ability": "design", "status": "pilot", "priority": 1, "concept": "在论文实际样品中选一个，使半满或完全填满密度、或莫尔周期最接近目标；四个候选各有 AA 图。", "domains": [ "quantum" ], "input_scale": "AA 图约 70–90 nm；原子块约 4 nm", "reasoning_scale": "莫尔周期 λ（12–18 nm）", "scale_cells": [ "10-100" ], "scale_limits": "按论文转角构建的模型；实测样品常有少量异质应变，使莫尔周期偏离理想值（Kerelsky et al. 讨论）。", "inputs": "转角双层石墨烯模型：AA 堆垛区中心图（nm，每行 \"AA x y\"，相当于 STM 形貌亮点，确定性旋转）或 AA 位点附近两层全部碳原子（Å）；设计题给四个候选图。", "physical_conditions": "刚性转角（无晶格弛豫、无异质应变）；石墨烯晶格常数 0.246 nm；自旋×谷四重简并（n_s = 4/A）。", "output_forms": [ "choice" ], "source_basis": "Cao et al., arXiv:1803.02342（Nature 2018）器件 M1/M2/D1 转角与 n_s = 4/A 定义；Kerelsky et al., arXiv:1812.08776（Nature 2019）STM 区域 1.10°、0.79°", "answer_method": "逐个候选由 AA 图计算周期与面积，按目标规则取唯一最优；要求领先幅度足够。", "distractor_mechanisms": [ "full_filling", "no_degeneracy", "spin_only", "square_cell", "aa_to_ab_distance", "second_neighbour", "farther_from_target" ], "four_choice_validation": "恰好一个候选满足目标；每个错误选项注明在哪种错误下会被选中。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/moire.py", "checkers/moire.py", "tools/build_moire_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文转角确定性生成，sources.json 记录出处与模型。", "gap": "与莫尔几何相同：两篇论文的 5 个转角（0.79–1.16°）。", "engine_families": [ "moire_design" ], "independent_checker": "checkers/moire.py:moire_design", "instance_ids": [ "FA-MO-DES-HALF", "FA-MO-DES-PERIOD", "FA-MO-DES-FULL" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "10-100" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "secondary_structure", "name": "由 φ/ψ 推断二级结构（HELIX/SHEET 为证据）", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由三残基 ATOM 片段计算中心残基的 φ/ψ，按题给参考构象（右手 α、β 链、PPII、左手 α）取最近者；条目自身的 HELIX/SHEET 记录作为审核证据，二者不一致即拒绝出题。", "domains": [ "biology" ], "input_scale": "三残基（约 1 nm）", "reasoning_scale": "定义 φ、ψ 的五个主链原子（约 0.3–0.4 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "单残基局部构象；不判断整段二级结构。", "inputs": "PDB 三残基 ATOM 片段（逐行照抄）。", "physical_conditions": "IUPAC 符号约定；(φ, ψ) 平面距离，360° 周期。", "output_forms": [ "choice" ], "source_basis": "RCSB PDB 1UBQ、6LYZ、1CRN（仓库已有）", "answer_method": "两种二面角实现一致；最近参考构象领先 ≥ 30%；与 HELIX/SHEET 记录一致。", "distractor_mechanisms": [ "second_nearest_reference", "mirror_image", "other_reference" ], "four_choice_validation": "四个选项恰为四个参考构象；唯一最近者。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/local_geometry.py", "checkers/migrated.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现。", "gap": "3 个残基（2 个 β、1 个 α）。", "engine_families": [ "secondary_structure" ], "independent_checker": "checkers/migrated.py:secondary_structure", "instance_ids": [ "FA-SS-1UBQ-43", "FA-SS-6LYZ-90", "FA-SS-1CRN-34", "EN-1A6M-SS-6", "EN-12E8-SS-3", "EN-1HHK-SS-3" ], "instance_count": 6, "independent_pass": 6, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "binding_evidence_choice", "name": "主客体结合强弱（实测证据）", "ability": "inference", "status": "pilot", "priority": 1, "concept": "给 WP6 主体与四个客体的 3D 结构，推断哪一个结合最强；答案依据 SAMPL9 实测 ITC ΔG（证据只在审核侧），最强者须领先 ≥ 0.8 kcal/mol。", "domains": [ "chemistry" ], "input_scale": "主体约 1.8 nm", "reasoning_scale": "主客体复合物（主体最大原子间距）", "scale_cells": [ "1-10" ], "scale_limits": "结合强弱取决于电荷、尺寸匹配、溶剂化等，不能单凭几何算出。", "inputs": "WP6 主体与四个客体 XYZ（SAMPL9 结构）。", "physical_conditions": "ITC，1× PBS，pH 7.40，298.15 K，1:1。", "output_forms": [ "choice" ], "source_basis": "SAMPL9 WP6 主客体挑战（MIT）：实测结合自由能与客体结构", "answer_method": "按实测 ΔG 排序；领先幅度检查。", "distractor_mechanisms": [ "next_strongest", "dication_but_weaker", "weaker_binder" ], "four_choice_validation": "四个候选均有实测值；最强者领先 ≥ 0.8 kcal/mol。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/hostguest.py", "checkers/hostguest.py", "tools/build_hostguest_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；实测表逐项转录并记录原表哈希。", "gap": "3 组候选。", "engine_families": [ "binding_evidence_choice" ], "independent_checker": "checkers/hostguest.py:binding_evidence_choice", "instance_ids": [ "FA-HG-EV1", "FA-HG-EV2", "FA-HG-EV3" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "1-10" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "binding_design_choice", "name": "客体设计选择（实测证据）", "ability": "design", "status": "pilot", "priority": 1, "concept": "从参考客体出发，在四个候选替换中选出唯一一个使实测结合增强至少给定幅度的；四个候选均有实测值，且都离阈值 ≥ 0.3 kcal/mol。", "domains": [ "chemistry" ], "input_scale": "主体约 1.8 nm", "reasoning_scale": "主客体复合物", "scale_cells": [ "1-10" ], "scale_limits": "结合强弱取决于电荷、尺寸匹配、溶剂化等，不能单凭几何算出；几何附件只用于识别候选客体。", "inputs": "WP6 主体与四个候选客体 XYZ。", "physical_conditions": "与主客体结合证据题相同：ITC，1× PBS，pH 7.40，298.15 K，1:1。", "output_forms": [ "choice" ], "source_basis": "SAMPL9 WP6（MIT）", "answer_method": "实测 ΔG 与阈值比较。", "distractor_mechanisms": [ "improves_but_below_target", "weakens_binding", "dication_not_sufficient" ], "four_choice_validation": "恰好一个候选达到目标；阈值附近无候选。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/hostguest.py", "checkers/hostguest.py", "tools/build_hostguest_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现。", "gap": "2 题。", "engine_families": [ "binding_design_choice" ], "independent_checker": "checkers/hostguest.py:binding_design_choice", "instance_ids": [ "FA-HG-DES1", "FA-HG-DES2" ], "instance_count": 2, "independent_pass": 2, "instance_failures": 0, "instance_reasoning_cells": [ "1-10" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "qdot_optics", "name": "CdSe 量子点尺寸与光学（经验曲线）", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由纳米晶模型坐标确定粒径（最大原子间距），用 Yu et al. (2003) 的经验曲线推断第一激子吸收峰、摩尔消光系数，或由吸光度求浓度。", "domains": [ "quantum" ], "input_scale": "整个纳米晶（2.8–5.0 nm）", "reasoning_scale": "纳米晶最大原子间距", "scale_cells": [ "1-10" ], "scale_limits": "经验曲线只在拟合尺寸范围内可靠；模型无配体。", "inputs": "闪锌矿 CdSe 纳米晶模型 XYZ（本项目构建）。", "physical_conditions": "题干给出经验多项式、适用范围与消光定律。", "output_forms": [ "choice", "numeric" ], "source_basis": "Yu, Qu, Guo, Peng, Chem. Mater. 15, 2854 (2003)：CdSe 尺寸多项式与 ε = 5857 D^2.65", "answer_method": "二分法反解 D(λ)；Beer–Lambert。", "distractor_mechanisms": [ "bounding_box_diagonal", "two_rg", "radius_as_diameter", "diameter_in_angstrom", "cubic_exponent", "exponent_as_factor", "path_in_mm" ], "four_choice_validation": "与正确值 > 3 nm、0.1×10⁵ 或 0.05 µM；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/quantum_dot.py", "checkers/quantum_dot.py", "tools/build_qdot_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现。", "gap": "5 个模型粒子。", "engine_families": [ "qdot_optics" ], "independent_checker": "checkers/quantum_dot.py:qdot_quantity", "instance_ids": [ "FA-QD-PEAK-C", "FA-QD-EPS-E", "FA-QD-CONC-B" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "1-10" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "qdot_design", "name": "量子点粒径选择设计", "ability": "design", "status": "pilot", "priority": 1, "concept": "在四个纳米晶模型中选一个，使第一激子吸收峰最接近目标，或选吸收峰不短于阈值的最小粒子（推断的逆问题）。", "domains": [ "quantum" ], "input_scale": "纳米晶 2.8–5.0 nm", "reasoning_scale": "候选最大粒径", "scale_cells": [ "1-10" ], "scale_limits": "经验曲线只在拟合尺寸范围内可靠；模型无配体。", "inputs": "四个 CdSe 纳米晶模型 XYZ。", "physical_conditions": "题干给出 Yu 2003 的经验多项式、适用范围与 Model conditions（与 CdSe 量子点光学题相同）。", "output_forms": [ "choice" ], "source_basis": "Yu et al. (2003) 经验曲线", "answer_method": "逐个候选反解吸收峰，按目标规则选择。", "distractor_mechanisms": [ "radius_as_diameter", "two_rg", "bounding_box_diagonal", "passes_if_bounding_box_diagonal", "meets_but_larger", "farther_from_target" ], "four_choice_validation": "恰好一个候选满足目标；领先幅度或阈值距离足够。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/quantum_dot.py", "checkers/quantum_dot.py", "tools/build_qdot_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现。", "gap": "2 题。", "engine_families": [ "qdot_design" ], "independent_checker": "checkers/quantum_dot.py:qdot_design", "instance_ids": [ "FA-QD-DES-590", "FA-QD-DES-565" ], "instance_count": 2, "independent_pass": 2, "instance_failures": 0, "instance_reasoning_cells": [ "1-10" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "diffraction_design", "name": "衍射设计：选晶体、选谱线或定波长", "ability": "design", "status": "pilot", "priority": 1, "concept": "first_diffraction_peak 的逆问题：在四种晶体中选首个非零反射最接近目标 2θ 者；为一种晶体在四条 X 射线谱线中选最接近目标角者；或求使首个非零反射落在目标角的波长。", "domains": [ "materials" ], "input_scale": "晶胞", "reasoning_scale": "首个非零反射的晶面间距 d", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "不含原子散射因子角度依赖与吸收。", "inputs": "COD 晶体（NaCl、MgO、CsCl、金刚石、MOF-5、SALEM-2）的整胞分数坐标表；谱线波长在题干给出。", "physical_conditions": "运动学结构因子、原子序数权重；一级 Bragg 衍射。", "output_forms": [ "choice" ], "source_basis": "COD 晶体结构（公有领域）", "answer_method": "从全部原子坐标求第一个 |F| ≠ 0 的反射，再按目标规则选择或反解。", "distractor_mechanisms": [ "lowest_index_ignoring_extinction", "theta_not_two_theta", "assumed_simple_cubic", "d_equals_a", "two_theta_as_theta", "missing_factor_two", "second_allowed", "farther_from_target" ], "four_choice_validation": "选择题：胜者领先 ≥ 1.5 倍且 ≥ 0.5°；数值题：与正确值 > 0.05 Å。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/diffraction_design.py", "checkers/design_extra.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现。", "gap": "6 题。", "engine_families": [ "diffraction_design" ], "independent_checker": "checkers/design_extra.py:diffraction_design", "instance_ids": [ "FA-XD-CRYSTAL-CU", "FA-XD-CRYSTAL-MO", "FA-XD-ANODE-NACL", "FA-XD-ANODE-MOF5", "FA-XD-ANODE-SALEM2", "FA-XD-WAVE-MOF5" ], "instance_count": 6, "independent_pass": 6, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1", "1-10" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "disulfide_design", "name": "二硫键工程位点选择", "ability": "design", "status": "pilot", "priority": 1, "concept": "蛋白工程：在真实蛋白的四个残基对中，选唯一满足题给 Cα–Cα 与 Cβ–Cβ 距离窗口、可突变为半胱氨酸形成二硫键的一对；其余每对至少偏离一个窗口 0.2 Å。", "domains": [ "biology" ], "input_scale": "涉及残基", "reasoning_scale": "候选残基对中最大的 Cα–Cα 距离（< 1 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "只按几何窗口筛选。", "inputs": "PDB ATOM 片段（涉及的残基，逐行照抄）。", "physical_conditions": "题给几何筛选窗口（含端点）。", "output_forms": [ "choice" ], "source_basis": "RCSB PDB 1UBQ、6LYZ（仓库已有）", "answer_method": "Cα–Cα、Cβ–Cβ 距离与窗口比较；恰好一对满足。", "distractor_mechanisms": [ "calpha_only", "cbeta_only", "both_out" ], "four_choice_validation": "恰好一对满足；其余偏离窗口 ≥ 0.2 Å。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/protein_design.py", "checkers/design_extra.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现。", "gap": "3 题。", "engine_families": [ "disulfide_design" ], "independent_checker": "checkers/design_extra.py:disulfide_design", "instance_ids": [ "FA-DS-1UBQ-A", "FA-DS-1UBQ-B", "FA-DS-6LYZ", "EN-1A6M-DS-2-6", "EN-1FHA-DS-6-9", "EN-12E8-DS-7-21", "EN-1HHK-DS-5-168" ], "instance_count": 7, "independent_pass": 7, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "fret_design", "name": "FRET 标记位点设计", "ability": "design", "status": "pilot", "priority": 1, "concept": "在真实蛋白或 tRNA 的四个候选标记对中，选点偶极 FRET 效率（题给 R0）最接近目标的一对；推断的逆问题。", "domains": [ "biology" ], "input_scale": "涉及残基", "reasoning_scale": "候选标记对中最大的点间距离（1–10 nm）", "scale_cells": [ "1-10" ], "scale_limits": "不含染料连接臂与构象涨落。", "inputs": "PDB ATOM 片段（涉及的残基或核苷酸，逐行照抄）。", "physical_conditions": "点偶极 Förster 定律；R0 为题设参数。", "output_forms": [ "choice" ], "source_basis": "RCSB PDB 1EHZ、1UBQ、6LYZ（仓库已有）", "answer_method": "E = 1/(1+(r/R0)^6)，选最接近目标者；领先幅度检查。", "distractor_mechanisms": [ "complement", "inverse_square", "farther_from_target" ], "four_choice_validation": "胜者领先 ≥ 1.5 倍且 ≥ 0.05。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/protein_design.py", "checkers/design_extra.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现。", "gap": "3 题。", "engine_families": [ "fret_design" ], "independent_checker": "checkers/design_extra.py:fret_design", "instance_ids": [ "FA-FD-1EHZ", "FA-FD-1UBQ", "FA-FD-6LYZ", "EN-1A6M-FD-25-1-18", "EN-1FHA-FD-25-5-42", "EN-12E8-FD-25-1-11", "EN-1HHK-FD-25-1-16" ], "instance_count": 7, "independent_pass": 7, "instance_failures": 0, "instance_reasoning_cells": [ "1-10" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "vortex_geometry", "name": "超导涡旋晶格几何（间距 / 密度）", "ability": "perception", "status": "pilot", "priority": 1, "concept": "从涡旋中心图识别 Abrikosov 涡旋晶格：晶格间距与涡旋面密度。", "domains": [ "quantum" ], "input_scale": "视场约 3.5–9 μm", "reasoning_scale": "涡旋晶格间距 a（356–855 nm）", "scale_cells": [ "100-1000" ], "scale_limits": "按论文实测磁场或实测晶格常数构建的模型；不含钉扎无序；论文指出 188.7 G 时单个涡旋已难分辨。", "inputs": "涡旋中心位置图（nm，每行 \"V x y\"，约 97 个，确定性旋转）；设计题给四张不同磁场的图。", "physical_conditions": "理想三角涡旋晶格，每个涡旋一个磁通量子 h/2e；题干写明 Model conditions。", "output_forms": [ "choice", "numeric" ], "source_basis": "Cottet et al., arXiv:1401.4062（MgB2，低温洛伦兹透射电镜，四个实测磁场）；Schaefermeier et al., arXiv:2602.13060（BSCCO，NV 扫描磁强计，a ≈ 0.855 μm）", "answer_method": "中心涡旋 6 个等距近邻；与论文参数一致。", "distractor_mechanisms": [ "row_spacing", "second_neighbour", "double_spacing", "square_cell", "half_density", "circle_cell" ], "four_choice_validation": "与正确值 > 10 nm 或 0.15 /μm²；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/vortex.py", "checkers/mesoscale.py", "tools/build_vortex_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文参数确定性生成。", "gap": "MgB2 四个磁场 + BSCCO 一个样品。", "engine_families": [ "vortex_geometry" ], "independent_checker": "checkers/mesoscale.py:vortex_quantity", "instance_ids": [ "FA-VX-SPACING-561", "FA-VX-DENSITY-1173", "FA-VX-SPACING-BSCCO" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "vortex_inference", "name": "超导涡旋晶格推断（磁场 / 磁通量子 / 中子散射峰）", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由涡旋图推断磁感应强度（每个涡旋一个 h/2e）、每个涡旋携带的磁通，以及小角中子散射首个 Bragg 峰。", "domains": [ "quantum" ], "input_scale": "视场约 3.5–9 μm", "reasoning_scale": "涡旋晶格间距 a（356–855 nm）", "scale_cells": [ "100-1000" ], "scale_limits": "按论文实测磁场或实测晶格常数构建的模型；不含钉扎无序；论文指出 188.7 G 时单个涡旋已难分辨。", "inputs": "涡旋中心位置图（nm，每行 \"V x y\"，约 97 个，确定性旋转）；设计题给四张不同磁场的图。", "physical_conditions": "理想三角涡旋晶格，每个涡旋一个磁通量子 h/2e；题干写明 Model conditions。", "output_forms": [ "choice", "numeric" ], "source_basis": "Cottet et al., arXiv:1401.4062（MgB2，低温洛伦兹透射电镜，四个实测磁场）；Schaefermeier et al., arXiv:2602.13060（BSCCO，NV 扫描磁强计，a ≈ 0.855 μm）", "answer_method": "B = Φ0/A_cell；Φ = B·A_cell；q10 = 4π/(√3 a)。", "distractor_mechanisms": [ "square_lattice", "flux_h_over_e", "flux_h_over_4e", "gauss_mT_slip", "second_neighbour_as_spacing", "h_over_e", "square_cell", "triangle_cell", "hbar_over_2e", "two_pi_over_a", "missing_two_pi", "second_order_peak", "square_lattice_q" ], "four_choice_validation": "与正确值 > 0.15 mT、0.15×10⁻¹⁵ Wb 或 0.5 μm⁻¹；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/vortex.py", "checkers/mesoscale.py", "tools/build_vortex_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文参数确定性生成。", "gap": "与涡旋晶格几何相同：MgB2 四个磁场 + BSCCO 一个样品。", "engine_families": [ "vortex_inference" ], "independent_checker": "checkers/mesoscale.py:vortex_quantity", "instance_ids": [ "FA-VX-FIELD-BSCCO", "FA-VX-FIELD-508", "FA-VX-FLUX-1887", "FA-VX-SANS-1173" ], "instance_count": 4, "independent_pass": 4, "instance_failures": 0, "instance_reasoning_cells": [ "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "vortex_design", "name": "涡旋晶格磁场选择设计", "ability": "design", "status": "pilot", "priority": 1, "concept": "在论文实际施加的四个磁场中选一个，使涡旋间距或中子散射峰最接近目标，或在密度要求下磁场最低。", "domains": [ "quantum" ], "input_scale": "视场约 3.5–9 μm", "reasoning_scale": "涡旋晶格间距 a（356–855 nm）", "scale_cells": [ "100-1000" ], "scale_limits": "按论文实测磁场或实测晶格常数构建的模型；不含钉扎无序；论文指出 188.7 G 时单个涡旋已难分辨。", "inputs": "涡旋中心位置图（nm，每行 \"V x y\"，约 97 个，确定性旋转）；设计题给四张不同磁场的图。", "physical_conditions": "理想三角涡旋晶格，每个涡旋一个磁通量子 h/2e；题干写明 Model conditions。", "output_forms": [ "choice" ], "source_basis": "Cottet et al., arXiv:1401.4062（MgB2，低温洛伦兹透射电镜，四个实测磁场）；Schaefermeier et al., arXiv:2602.13060（BSCCO，NV 扫描磁强计，a ≈ 0.855 μm）", "answer_method": "逐张图计算晶胞，按目标规则选择；领先幅度检查。", "distractor_mechanisms": [ "row_spacing", "second_neighbour", "two_pi_over_a", "second_order_peak", "passes_if_square_cell", "meets_but_higher_field", "too_sparse", "farther_from_target" ], "four_choice_validation": "恰好一个候选满足目标；每个错误选项注明误选原因。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/vortex.py", "checkers/mesoscale.py", "tools/build_vortex_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文参数确定性生成。", "gap": "3 题。", "engine_families": [ "vortex_design" ], "independent_checker": "checkers/mesoscale.py:vortex_design", "instance_ids": [ "FA-VX-DES-SPACING", "FA-VX-DES-SANS", "FA-VX-DES-DENSITY" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "cholesteric_geometry", "name": "胆甾液晶螺旋几何（螺距 / 手性）", "ability": "perception", "status": "pilot", "priority": 1, "concept": "从指向矢场识别胆甾螺旋的螺距（转 360° 的距离）与手性（右手定则）。", "domains": [ "chemistry" ], "input_scale": "1.5 μm 指向矢场", "reasoning_scale": "螺距 p（274–423 nm）", "scale_cells": [ "100-1000" ], "scale_limits": "按论文参数构建的模型；模型反射与实测相差约 1–2%。", "inputs": "胆甾液晶指向矢场（每行 \"site x y z nx ny nz\"，每 5 nm 一点，1.5 μm，任意取向）；设计题给四份。", "physical_conditions": "理想均匀扭曲的螺旋（平面 Grandjean 织构）；λc = ⟨n⟩p，Δλ = Δn·p，p = 1/(HTP·c)；题干写明 Model conditions。", "output_forms": [ "choice", "numeric" ], "source_basis": "Shaban et al., Materials 17, 5080 (2024)（E44 + R5011/S5011，HTP 107.5 μm⁻¹）；Dinc et al., Org. Chem. Front. (2024)（反应型异山梨醇掺杂剂涂层）", "answer_method": "每步转角求螺距；(n_i × n_i+1)·前进方向 的符号定手性。", "distractor_mechanisms": [ "half_pitch", "double_pitch", "quarter_pitch", "reflection_wavelength", "opposite_handedness", "both_wrong" ], "four_choice_validation": "数值与正确值 > 10 nm；手性题四项为手性 × {p, p/2}。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/cholesteric.py", "checkers/mesoscale.py", "tools/build_cholesteric_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文参数确定性生成。", "gap": "Shaban 两种掺杂 + Dinc 两种涂层 + 模型浓度。", "engine_families": [ "cholesteric_geometry" ], "independent_checker": "checkers/mesoscale.py:cholesteric_quantity", "instance_ids": [ "FA-CH-PITCH-R5011", "FA-CH-HAND-S5011", "FA-CH-PITCH-LRCD" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "cholesteric_optics", "name": "胆甾液晶选择反射与扭曲力（推断）", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由螺距与题给折射率推断反射中心波长和带宽，或由螺距与浓度推断手性掺杂剂的螺旋扭曲力 HTP。", "domains": [ "chemistry" ], "input_scale": "1.5 μm 指向矢场", "reasoning_scale": "螺距 p（274–423 nm）", "scale_cells": [ "100-1000" ], "scale_limits": "按论文参数构建的模型；模型反射与实测相差约 1–2%。", "inputs": "胆甾液晶指向矢场（每行 \"site x y z nx ny nz\"，每 5 nm 一点，1.5 μm，任意取向）；设计题给四份。", "physical_conditions": "理想均匀扭曲的螺旋（平面 Grandjean 织构）；λc = ⟨n⟩p，Δλ = Δn·p，p = 1/(HTP·c)；题干写明 Model conditions。", "output_forms": [ "choice", "numeric" ], "source_basis": "Shaban et al., Materials 17, 5080 (2024)（E44 + R5011/S5011，HTP 107.5 μm⁻¹）；Dinc et al., Org. Chem. Front. (2024)（反应型异山梨醇掺杂剂涂层）", "answer_method": "λc = ⟨n⟩p；Δλ = Δn·p；HTP = 1/(p c)。", "distractor_mechanisms": [ "extraordinary_index", "ordinary_index", "half_pitch", "no_index", "double_pitch", "central_wavelength", "relative_birefringence", "half_birefringence", "percent_not_fraction", "reflection_for_pitch", "nm_um_slip" ], "four_choice_validation": "与正确值 > 5–8 nm 或 5 μm⁻¹；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/cholesteric.py", "checkers/mesoscale.py", "tools/build_cholesteric_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文参数确定性生成。", "gap": "与胆甾液晶几何相同：Shaban 两种掺杂 + Dinc 两种涂层 + 模型浓度。", "engine_families": [ "cholesteric_optics" ], "independent_checker": "checkers/mesoscale.py:cholesteric_quantity", "instance_ids": [ "FA-CH-LAMBDA-R5011", "FA-CH-BAND-R5011", "FA-CH-HTP-S5011", "FA-CH-LAMBDA-RRCD" ], "instance_count": 4, "independent_pass": 4, "instance_failures": 0, "instance_reasoning_cells": [ "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "cholesteric_design", "name": "胆甾液晶结构色设计", "ability": "design", "status": "pilot", "priority": 1, "concept": "在四份膜中选择：反射最接近目标色、左手且最接近目标、或在反射上限内掺杂量最少。", "domains": [ "chemistry" ], "input_scale": "1.5 μm 指向矢场", "reasoning_scale": "螺距 p（274–423 nm）", "scale_cells": [ "100-1000" ], "scale_limits": "按论文参数构建的模型；模型反射与实测相差约 1–2%。", "inputs": "胆甾液晶指向矢场（每行 \"site x y z nx ny nz\"，每 5 nm 一点，1.5 μm，任意取向）；设计题给四份。", "physical_conditions": "理想均匀扭曲的螺旋（平面 Grandjean 织构）；λc = ⟨n⟩p，Δλ = Δn·p，p = 1/(HTP·c)；题干写明 Model conditions。", "output_forms": [ "choice" ], "source_basis": "Shaban et al., Materials 17, 5080 (2024)（E44 + R5011/S5011，HTP 107.5 μm⁻¹）；Dinc et al., Org. Chem. Front. (2024)（反应型异山梨醇掺杂剂涂层）", "answer_method": "逐份拟合螺距与手性，按目标规则选择。", "distractor_mechanisms": [ "extraordinary_index", "half_pitch", "right_handed_closer", "right_handed", "reflects_too_red", "meets_but_more_dopant", "farther_from_target" ], "four_choice_validation": "恰好一个候选满足目标；每个错误选项注明误选原因。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/cholesteric.py", "checkers/mesoscale.py", "tools/build_cholesteric_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；资产由论文参数确定性生成。", "gap": "3 题。", "engine_families": [ "cholesteric_design" ], "independent_checker": "checkers/mesoscale.py:cholesteric_design", "instance_ids": [ "FA-CH-DES-550", "FA-CH-DES-LEFT600", "FA-CH-DES-BELOW620" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "lattice_plane_geometry", "name": "晶面间距", "ability": "perception", "status": "existing_legacy", "priority": 5, "concept": "立方晶胞 (hkl) 晶面间距 d = a/√(h²+k²+l²)。", "domains": [ "materials" ], "input_scale": "晶胞边长", "reasoning_scale": "晶面间距", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "公式题，坐标仅提供晶格参数。", "inputs": "晶胞参数、晶面指数。", "physical_conditions": "立方晶系。", "output_forms": [ "numeric" ], "source_basis": "COD", "answer_method": "立方晶系面间距解析公式 d = a/√(h²+k²+l²)。", "distractor_mechanisms": [ "a/(h+k+l)", "a·√(h²+k²+l²)", "diagonal_confusion" ], "four_choice_validation": "未实现（旧数值短答案，未改为四选一）。", "legacy_families": [ "cubic_plane_geometry" ], "legacy_qa": [ "MNP003", "MNP008", "MNP009" ], "code": [ "tools/recompute_materials_batch.py" ], "tests": [ "tests/test_materials_batch.py" ], "builder_modules": "旧 pipeline。", "gap": "优先级低：易退化为公式题。", "engine_families": [], "independent_checker": "", "instance_ids": [], "instance_count": 0, "independent_pass": 0, "instance_failures": 0, "instance_reasoning_cells": [], "legacy_decisions": { "retain": 3 }, "legacy_choice_v01": 0 }, { "id": "biopolymer_path_geometry", "name": "生物大分子路径与局部几何", "ability": "perception", "status": "existing_legacy", "priority": 6, "concept": "DNA/RNA 骨架路径、反密码子环、螺旋局部间距、催化裂隙等。", "domains": [ "biology" ], "input_scale": "整条链（1–10 nm）", "reasoning_scale": "路径或局部片段", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "各题定义不一，需统一标记原子。", "inputs": "PDB、链与残基选择。", "physical_conditions": "单一实验结构模型（PDB 主构象）。", "output_forms": [ "numeric" ], "source_basis": "PDB 1BNA、1EHZ、1UBQ、6LYZ", "answer_method": "旧专用脚本（tools/recompute_bio_batch*.py）。", "distractor_mechanisms": [ "wrong_marker_atom", "chord_vs_path", "neighbor_residue" ], "four_choice_validation": "未实现（旧数值短答案，未改为四选一）。", "legacy_families": [ "dna_marker_path", "rna_backbone_path", "anticodon_loop_geometry", "helix_local_spacing", "donor_orientation_screen", "protein_extent_and_catalytic_cleft" ], "legacy_qa": [ "BNP003", "BNP005", "BNP006", "BNP009", "BNP011", "HBP002", "BNP004" ], "code": [ "tools/recompute_bio_batch.py", "tools/recompute_bio_batch2.py" ], "tests": [ "tests/test_bio_geometry.py" ], "builder_modules": "旧 pipeline。", "gap": "需要先拆成若干可复用子族（路径长度、弦长、局部间距）。", "engine_families": [], "independent_checker": "", "instance_ids": [], "instance_count": 0, "independent_pass": 0, "instance_failures": 0, "instance_reasoning_cells": [], "legacy_decisions": { "retain": 3, "revise": 4 }, "legacy_choice_v01": 0 }, { "id": "stereo_relationship", "name": "立体关系判别", "ability": "perception", "status": "pilot", "priority": 1, "concept": "判断结构 X 相对 S0 是同一化合物（仅构象/取向不同）、对映体、非对映体还是构造异构体；需要在三维中区分旋转与镜像、构象改变与构型改变（会议 2 立体化学梯子第一级）。", "domains": [ "chemistry" ], "input_scale": "两个分子", "reasoning_scale": "整分子立体构型（0.4–0.8 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "立体中心按第一配位层判定；与数据集标签不一致的实例拒绝。", "inputs": "S0.xyz 与 X.xyz（Å）；X 经刚性旋转平移，原子顺序可能不同。", "physical_conditions": "单键旋转可互变的结构视为同一化合物。", "output_forms": [ "choice" ], "source_basis": "QM7-X 立体异构体与构象（CC BY 4.0），15 个分子的全部优化构象", "answer_method": "共价半径成键图 + 图不变量判定键连；四配位中心、锥形 S/P、氮杂环丙烷 N 的手性符号与双键顺反描述符判定构型关系。", "distractor_mechanisms": [ "conformation_mistaken_for_configuration", "rotation_mistaken_for_reflection", "partial_vs_full_inversion", "same_formula_assumed_same_bonding" ], "four_choice_validation": "四个固定类别恰好一个成立；批次内每类 2 例以平衡位置；独立检查器用自有实现复核。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/stereo.py", "task_families/kit.py", "checkers/stereo.py", "tools/build_family_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现 M0–M6；所有判定与数据集立体异构体标签交叉核对，不一致即拒绝。", "gap": "R/S、E/Z 命名需 CIP 优先级；可在获准安装 RDKit 后加入。", "engine_families": [ "stereo_relationship" ], "independent_checker": "checkers/stereo.py:stereo_relationship", "instance_ids": [ "FA-SREL-7029-SAME", "FA-SREL-7052-SAME", "FA-SREL-7208-ENAN", "FA-SREL-7111-ENAN", "FA-SREL-7110-DIAS", "FA-SREL-7043-DIAS", "FA-SREL-7082-CONS", "FA-SREL-7033-CONS" ], "instance_count": 8, "independent_pass": 8, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "stereo_property_inference", "name": "立体异构体性质推断", "ability": "inference", "status": "pilot", "priority": 1, "concept": "同一分子的两个或多个立体异构体，由三维构型推断哪一个的指定性质（偶极、能量或文献报道的活性）更高或落在哪个区间。", "domains": [ "chemistry" ], "input_scale": "分子", "reasoning_scale": "立体中心周围与整分子取向", "scale_cells": [ "0.1-1" ], "scale_limits": "构型差异须由几何决定；性质差须超过阈值。", "inputs": "S0.xyz 与 X.xyz：同一分子两个非对映体各自的数据集最低能构象；X 刚性旋转。", "physical_conditions": "性质为数据集计算值或论文实测值，题干写明来源类型与条件。", "output_forms": [ "choice", "numeric" ], "source_basis": "QM7-X 立体异构体与构象（CC BY 4.0），15 个分子的全部优化构象", "answer_method": "由数据集 PBE0+MBD 能量与偶极得到 ΔE、Δμ 的符号；两者均须超过阈值（1 kcal/mol、0.10 e·Å）。", "distractor_mechanisms": [ "energy_direction_reversed", "dipole_direction_reversed", "both_reversed" ], "four_choice_validation": "四个陈述覆盖符号组合，恰好一个成立；位置按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/stereo.py", "task_families/kit.py", "checkers/stereo.py", "tools/build_family_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现 M0–M6；所有判定与数据集立体异构体标签交叉核对，不一致即拒绝。", "gap": "仅 4 对满足阈值的非对映体；更多需下载 QM7-X 其他分卷。", "engine_families": [ "stereo_property_inference" ], "independent_checker": "checkers/stereo.py:stereo_property_inference", "instance_ids": [ "FA-SPROP-7082", "FA-SPROP-7089", "FA-SPROP-7110", "FA-SPROP-7114" ], "instance_count": 4, "independent_pass": 4, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "stereo_design_selection", "name": "立体构型设计选择", "ability": "design", "status": "pilot", "priority": 1, "concept": "给定 S0 与目标（能量降低 ≥1 kcal/mol），只允许改变构型（立体中心手性或双键顺反），从四个候选中选唯一满足者；陷阱：镜像（构型变但能量不变）、只改构象（能量更低但不允许）、构造异构体（键连不同）。", "domains": [ "chemistry" ], "input_scale": "分子", "reasoning_scale": "立体中心与整分子", "scale_cells": [ "0.1-1" ], "scale_limits": "候选须满足约束且互不等价（排除镜像/构象重复）。", "inputs": "S0 与四个候选 XYZ（均刚性旋转；构造异构体原子顺序不同）；性质证据在审核侧。", "physical_conditions": "同 stereo_property_inference。", "output_forms": [ "choice" ], "source_basis": "QM7-X 立体异构体与构象（CC BY 4.0），15 个分子的全部优化构象", "answer_method": "逐项判定与 S0 的关系（是否只改构型）+ 能量证据（数据集值；镜像按反演对称等于 S0）。", "distractor_mechanisms": [ "mirror_image_no_energy_change", "conformation_only_change", "constitution_changed" ], "four_choice_validation": "恰好一项同时满足约束与目标；独立检查器复核每一项。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/stereo.py", "task_families/kit.py", "checkers/stereo.py", "tools/build_family_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现 M0–M6；所有判定与数据集立体异构体标签交叉核对，不一致即拒绝。", "gap": "3 个分子；目标暂只有降低能量。", "engine_families": [ "stereo_design_selection" ], "independent_checker": "checkers/stereo.py:stereo_design_selection", "instance_ids": [ "FA-SDES-7029", "FA-SDES-7033", "FA-SDES-7110" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "force_path_derivative", "name": "受力沿路径的一阶能量变化", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由 F = −∂E/∂r 与两结构间位移推断沿指定路径在 B 处的能量方向导数：几何（位移方向）+ 物理条件（受力）→ 性质。", "domains": [ "quantum" ], "input_scale": "分子最大间距", "reasoning_scale": "全部原子参与（分子尺度 0.4–0.8 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "局部方向导数，不是 E(B)−E(A)，不证明极小点。", "inputs": "A、B 两个 XYZ（同原子序、同坐标系）+ B 处全原子受力表。", "physical_conditions": "PBE0+MBD 总力；直线路径；方向在题干中给定（A→B 或 B→A）。", "output_forms": [ "numeric", "choice" ], "source_basis": "QM7-X（CC BY 4.0）", "answer_method": "Decimal 与浮点投影求和一致。", "distractor_mechanisms": [ "sign_convention", "heavy_atoms_only", "trapezoid_energy_estimate", "abs_projection_sum", "magnitude_product" ], "four_choice_validation": "与正确值 > 0.05 eV；升序排列，秩按批次平衡；路径方向在实例间变化以削弱“选正数”捷径。", "legacy_families": [ "force_displacement_projection", "force_displacement_projection_comparison" ], "legacy_qa": [ "QNP001", "QNI001", "QNI002", "QNI003" ], "code": [ "task_families/force_path.py", "builder_modules/m4_construction/force_projection.py", "tools/build_family_assets.py", "checkers/families.py" ], "tests": [ "tests/test_task_families.py", "tests/test_force_projection.py", "tests/test_checkers.py" ], "builder_modules": "M1 资产由 tools/build_family_assets.py 从 QM7-X 提取并与已发布旧题坐标/受力逐项核对。", "gap": "可扩到“受力是否指向降低能量的方向”的结论选择题。", "engine_families": [ "force_path_derivative" ], "independent_checker": "checkers/families.py:force_path", "instance_ids": [ "FA-FP-QM7X-7001", "FA-FP-QM7X-7002", "FA-FP-QM7X-7035", "FA-FP-QM7X-7040-D1", "FA-FP-QM7X-7040-D2" ], "instance_count": 5, "independent_pass": 5, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": { "retain": 4 }, "legacy_choice_v01": 0 }, { "id": "kinematic_extinction", "name": "晶体衍射消光判别", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由晶胞内全部原子坐标与运动学结构因子推断哪一个反射强度为零；优先选择由基元/滑移而非仅中心化导致的消光。", "domains": [ "materials" ], "input_scale": "晶胞边长", "reasoning_scale": "晶胞边长（相位求和跨越整个晶胞）", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "弱反射（|F|/F000 < 0.05）不作选项；非立方时 (h k l) 取 0–4，并排除彼此对称等价（同 d、同 |F|）的干扰项，避免一次排除两项。", "inputs": "常规晶胞全部原子分数坐标表 + 晶胞参数 + 原子序数权重。", "physical_conditions": "运动学近似；零角度原子序数权重；忽略热振动。", "output_forms": [ "choice" ], "source_basis": "COD（NaCl、金刚石、SALEM-2、MOF-5；金红石、板钛矿、ZnO、α-石英）", "answer_method": "对学生看到的同一坐标表求 F(hkl)；|F|/F000 ≤ 1e-6 判为消光。非立方晶胞以“换轴误用消光条件”（指数换轴后的反射才消光）代替中心化规则作为错误机制。", "distractor_mechanisms": [ "wrong_rule:I", "wrong_rule:F", "wrong_rule:F+d", "identical_atom_assumption", "weak_not_absent", "axis_confusion" ], "four_choice_validation": "恰好一项消光、三项 ≥ 0.05 F000；另记录 P/I/F/F+d 规则能否单独答对。", "legacy_families": [ "basis_phase_extinction", "bcc_superlattice_extinction", "centering_type_comparison", "sublattice_scattering_contrast" ], "legacy_qa": [ "MNI002", "MNI003", "MNI005", "MNI008", "MNI009", "MNI010" ], "code": [ "task_families/extinction.py", "tools/recompute_materials_batch.py", "checkers/families.py" ], "tests": [ "tests/test_task_families.py", "tests/test_materials_batch.py", "tests/test_checkers.py" ], "builder_modules": "M1 直接读取已发布 CIF/XYZ；M5 记录规则捷径。", "gap": "旧 MOF-5 题来源误记为 COD 4118891（应为 1516287），旧“ZIF-8”实为未取代咪唑的 SALEM-2；新族已更正，旧目录待修。", "engine_families": [ "kinematic_extinction" ], "independent_checker": "checkers/families.py:extinction", "instance_ids": [ "FA-EXT-NACL", "FA-EXT-DIAMOND", "FA-EXT-SALEM2", "FA-EXT-MOF5", "FA-EXT-RUTILE", "FA-EXT-BROOKITE", "FA-EXT-ZNO", "FA-EXT-SI" ], "instance_count": 8, "independent_pass": 8, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1", "1-10" ], "legacy_decisions": { "retain": 2, "revise": 4 }, "legacy_choice_v01": 0 }, { "id": "bragg_angle_inference", "name": "首个衍射峰（布拉格角）", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由晶胞全部原子坐标判断哪一个反射最先出现（跳过消光反射），再用布拉格定律求其 2θ；不同于只代入给定 d 的公式题。", "domains": [ "materials" ], "input_scale": "晶胞", "reasoning_scale": "晶面间距", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "只需晶胞与原子位置；若只用晶胞边长而不算结构因子，会误选被消光的反射（这正是一种错误机制）。", "inputs": "常规晶胞全部原子分数坐标表 + 晶胞参数；原子序数权重；波长。", "physical_conditions": "运动学近似、零角度原子序数权重、一级衍射。", "output_forms": [ "numeric", "choice" ], "source_basis": "COD", "answer_method": "立方：h ≥ k ≥ l ≥ 0 的低指数反射；其他晶系：|h|、|k|、|l| ≤ 4 的全部反射（Friedel 对取一个），d 由倒格矢求，取 |F|≠0 的最小 2θ；等价反射同角度视为一个峰。非立方的登记错误包括套用立方公式 d = a/√(h²+k²+l²)。", "distractor_mechanisms": [ "lowest_index_ignoring_extinction", "assumed_lattice_rule", "theta_not_two_theta", "second_allowed_peak", "d_equals_a", "cubic_d_formula", "d_equals_c", "bragg_missing_factor_two", "strongest_not_first", "index_sum_for_root" ], "four_choice_validation": "与正确值 > 0.3°；升序排列，秩按批次轮换。 枚举探针发现正确项多为 C（错误机制几乎都给出更小的角），v0.3.0 加入三种给出更大角的错误（布拉格式漏 2、把最强反射当第一个峰、立方公式把根号写成指数和），并把 d = a 的迷惑程度调为 2；立方晶胞即使经通用读取器也走立方路径。", "legacy_families": [ "bragg_geometry_inference" ], "legacy_qa": [ "MNI001", "MNI004", "MNI006", "MNI007" ], "code": [ "task_families/crystal.py", "checkers/migrated.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；旧 MNI004/006/007 由新版取代（同时更正 MOF-5 来源为 COD 1516287）。", "gap": "未含原子散射因子角度依赖。", "engine_families": [ "first_diffraction_peak" ], "independent_checker": "checkers/migrated.py:first_diffraction_peak", "instance_ids": [ "FA-PEAK-NACL", "FA-PEAK-CSCL", "FA-PEAK-MGO", "FA-PEAK-SALEM2", "FA-PEAK-MOF5", "FA-PEAK-RUTILE", "FA-PEAK-ZNO", "FA-PEAK-QUARTZ", "FA-PEAK-SI", "FA-PEAK-DIAMOND", "FA-PEAK-CU", "FA-PEAK-FE", "FA-PEAK-MG", "FA-PEAK-ALNI", "FA-PEAK-B5H9", "FA-PEAK-K2CUF4", "FA-PEAK-MN2CO10", "FA-PEAK-PARA-I", "FA-PEAK-PARA-II", "FA-PEAK-PENTACENE", "FA-PEAK-HAP", "FA-PEAK-ARAGONITE" ], "instance_count": 22, "independent_pass": 22, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1", "1-10" ], "legacy_decisions": { "retain": 1, "revise": 3 }, "legacy_choice_v01": 0 }, { "id": "scattering_model_inference", "name": "散射/FRET 模型推断", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由结构几何与给定模型参数推断可观测量：FRET 效率（点探针、给定 R0）与 Guinier 近似下的归一化小角散射强度（等权点散射体、给定 q）。题干给出模型输入但不给公式。", "domains": [ "chemistry", "biology" ], "input_scale": "蛋白（1–10 nm）", "reasoning_scale": "全局尺度", "scale_cells": [ "1-10", "10-100" ], "scale_limits": "模型假设需人工审核。暂无数据源的格：chemistry 1-10（规划器不据此给出路线）。", "inputs": "PDB 链 ATOM 片段（FRET）或整分子重原子 XYZ（Guinier）。", "physical_conditions": "Förster 点偶极模型（R0 含取向与光谱因子）；Guinier 近似仅在 qRg ≤ 1.3 出题。", "output_forms": [ "numeric", "choice" ], "source_basis": "PDB（含 7ARQ DNA 折纸约简点表）+ 教科书/论文模型", "answer_method": "FRET：直线距离 + E=1/(1+(r/R0)^6)；Guinier：等权 Rg + exp(−q²Rg²/3)。", "distractor_mechanisms": [ "path_length_distance", "complement", "closest_atom_pair", "r_equals_R0_default", "wrong_acceptor_atom", "missing_one_third", "half_dmax_as_rg", "amplitude_not_intensity", "unit_mixups" ], "four_choice_validation": "与正确值 > 0.02–0.03；升序排列，秩按批次轮换。FRET 可由题设 R0 列表中选取使目标秩可达的值（某些秩只在 E < 0.5 或 E > 0.5 时可达；枚举探针发现只用固定 R0 时正确项总是最大）。 Guinier：q 必须恰为题干显示的两位小数（之前更精细的 q 会使题干与答案不一致）；枚举实例设 strict_rank。", "legacy_families": [ "conditional_guinier_inference", "conditional_fret_inference", "fret_range_mismatch_inference" ], "legacy_qa": [ "BNI001", "BNI002", "BNI004" ], "code": [ "checkers/migrated.py", "task_families/scattering.py", "tools/build_assembly_assets.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "族引擎内等价实现；旧 BNI004 由新版取代。", "gap": "R0 为题设虚拟值，不对应具体染料。", "engine_families": [ "fret_efficiency", "guinier_intensity" ], "independent_checker": "checkers/migrated.py:fret_efficiency, guinier_intensity", "instance_ids": [ "FA-FRET-1EHZ", "FA-FRET-1UBQ", "FA-FRET-6LYZ", "FA-FRET-1BNA", "EN-1A6M-FRET-1-15", "EN-1FHA-FRET-5-22", "EN-12E8-FRET-1-8", "EN-1HHK-FRET-1-7", "FA-GUIN-6LYZ", "FA-GUIN-1EHZ", "FA-GUIN-1BNA", "FA-ASM-GUIN-1AON", "FA-ASM-GUIN-1KX5", "FA-ASM-GUIN-1A34", "FA-ASM-GUIN-7ARQ", "FA-ASM-GUIN-1RYP", "FA-ASM-GUIN-1STM", "FA-GUIN-1UBQ", "FA-GUIN-1CRN" ], "instance_count": 19, "independent_pass": 19, "instance_failures": 0, "instance_reasoning_cells": [ "1-10", "10-100" ], "legacy_decisions": { "retain": 2, "revise": 1 }, "legacy_choice_v01": 0 }, { "id": "evidence_inference_choice", "name": "论文证据支持的物理推断选择", "ability": "inference", "status": "planned", "priority": 3, "concept": "几何特征 + 论文给出的关系 → 性质或结论；证据只给审核方。", "domains": [ "chemistry", "biology" ], "input_scale": "按实例", "reasoning_scale": "按实例", "scale_cells": [ "0.1-1" ], "scale_limits": "未知不能冒充错误。", "inputs": "学生几何条件与选项；审核侧保留论文原文定位、条件及答案依据。", "physical_conditions": "按论文适用范围写入题干。", "output_forms": [ "choice", "numeric" ], "source_basis": "论文（HBI001/HBI002 氢键红移、刚度查表为已有例）", "answer_method": "证据映射 + 数值校验。", "distractor_mechanisms": [ "reverse_trend", "outside_validity_range", "confused_quantity" ], "four_choice_validation": "需证据映射与区间不重叠。", "legacy_families": [ "geometry_redshift_inference", "geometry_conditioned_stiffness_lookup" ], "legacy_qa": [ "HBI001", "HBI002" ], "code": [], "tests": [], "builder_modules": "旧 pipeline M2 证据模块。", "gap": "待建立首个经证据支持的通用模板。 已由 secondary_structure（生物 0.1–1）与 binding_evidence_choice（化学 1–10）部分实现。 2026-10-16：论文“证据 → 结论”已由 evidence_quantity 与 evidence_elimination 实现（HBI001/HBI002 由 CNI027/CNI028 取代）。", "engine_families": [], "independent_checker": "", "instance_ids": [], "instance_count": 0, "independent_pass": 0, "instance_failures": 0, "instance_reasoning_cells": [], "legacy_decisions": { "retain": 2 }, "legacy_choice_v01": 0 }, { "id": "structure_comparison_inference", "name": "结构比较推断", "ability": "inference", "status": "existing_legacy", "priority": 6, "concept": "两结构/两链之间的几何比较及其结论边界。", "domains": [ "biology", "chemistry" ], "input_scale": "按实例", "reasoning_scale": "按实例", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "多题只做算术比较，推断成分弱。", "inputs": "两个结构。", "physical_conditions": "按每道旧题给出的两个结构与比较条件。", "output_forms": [ "numeric", "choice" ], "source_basis": "PDB", "answer_method": "旧专用脚本（见对应旧题的计算记录）。", "distractor_mechanisms": [ "reverse_comparison", "overgeneralization" ], "four_choice_validation": "未实现（旧数值短答案，未改为四选一）。", "legacy_families": [ "cross_chain_gyration_comparison", "cross_protein_gyration_comparison", "disulfide_bond_length_comparison", "helix_regularity_cross_protein", "helix_regularity_inference", "terminal_vs_global_extent", "assembly_extent_arithmetic", "contact_evidence_scope", "substituent_effect_on_distal_geometry" ], "legacy_qa": [ "CNI004", "BNI003", "BNI005", "BNI007", "BNI008", "BNI006", "CNI002", "CNI003", "CNI005", "CNI006", "CNI001" ], "code": [ "tools/recompute_bio_batch2.py" ], "tests": [ "tests/test_bio_geometry2.py" ], "builder_modules": "旧 pipeline。", "gap": "需要先确认哪些比较真正需要推断；否则并入感知族。", "engine_families": [], "independent_checker": "", "instance_ids": [], "instance_count": 0, "independent_pass": 0, "instance_failures": 0, "instance_reasoning_cells": [], "legacy_decisions": { "revise": 11 }, "legacy_choice_v01": 0 }, { "id": "electronic_property_from_geometry", "name": "电子性质（能隙、偶极）", "ability": "inference", "status": "rebuild_needed", "priority": 8, "concept": "旧题直接给出能级/偶极向量，删去几何仍可作答。", "domains": [ "quantum" ], "input_scale": "分子", "reasoning_scale": "无（几何非必要）", "scale_cells": [], "scale_limits": "几何非必要。", "inputs": "OE62/QM7-X 能级列表或偶极矢量。", "physical_conditions": "给定泛函。", "output_forms": [ "numeric" ], "source_basis": "OE62、QM7-X", "answer_method": "算术（旧题）。", "distractor_mechanisms": [], "four_choice_validation": "不适用。", "legacy_families": [ "homo_lumo_gap", "gap_method_comparison", "cross_molecule_gap_comparison", "dipole_magnitude_change" ], "legacy_qa": [ "QNI004", "QNI005", "QNI006", "QNI007", "QNI008", "QNI009" ], "code": [], "tests": [], "builder_modules": "旧 pipeline。", "gap": "重建方向：改为 conformer_target_selection 式的“由结构选性质”题，而不是给出能级。", "engine_families": [], "independent_checker": "", "instance_ids": [], "instance_count": 0, "independent_pass": 0, "instance_failures": 0, "instance_reasoning_cells": [], "legacy_decisions": { "rebuild": 6 }, "legacy_choice_v01": 0 }, { "id": "conformer_target_selection", "name": "构象设计选择", "ability": "design", "status": "pilot", "priority": 1, "concept": "给定起始构象 S0、允许修改（仅单键旋转：原子、键连与立体中心构型不变）和目标性质（最低能量或最大偶极），从四个真实候选构象中选出唯一满足目标者。", "domains": [ "quantum", "chemistry" ], "input_scale": "S0 最大间距", "reasoning_scale": "整分子构象（0.4–0.8 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "计算证据，不是实验构象分布或自由能；不是开放式生成。", "inputs": "S0 与四个候选的 XYZ（Å）；性质证据只在审核侧。", "physical_conditions": "性质为 QM7-X PBE0+MBD 单点值（DFTB3+MBD 优化几何）。", "output_forms": [ "choice" ], "source_basis": "QM7-X（CC BY 4.0）", "answer_method": "逐项检查约束（原子、键连图、立体中心带符号体积、与 S0 不同、偶极矢量一致）；去除同一/镜像构象；最优项领先第二名 ≥ 阈值（能量 0.0434 eV，偶极 0.10 e·Å）。", "distractor_mechanisms": [ "real_candidate_worse_by_computed_property", "compactness_heuristic (选 Rg 最小/最大；选择时优先让正确项不是极值)", "minimal_change_heuristic (选与 S0 最接近者；记录诊断)" ], "four_choice_validation": "四个候选均满足约束；唯一最优；数据集 ID 不出现在学生包；凑不齐则记录失败。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/conformer_design.py", "tools/build_family_assets.py", "checkers/families.py" ], "tests": [ "tests/test_task_families.py", "tests/test_checkers.py" ], "builder_modules": "旧 M3 仍拒绝 design；本族在族引擎内实现独立约束与唯一性检查，未改动旧门槛。", "gap": "论文 S0→S1 优化案例（design_choice）仍缺实例。", "engine_families": [ "conformer_target_selection" ], "independent_checker": "checkers/families.py:conformer_selection", "instance_ids": [ "FA-CONF-7050-max_dipole", "FA-CONF-7206-min_energy", "FA-CONF-7095-min_energy", "FA-CONF-7032-max_dipole" ], "instance_count": 4, "independent_pass": 4, "instance_failures": 1, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "design_choice", "name": "有证据的结构优化选择（论文案例）", "ability": "design", "status": "pilot", "priority": 2, "concept": "从论文中记录在案的结构优化构题：作者为某个目标刻意改了一处结构特征（芳香墙加长、门户取代基加大、孔口收窄），并对每个变体都测了性质。题干给出目标（实测性质的阈值）与四个只用结构描述呈现的候选；实测表留在审核侧，所以这是设计选择，不是查表。", "domains": [ "chemistry", "materials" ], "input_scale": "候选结构（分子夹–染料复合物、大环空腔、框架孔口）", "reasoning_scale": "决定性质的那一处几何（芳香墙长轴 0.70–0.92 nm；C–H···X⁻ 接触 0.31 nm；孔口 0.32–0.80 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "尺度取决定目标性质的那一处几何，不取整个候选分子或整块晶体。", "inputs": "candidates.txt（四个候选的结构描述，逐项转述并带原文位置）；实测性质表与原文位置只在审核侧。", "physical_conditions": "按论文的实测条件写入题干（缓冲液与温度、溶剂与温度、暴露压力与温度）。", "output_forms": [ "choice" ], "source_basis": "2026-10-19 下载的三篇 CC BY 全文（本地 inputs/papers/，不入库）：Vincent 2025（分子夹芳香墙 H1→H2，Table 1 的 14 种染料 × 2 主体）、Rando 2025（竹环脲门户取代基 BnBU/BU1/BU2，Table 1 与 Table S3 的 10 种阴离子）、Chetry 2026（MOF 孔口与 D₂/H₂ 分离，Table 1 与 TDS 结果）。设计卡 docs/assets/families/design/*.json 逐项记录原文位置。", "answer_method": "逐个候选、逐条约束比对实测值；恰好一个候选满足全部约束，且必须是原文选定的那一个。", "distractor_mechanisms": [ "measured_inferior_analogue（同篇论文测过但不达标的同类结构）", "violates_constraint（主性质达标却破坏另一条约束，如最窄孔把容量压到 0.2 mmol g⁻¹）", "host_walls_not_extended / guest_not_extended / neither_extended（只改一侧的 π 延长）", "extended_walls_weaken_this_guest（实测反例：延长墙反而变弱）", "strongest_binder_cue（绝对结合最强但选择性方向错）", "distorted_cavity（取代基一样大，位置不同导致空腔扭曲）", "solvent_change_insufficient（换溶剂不改变几何）", "narrowest_pore_cue / pore_too_wide / interpenetrated_network（孔口几何的三种误判）" ], "four_choice_validation": "四个候选必须对每条受约束的性质都有实测值且带原文位置（“未测过”不会被当成“不行”）；每个候选与每个阈值之间必须有 margin（默认 1.2 倍）的距离；恰好一个候选满足全部约束。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/paper_design.py", "checkers/paper_design.py" ], "tests": [ "tests/test_paper_design.py" ], "builder_modules": "族引擎内实现；设计卡人工转录，候选结构描述与实测值各带原文位置供 L1 对照。", "gap": "第一批 6 道（3 篇论文，每篇两个目标共用一张实测表）。只覆盖化学 C4a/C4b 与材料 M4a，都不是空白子领域；尚未覆盖生物、量子，也未覆盖 C5 反应与催化（检索中未找到既开放获取、又有四个候选的实测性能表的优化论文）。", "engine_families": [ "design_choice" ], "independent_checker": "checkers/paper_design.py:design_choice", "instance_ids": [ "FA-DES-VIN-WALLS", "FA-DES-VIN-MV", "FA-DES-RAN-NITRATE", "FA-DES-RAN-BROMIDE", "FA-DES-CHE-SIEVE", "FA-DES-CHE-FLOW" ], "instance_count": 6, "independent_pass": 6, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "open_structure_design", "name": "开放结构生成", "ability": "design", "status": "blocked", "priority": 10, "concept": "输出新结构；可能有多个有效解。", "domains": [ "chemistry", "materials" ], "input_scale": "按实例", "reasoning_scale": "按实例", "scale_cells": [], "scale_limits": "暂无评价器。", "inputs": "起始结构、可变部分、目标、约束与输出结构格式。", "physical_conditions": "按目标。", "output_forms": [ "structure" ], "source_basis": "—", "answer_method": "需结构合法性、约束和目标评价器。", "distractor_mechanisms": [], "four_choice_validation": "不适用。", "legacy_families": [], "legacy_qa": [], "code": [], "tests": [], "builder_modules": "未开放。", "gap": "保留路线。", "engine_families": [], "independent_checker": "", "instance_ids": [], "instance_count": 0, "independent_pass": 0, "instance_failures": 0, "instance_reasoning_cells": [], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "rotational_line", "name": "转动跃迁频率", "ability": "inference", "status": "pilot", "priority": 1, "concept": "由分子几何预测转动光谱：质心 → 惯量张量 → 主转动惯量 → 转动常数 A ≥ B ≥ C → J = 1 能级 → 1_01、1_11 或 1_10 ← 0_00 跃迁频率。几何决定可观测谱线（量子力学刚性转子）。", "domains": [ "quantum", "chemistry" ], "input_scale": "整个分子（13–56 个原子）", "reasoning_scale": "整分子转动惯量（分子最大原子间距 0.5–1.2 nm）", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "推理需要整个分子的质量分布；大于 60 个原子的主体分子不出题（超出常见转动光谱范围）。", "inputs": "XYZ（QM7-X 优化几何或 SAMPL9 客体结构，按行号计数）。", "physical_conditions": "刚性转子、给定平衡几何；题干写明最丰同位素质量、h 与原子质量单位；不含振动平均、离心畸变与超精细结构。", "output_forms": [ "numeric", "choice" ], "source_basis": "QM7-X（CC BY 4.0）、SAMPL9 客体（仓库内已有资产，不新下载）", "answer_method": "质量加权、平移到质心的惯量张量，循环 Jacobi 对角化；跃迁频率按非对称陀螺 J = 1 能级（B+C、A+C、A+B）；以张量迹不变量自检。", "distractor_mechanisms": [ "other_J1_level", "linear_rotor_2B", "unrotated_tensor_diagonal", "origin_not_centre_of_mass", "heavy_atoms_only", "four_pi_squared", "polar_moment_as_I", "average_atomic_weights", "equal_masses", "single_constant", "hbar_instead_of_h" ], "four_choice_validation": "与正确值相差 > 2%（且 ≥ 20 个末位），干扰项之间 > 1%；升序排列，秩按批次轮换；枚举实例设 strict_rank。2026-10-06 探针 runs/enum-v32：99 个候选接受 81，独立检查全部一致；1_10 线答案几乎不在 A、1_01 线几乎不在 D（其余能级在物理上位于一侧），三条线合起来位置均衡。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/rotational.py", "checkers/rotational.py", "enumerators.py" ], "tests": [ "tests/test_rotational.py" ], "builder_modules": "族引擎内实现（M3–M5）；枚举器 enumerators.py。", "gap": "尚无 J = 2 ← 1 等更高跃迁与离心畸变；题干不给能级公式（考查是否知道非对称陀螺能级标号）。", "engine_families": [ "rotational_line" ], "independent_checker": "checkers/rotational.py:rotational_line", "instance_ids": [ "FA-ROT-7029-111", "FA-ROT-G9-101", "FA-ROT-G2-110" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1", "1-10" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "isotopologue_design", "name": "单氘代位点设计", "ability": "design", "status": "pilot", "priority": 1, "concept": "为替代法（Kraitchman）结构测定选择单氘代位点：四个候选 H 中，换成 D 后使指定转动跃迁位移最大者。设计 = 转动谱线推断的逆问题；位移取决于到主轴的距离，而不只是到质心的距离。 题干写明每个候选 H 所连原子及其官能团角色（如“H10 连在甲基碳 C1 上”）。", "domains": [ "quantum", "chemistry" ], "input_scale": "整个分子（13–56 个原子）", "reasoning_scale": "整分子转动惯量（分子最大原子间距 0.5–1.2 nm）", "scale_cells": [ "0.1-1", "1-10" ], "scale_limits": "推理需要整个分子的质量分布；大于 60 个原子的主体分子不出题（超出常见转动光谱范围）。", "inputs": "XYZ（QM7-X 优化几何或 SAMPL9 客体结构，按行号计数）。", "physical_conditions": "刚性转子、给定平衡几何；题干写明最丰同位素质量、h 与原子质量单位；不含振动平均、离心畸变与超精细结构。", "output_forms": [ "choice" ], "source_basis": "QM7-X（CC BY 4.0）、SAMPL9 客体（仓库内已有资产，不新下载）", "answer_method": "对每个候选 H 换成 D 质量，重新计算质心、惯量张量与跃迁频率，取绝对位移最大者；要求领先第二名 ≥ 15% 且 ≥ 0.5 MHz。", "distractor_mechanisms": [ "farthest_from_centre_of_mass", "best_for_other_line", "best_for_single_constant", "smaller_shift" ], "four_choice_validation": "恰好一个候选位移最大且领先足够；正确项按批次目标位置放置。枚举器优先提出“离质心更远的 H 是干扰项”的集合，并按约 1:3 加入捷径能答对的集合；runs/enum-v32 中捷径命中 52/524，改进后的候选中为 90/610（约 15%，低于随机）。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/rotational.py", "task_families/chem_names.py", "checkers/rotational.py", "enumerators.py" ], "tests": [ "tests/test_rotational.py", "tests/test_chem_names.py" ], "builder_modules": "族引擎内实现（M3–M5）；枚举器 enumerators.py。", "gap": "尚未包含双取代、13C 取代与 Kraitchman 坐标精度（位点靠近主轴时不确定度大）的定量设计。", "engine_families": [ "isotopologue_design" ], "independent_checker": "checkers/rotational.py:isotopologue_design", "instance_ids": [ "FA-ISO-7095-111", "FA-ISO-G9-110", "FA-ISO-G13-111" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1", "1-10" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "hbond_geometry", "name": "氢键几何量", "ability": "perception", "status": "pilot", "priority": 1, "concept": "分子内氢键 D–H···A 的几何量：给体–受体距离、氢–受体距离或 D–H···A 角；原子按化学角色命名（如羟基氧、羰基氧、酰胺氮）。", "domains": [ "chemistry" ], "input_scale": "整个分子（14 个原子）", "reasoning_scale": "给体–受体或氢–受体距离（0.15–0.5 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "分子内单个接触的局部几何。", "inputs": "XYZ（论文 SI 的计算构象，按行号计数）。", "physical_conditions": "单一计算构象，坐标原样；不含振动平均与溶剂。", "output_forms": [ "numeric", "choice" ], "source_basis": "Altun, Bleda & Trindle, Molecules 26, 2623 (2021)（CC BY 4.0）SI 第 3–6 页坐标（仓库内已有资产）", "answer_method": "指定原子的欧氏距离或以 H 为顶点的向量夹角；共价半径规则确认 D–H 成键、H–A 不成键。", "distractor_mechanisms": [ "h_acceptor_instead", "donor_acceptor_instead", "path_through_h", "linear_assumption", "covalent_d_h", "other_acceptor", "angle_at_donor", "angle_at_acceptor", "supplement", "acceptor_side" ], "four_choice_validation": "距离与正确值相差 > 0.03 Å、角度 > 3°；升序排列，秩按批次轮换。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/hbond.py", "checkers/hbonds.py", "task_families/chem_names.py" ], "tests": [ "tests/test_hbond.py" ], "builder_modules": "族引擎内实现（M3–M5）；数据只有 4 个构象，未写枚举器。", "gap": "只有一篇论文的 4 个构象；扩量需要更多含分子内氢键的结构（如 GEOM、QM7-X 中的羟基 / 酰胺分子）。", "engine_families": [ "hbond_geometry" ], "independent_checker": "checkers/hbonds.py:hbond_geometry", "instance_ids": [ "FA-HBG-TWO-OO", "FA-HBG-FIVE-HN", "FA-HBG-FOUR-ANG" ], "instance_count": 3, "independent_pass": 3, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "hbond_criteria", "name": "按几何判据识别氢键", "ability": "inference", "status": "pilot", "priority": 1, "concept": "给定操作判据（H···A ≤ d、D–H···A ≥ θ），在四个候选接触中判断哪一个构成分子内氢键；考查“只看给体–受体距离不够，氢要指向受体”。", "domains": [ "chemistry" ], "input_scale": "整个分子（14 个原子）", "reasoning_scale": "给体–受体或氢–受体距离（0.15–0.5 nm）", "scale_cells": [ "0.1-1" ], "scale_limits": "分子内单个接触的局部几何。", "inputs": "XYZ（论文 SI 的计算构象，按行号计数）。", "physical_conditions": "单一计算构象，坐标原样；不含振动平均与溶剂。", "output_forms": [ "choice" ], "source_basis": "Altun, Bleda & Trindle, Molecules 26, 2623 (2021)（CC BY 4.0）SI 第 3–6 页坐标（仓库内已有资产）", "answer_method": "逐个候选计算 H···A 与 D–H···A 角，恰好一个同时满足两项判据；每个候选与阈值至少相差 0.10 Å / 5°。", "distractor_mechanisms": [ "donor_acceptor_close", "distance_only", "angle_only", "neither" ], "four_choice_validation": "恰好一个候选满足两项判据；其余各记录不满足的原因；正确项按批次目标位置放置。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/hbond.py", "checkers/hbonds.py", "task_families/chem_names.py" ], "tests": [ "tests/test_hbond.py" ], "builder_modules": "族引擎内实现（M3–M5）；数据只有 4 个构象，未写枚举器。", "gap": "同上：构象来源有限。", "engine_families": [ "hbond_criteria" ], "independent_checker": "checkers/hbonds.py:hbond_criteria", "instance_ids": [ "FA-HBC-SET1", "FA-HBC-SET2" ], "instance_count": 2, "independent_pass": 2, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "evidence_quantity", "name": "论文证据 → 计算 → 结论", "ability": "inference", "status": "pilot", "priority": 1, "concept": "给出论文中的若干事实（数值与观测），先算一个几何量（载流子间距、环路面积、轨道临界场、等面积转角、涡旋数、反射带宽、HTP 比值、给体–受体距离差），再判断论文的关键结论。", "domains": [ "quantum", "chemistry" ], "input_scale": "证据卡（论文事实）", "reasoning_scale": "由证据算出的几何量（0.26 nm–0.86 μm）", "scale_cells": [ "0.1-1", "10-100", "100-1000" ], "scale_limits": "尺度取所算几何量（间距、厚度、莫尔周期、螺距、给体–受体距离等）。", "inputs": "evidence.txt（论文事实逐条转述，E1–En）；氢键题另附论文 SI 构象 XYZ。", "physical_conditions": "按论文的实验或计算条件写入证据条目。", "output_forms": [ "choice" ], "source_basis": "第 1 步下载的论文全文（本地 inputs/papers/，不入库）与 Altun 2021 SI；证据卡 docs/assets/families/evidence/*.json 逐条记录原文位置。", "answer_method": "数值由代码从证据卡计算；四个选项 = 数值（正确 / 登记的错误做法）× 结论（原文结论 / 原文排除的解释），恰好一个两者都对。", "distractor_mechanisms": [ "rejected_conclusion（原文排除的解释）", "wrong_constant（如 h/e 代替 h/2e）", "wrong_input_fact（如用总密度代替有效密度）", "wrong_formula（如垂直场公式、只取单一方向）", "confused_quantity（如 H···A 代替 D···A、中心代替带宽）" ], "four_choice_validation": "两个数值相差 ≥ 规定间隔且显示不同；四个组合齐全；独立检查器以 Decimal 复算并逐项解析选项。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/evidence.py", "checkers/evidence.py", "tools/build_evidence_assets.py" ], "tests": [ "tests/test_evidence.py" ], "builder_modules": "族引擎内实现（M3–M5）；证据卡人工转录，每条带原文位置供 L1 对照。", "gap": "第一批 9 道（7 篇论文）；尚未覆盖材料、生物。", "engine_families": [ "evidence_quantity" ], "independent_checker": "checkers/evidence.py:evidence_quantity", "instance_ids": [ "FA-EVQ-CAO-COHERENCE", "FA-EVQ-CAO-SQUID", "FA-EVQ-CAO-PAULI", "FA-EVQ-KER-STRAIN", "FA-EVQ-SCH-FLUX", "FA-EVQ-SHA-BAND", "FA-EVQ-DIN-HTP", "FA-EVQ-ALT-OHSHIFT", "FA-EVQ-ALT-NHO", "FA-EVQ-FOO-LAYERS", "FA-EVQ-FIL-RBCS", "FA-EVQ-JIA-CLAMP" ], "instance_count": 12, "independent_pass": 12, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1", "1-10", "10-100", "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 }, { "id": "evidence_elimination", "name": "论文证据 → 排除 → 结论", "ability": "inference", "status": "pilot", "priority": 1, "concept": "给出论文中的若干观测，四个候选解释中只有一个与全部观测一致；其余三个是论文讨论并排除的解释（或超出证据的说法）。", "domains": [ "quantum", "chemistry" ], "input_scale": "证据卡（论文观测）", "reasoning_scale": "证据所描述结构的特征尺度（0.26 nm–0.45 μm）", "scale_cells": [ "0.1-1", "10-100", "100-1000" ], "scale_limits": "尺度取证据所描述结构的特征长度（莫尔晶胞、涡旋间距、螺距、O···O 距离）。", "inputs": "evidence.txt（论文事实逐条转述，E1–En）；氢键题另附论文 SI 构象 XYZ。", "physical_conditions": "按论文的实验或计算条件写入证据条目。", "output_forms": [ "choice" ], "source_basis": "第 1 步下载的论文全文（本地 inputs/papers/，不入库）与 Altun 2021 SI；证据卡 docs/assets/families/evidence/*.json 逐条记录原文位置。", "answer_method": "证据卡为每个解释写明对相关观测的预期；代码逐条比对观测，恰好一个解释无矛盾且为原文结论，其余每个至少与一条给学生的事实矛盾。", "distractor_mechanisms": [ "rejected_alternative（原文排除的解释）", "overreach（超出证据的结论）", "surface_cue（只看表面线索，如有键径即判氢键）" ], "four_choice_validation": "每个预期都对应某条观测；恰好一个解释存活且为原文结论；独立检查器重新比对。", "legacy_families": [], "legacy_qa": [], "code": [ "task_families/evidence.py", "checkers/evidence.py", "tools/build_evidence_assets.py" ], "tests": [ "tests/test_evidence.py" ], "builder_modules": "族引擎内实现（M3–M5）；证据卡人工转录，每条带原文位置供 L1 对照。", "gap": "第一批 4 道（4 篇论文）。", "engine_families": [ "evidence_elimination" ], "independent_checker": "checkers/evidence.py:evidence_elimination", "instance_ids": [ "FA-EVE-KER-NEMATIC", "FA-EVE-COT-BEND", "FA-EVE-SHA-IONS", "FA-EVE-ALT-THREE", "FA-EVE-CAO-POCKETS", "FA-EVE-KER-BANDWIDTH", "FA-EVE-KER-GAP", "FA-EVE-COT-LATTICE", "FA-EVE-ALT-STABILITY", "FA-EVE-MUH-SKYRMION", "FA-EVE-FIL-ANGLE", "FA-EVE-JIA-CHANNEL", "FA-EVE-JIA-SEQUENCE", "FA-EVE-KCS-COLLAPSE", "FA-EVE-RAS-DOMAIN" ], "instance_count": 15, "independent_pass": 15, "instance_failures": 0, "instance_reasoning_cells": [ "0.1-1", "1-10", "10-100", "100-1000" ], "legacy_decisions": {}, "legacy_choice_v01": 0 } ] }