evaluator_reference.json

{
  "schema": "sciencegym.evaluator_reference.design.v2",
  "agent_visible": false,
  "independent_of_agent_self_report": true,
  "authority": "Trusted world event/state history, immutable acquisition records and material/object lineage; never submitted success booleans",
  "branch_requirements": [
    {
      "branch_id": "tubulin",
      "entry": "U-2 OS stock identity token, not a pre-stained or pre-mounted specimen",
      "required_stage_ids": [
        "tubulin__tub_seed",
        "tubulin__tub_grow",
        "tubulin__tub_fix",
        "tubulin__tub_permeabilize",
        "tubulin__tub_block",
        "tubulin__tub_primary",
        "tubulin__tub_wash1",
        "tubulin__tub_reblock",
        "tubulin__tub_secondary",
        "tubulin__tub_wash2",
        "tubulin__tub_mount"
      ],
      "sequence_edges": [
        [
          "tubulin__tub_seed",
          "tubulin__tub_grow"
        ],
        [
          "tubulin__tub_grow",
          "tubulin__tub_fix"
        ],
        [
          "tubulin__tub_fix",
          "tubulin__tub_permeabilize"
        ],
        [
          "tubulin__tub_permeabilize",
          "tubulin__tub_block"
        ],
        [
          "tubulin__tub_block",
          "tubulin__tub_primary"
        ],
        [
          "tubulin__tub_primary",
          "tubulin__tub_wash1"
        ],
        [
          "tubulin__tub_wash1",
          "tubulin__tub_reblock"
        ],
        [
          "tubulin__tub_reblock",
          "tubulin__tub_secondary"
        ],
        [
          "tubulin__tub_secondary",
          "tubulin__tub_wash2"
        ],
        [
          "tubulin__tub_wash2",
          "tubulin__tub_mount"
        ]
      ],
      "required_acquisition_ids": [
        "tub_confocal",
        "tub_widefield"
      ],
      "allowed_optional_choices": {},
      "analysis_requirements": [
        "Preserve paired raw stacks and sampling per modality",
        "Select the matching widefield PSF metadata and obtain preset DW derivative",
        "Register the same FOV across modalities; inspect lateral and axial views",
        "Compare structures and retain low-intensity discrepancies rather than declaring every sharper output true"
      ]
    },
    {
      "branch_id": "ifish",
      "entry": "HAP1 stock identity token and separately supplied probe-lot tokens",
      "required_stage_ids": [
        "ifish__if_seed",
        "ifish__if_grow",
        "ifish__if_fix",
        "ifish__if_quench",
        "ifish__if_wash1",
        "ifish__if_perm",
        "ifish__if_wash2",
        "ifish__if_acid",
        "ifish__if_wash3",
        "ifish__if_rinse1",
        "ifish__if_store",
        "ifish__if_equilibrate",
        "ifish__if_prehyb",
        "ifish__if_mix",
        "ifish__if_seal",
        "ifish__if_denature",
        "ifish__if_hyb1",
        "ifish__if_rinse2",
        "ifish__if_hotwash",
        "ifish__if_rinse3",
        "ifish__if_rinse4",
        "ifish__if_wash4",
        "ifish__if_hyb2",
        "ifish__if_wash5",
        "ifish__if_hoechst",
        "ifish__if_wash6",
        "ifish__if_mount"
      ],
      "sequence_edges": [
        [
          "ifish__if_seed",
          "ifish__if_grow"
        ],
        [
          "ifish__if_grow",
          "ifish__if_fix"
        ],
        [
          "ifish__if_fix",
          "ifish__if_quench"
        ],
        [
          "ifish__if_quench",
          "ifish__if_wash1"
        ],
        [
          "ifish__if_wash1",
          "ifish__if_perm"
        ],
        [
          "ifish__if_perm",
          "ifish__if_wash2"
        ],
        [
          "ifish__if_wash2",
          "ifish__if_acid"
        ],
        [
          "ifish__if_acid",
          "ifish__if_wash3"
        ],
        [
          "ifish__if_wash3",
          "ifish__if_rinse1"
        ],
        [
          "ifish__if_rinse1",
          "ifish__if_store"
        ],
        [
          "ifish__if_store",
          "ifish__if_equilibrate"
        ],
        [
          "ifish__if_equilibrate",
          "ifish__if_prehyb"
        ],
        [
          "ifish__if_prehyb",
          "ifish__if_mix"
        ],
        [
          "ifish__if_mix",
          "ifish__if_seal"
        ],
        [
          "ifish__if_seal",
          "ifish__if_denature"
        ],
        [
          "ifish__if_denature",
          "ifish__if_hyb1"
        ],
        [
          "ifish__if_hyb1",
          "ifish__if_rinse2"
        ],
        [
          "ifish__if_rinse2",
          "ifish__if_hotwash"
        ],
        [
          "ifish__if_hotwash",
          "ifish__if_rinse3"
        ],
        [
          "ifish__if_rinse3",
          "ifish__if_rinse4"
        ],
        [
          "ifish__if_rinse4",
          "ifish__if_wash4"
        ],
        [
          "ifish__if_wash4",
          "ifish__if_hyb2"
        ],
        [
          "ifish__if_hyb2",
          "ifish__if_wash5"
        ],
        [
          "ifish__if_wash5",
          "ifish__if_hoechst"
        ],
        [
          "ifish__if_hoechst",
          "ifish__if_wash6"
        ],
        [
          "ifish__if_wash6",
          "ifish__if_mount"
        ]
      ],
      "required_acquisition_ids": [
        "if_widefield"
      ],
      "allowed_optional_choices": {
        "ifish__if_store": [
          "immediate",
          "source-permitted storage"
        ]
      },
      "analysis_requirements": [
        "Keep four iFISH channels and nuclear identity separate",
        "Preserve raw and preset DW derivatives",
        "Compare per-nucleus/per-channel nearest neighbors at source threshold 260 nm",
        "Label Raw&DW, DW&Raw, Lost and New; exclude nuclei with no dots in either comparison image",
        "Retain FWHM/NCR outputs; upscaling to 23 nm for ED7 is display processing, not acquisition sampling"
      ]
    },
    {
      "branch_id": "gapdh",
      "entry": "SKBR3 stock identity token, source-specific probe lots and empty authored carrier",
      "required_stage_ids": [
        "gapdh__gap_culture",
        "gapdh__gap_unreported_prep",
        "gapdh__fish_probe",
        "gapdh__fish_hyb",
        "gapdh__fish_wash1",
        "gapdh__fish_secondary",
        "gapdh__fish_wash2",
        "gapdh__fish_counterstain",
        "gapdh__fish_mount"
      ],
      "sequence_edges": [
        [
          "gapdh__gap_culture",
          "gapdh__gap_unreported_prep"
        ],
        [
          "gapdh__gap_unreported_prep",
          "gapdh__fish_probe"
        ],
        [
          "gapdh__fish_probe",
          "gapdh__fish_hyb"
        ],
        [
          "gapdh__fish_hyb",
          "gapdh__fish_wash1"
        ],
        [
          "gapdh__fish_wash1",
          "gapdh__fish_secondary"
        ],
        [
          "gapdh__fish_secondary",
          "gapdh__fish_wash2"
        ],
        [
          "gapdh__fish_wash2",
          "gapdh__fish_counterstain"
        ],
        [
          "gapdh__fish_counterstain",
          "gapdh__fish_mount"
        ]
      ],
      "required_acquisition_ids": [
        "gap_widefield"
      ],
      "allowed_optional_choices": {},
      "analysis_requirements": [
        "Retain matching raw, DW, DL2 and Huygens preset records",
        "For ED6 display comparison manually choose DW intensity threshold and match raw brightest-dot count",
        "Separately compare Intensity and DoG detection across paired raw/DW fields; do not collapse the two tests",
        "Record FWHM and per-FOV counts without requiring raw counts always be lower"
      ]
    },
    {
      "branch_id": "tissue",
      "entry": "Purchased 5 µm frozen TMA section token, US Biomax FMC282e; no donated-person identifying data",
      "required_stage_ids": [
        "tissue__tis_fix",
        "tissue__tis_pbs",
        "tissue__tis_ethanolrinse",
        "tissue__tis_ethanol",
        "tissue__tis_rehydrate",
        "tissue__tis_fish_probe",
        "tissue__tis_fish_hyb",
        "tissue__tis_fish_wash1",
        "tissue__tis_fish_secondary",
        "tissue__tis_fish_wash2",
        "tissue__tis_fish_counterstain",
        "tissue__tis_fish_mount"
      ],
      "sequence_edges": [
        [
          "tissue__tis_fix",
          "tissue__tis_pbs"
        ],
        [
          "tissue__tis_pbs",
          "tissue__tis_ethanolrinse"
        ],
        [
          "tissue__tis_ethanolrinse",
          "tissue__tis_ethanol"
        ],
        [
          "tissue__tis_ethanol",
          "tissue__tis_rehydrate"
        ],
        [
          "tissue__tis_rehydrate",
          "tissue__tis_fish_probe"
        ],
        [
          "tissue__tis_fish_probe",
          "tissue__tis_fish_hyb"
        ],
        [
          "tissue__tis_fish_hyb",
          "tissue__tis_fish_wash1"
        ],
        [
          "tissue__tis_fish_wash1",
          "tissue__tis_fish_secondary"
        ],
        [
          "tissue__tis_fish_secondary",
          "tissue__tis_fish_wash2"
        ],
        [
          "tissue__tis_fish_wash2",
          "tissue__tis_fish_counterstain"
        ],
        [
          "tissue__tis_fish_counterstain",
          "tissue__tis_fish_mount"
        ]
      ],
      "required_acquisition_ids": [
        "tis20",
        "tis60"
      ],
      "allowed_optional_choices": {
        "tissue__tis_fish_secondary": [
          "direct-label interpretation: skip secondary-specific application while retaining wash/counterstain",
          "literal inherited-steps interpretation: execute secondary stage with source ambiguity recorded"
        ]
      },
      "analysis_requirements": [
        "Survey 10 labeled authored FOV slots at 20× and retain matching selected 60× regions",
        "Use DoG dot detection, 60× nuclear masks and rescale to corresponding 20× sampling",
        "Review intensity/CNR plots; select local-minimum threshold where the two clouds exist",
        "20× raw has a distinct tail-selection condition; do not force the same threshold logic",
        "Compare HQ dots against shared 60× raw/DW reference and retain translation/deformation registration metadata"
      ]
    },
    {
      "branch_id": "gfap",
      "entry": "FFPE human cerebral-cortex TMA block token from source-described biobank-derived array",
      "required_stage_ids": [
        "gfap__g_cut",
        "gfap__g_bake",
        "gfap__g_dewax",
        "gfap__g_retrieve",
        "gfap__g_primary",
        "gfap__g_secondary",
        "gfap__g_mount"
      ],
      "sequence_edges": [
        [
          "gfap__g_cut",
          "gfap__g_bake"
        ],
        [
          "gfap__g_bake",
          "gfap__g_dewax"
        ],
        [
          "gfap__g_dewax",
          "gfap__g_retrieve"
        ],
        [
          "gfap__g_retrieve",
          "gfap__g_primary"
        ],
        [
          "gfap__g_primary",
          "gfap__g_secondary"
        ],
        [
          "gfap__g_secondary",
          "gfap__g_mount"
        ]
      ],
      "required_acquisition_ids": [
        "g60",
        "g63",
        "g100"
      ],
      "allowed_optional_choices": {},
      "analysis_requirements": [
        "Acquire same ROI mosaic on 60× widefield and 63× confocal; either order is allowed",
        "Create separate 100× comparison branch with unknown settings labeled task presets",
        "Preserve raw/DW/DL2 comparisons and axial views",
        "Record structural agreement and loss of low-intensity details as separate observations"
      ]
    },
    {
      "branch_id": "nuclear_pores",
      "entry": "Source-literal PtK2 identity token with unresolved species label; no live biological material instantiated",
      "required_stage_ids": [
        "nuclear_pores__p_identity",
        "nuclear_pores__p_upstream",
        "nuclear_pores__p_primary",
        "nuclear_pores__p_secondary",
        "nuclear_pores__p_mount"
      ],
      "sequence_edges": [
        [
          "nuclear_pores__p_identity",
          "nuclear_pores__p_upstream"
        ],
        [
          "nuclear_pores__p_upstream",
          "nuclear_pores__p_primary"
        ],
        [
          "nuclear_pores__p_primary",
          "nuclear_pores__p_secondary"
        ],
        [
          "nuclear_pores__p_secondary",
          "nuclear_pores__p_mount"
        ]
      ],
      "required_acquisition_ids": [
        "p_sted",
        "p_confocal"
      ],
      "allowed_optional_choices": {
        "nuclear_pores__p_identity": [
          "immediate",
          "source-permitted storage"
        ]
      },
      "analysis_requirements": [
        "Keep STED-first and thereafter-confocal acquisitions linked to one specimen",
        "Compare raw/Huygens/DW results with correct modality-specific PSF card",
        "Keep widefield Born-Wolf, confocal and STED PSF choices distinct",
        "Report unresolved physical fluorescence compatibility; task preset is not experimental validation"
      ]
    }
  ],
  "data_requirements": [
    {
      "id": "software_benchmark",
      "title": "Compare deconvolution tools on source-described benchmark inputs",
      "inputs": [
        "synthetic microtubule reference",
        "synthetic hollow bar reference",
        "external C. elegans whole-embryo reference"
      ],
      "actions": [
        "Select source-matched raw/truth pair; preserve identity",
        "Choose DW SHB versus unaccelerated RL iteration series",
        "Compare DW, DL2, Huygens and RedLionfish preset results without executing proprietary software",
        "Inspect MSE-versus-iteration and time records with hardware context",
        "Use wrap boundary assumption only on source synthetic inputs identified as wrapped"
      ],
      "source_refs": [
        "R_BENCH",
        "ED1",
        "SIN4"
      ],
      "unknowns": [
        "No real input arrays, executable comparisons or timing measurements supplied here"
      ]
    },
    {
      "id": "psf",
      "title": "Compare PSF calculator versus PSF Generator",
      "inputs": [
        "ChrX-36plex PSF-comparison source token"
      ],
      "actions": [
        "Verify 36plex identity, distinct from 46plex tracing data",
        "Select two PSF-method cards with all other DW parameters held equal",
        "Link raw to both preset derivatives",
        "Compare dot FWHM/PSF displays",
        "Record main Lanczos-3 versus SI Lanczos-5 discrepancy without silently choosing"
      ],
      "source_refs": [
        "R_BENCH",
        "M87",
        "SI1",
        "SIN13"
      ],
      "unknowns": [
        "Full optical metadata and original image arrays not retained"
      ]
    },
    {
      "id": "boundary",
      "title": "Handle lateral and axial boundary comparisons",
      "inputs": [
        "external embryo stack",
        "HAP1 Hoechst boundary-example stack with 51 planes"
      ],
      "actions": [
        "Select original and derived cropped copies; preserve raw full stack",
        "Split embryo into four cuboids for lateral comparison",
        "Create distinct HAP1 derivative after removing bottom 12 planes for Fig. 2 comparison",
        "Select DW versus DL2 boundary/padding/apodization presets",
        "Inspect lateral/axial profiles and label crops; never treat crop as sample removal"
      ],
      "source_refs": [
        "R_BENCH",
        "ED2",
        "SI2",
        "SIN4"
      ],
      "unknowns": [
        "Connection between boundary-example HAP1 specimen and iFISH specimen not established; no forced same-sample link"
      ]
    },
    {
      "id": "synthetic_dots",
      "title": "Inspect dot-density/noise simulation comparisons through preset outputs",
      "inputs": [
        "authored stand-ins for source synthetic truth/noisy widefield/confocal conditions"
      ],
      "actions": [
        "Choose declared density/noise condition",
        "Keep truth separate from observed noisy input",
        "Choose widefield/confocal and DW/DL2 derivative records",
        "Compare detection overlap against truth and inspect MSE",
        "Tag source synthetic result as best-case matched-PSF condition; do not present as empirical specimen result"
      ],
      "source_refs": [
        "M97",
        "R_FISH",
        "ED4",
        "SI56"
      ],
      "unknowns": [
        "No optical simulation implemented; all task observations are presets"
      ]
    },
    {
      "id": "isst",
      "title": "Process reused ISST dataset without inventing new wet-lab preparation",
      "inputs": [
        "previously generated 120-gene middle-temporal-gyrus image dataset token"
      ],
      "actions": [
        "Import raw cycle/FOV/channel lineage",
        "Select raw and preset DW branches",
        "Align cycles and stitch FOVs using operator workstation cards",
        "Select FindSpots masking radius 15 and compare threshold sweep",
        "Record source selected normalized threshold 2% and decode assigned versus improper barcode outputs",
        "Compare cell-typing results for 18 types with identical underlying cell inventory"
      ],
      "source_refs": [
        "M105",
        "R_ISST",
        "ED10"
      ],
      "unknowns": [
        "Raw images and gene table absent; preset data token only",
        "This paper reused images, so no new ISST wet-lab protocol is invented"
      ]
    },
    {
      "id": "oligofisseq",
      "title": "Inspect reused ChrX-46plex decoding and tracing comparisons",
      "inputs": [
        "ChrX-46plex O-eLIT source dataset token; seven replicate dataset slots"
      ],
      "actions": [
        "Keep five imaging/sequencing cycles in source data lineage",
        "Select raw, NIS and DW derivatives",
        "Manually match nucleus-pair records using fixture identification cues; source reports 168 nuclei",
        "Run two-tier every-pixel analysis card on the paired dataset tokens",
        "Inspect tracing nodes and interpolation flags",
        "Compare contact maps against linked Hi-C reference",
        "Keep seven biological replicate datasets distinct from 1,000 bootstrap resamples"
      ],
      "source_refs": [
        "M107",
        "R_OLIGO",
        "R_ISST",
        "ED10"
      ],
      "unknowns": [
        "No upstream OligoFISSEQ probe production or sequencing procedure invented",
        "168 match operations may be task-batched with coverage record, never claimed as observed here"
      ]
    },
    {
      "id": "rln_comparator",
      "title": "Inspect downloaded raw image against a published RLN screenshot comparator",
      "inputs": [
        "external U-2 OS mitochondria/actin/tubulin raw token",
        "published RLN screenshot-reference token"
      ],
      "actions": [
        "Keep this external U-2 OS sample separate from in-paper alpha-tubulin specimen",
        "Prepare visualization-only derivative in image-viewer card",
        "Select DW 50-iteration preset derivative",
        "Inspect paired zoom regions",
        "Mark comparator as screenshot from published figure, not rerun RLN"
      ],
      "source_refs": [
        "SI9"
      ],
      "unknowns": [
        "No screenshot or raw image redistributed; task uses authored stand-ins"
      ]
    }
  ],
  "hard_gates": [
    "No required reported preparation step omitted; an initial downstream specimen cannot substitute",
    "Every source-unreported operation executed through an explicit authored gap card, without claiming scientific provenance",
    "Required reagent identity and ordered wash/incubation events match branch reference",
    "Source temperatures/times/concentrations and acquisition parameters preserved exactly when reported",
    "No glass/carrier/specimen identity swap",
    "Same-FOV and modality controls satisfied from fixture mapping and immutable metadata",
    "Every required acquisition saved and every derivative parent valid",
    "All six specimens archived with location and no active optical exposure",
    "All touched station shutters closed, objectives in load/park state, mounts cleared, tools returned and waste/tips accounted"
  ],
  "specific_controls": [
    {
      "id": "C_tub_pair",
      "rule": "tub_confocal acquisition precedes tub_widefield, same specimen and same FOV",
      "source_refs": [
        "M80"
      ]
    },
    {
      "id": "C_gfap_pair",
      "rule": "g60/g63 use same 4×4 ROI mosaic with 10% overlap; no prescribed inter-modality chronology",
      "source_refs": [
        "M78"
      ]
    },
    {
      "id": "C_g100",
      "rule": "g100 uses separate caption-supported configuration, unresolved camera/stack not silently inherited",
      "source_refs": [
        "R_IF",
        "SI3",
        "ED3"
      ]
    },
    {
      "id": "C_tissue_pair",
      "rule": "10 20× FOVs and matched selected 60× subset; exact fixture field map is authored",
      "source_refs": [
        "M103"
      ]
    },
    {
      "id": "C_pore_order",
      "rule": "STED first, then confocal; preserve unresolved source identity/fluorophore annotations",
      "source_refs": [
        "M82"
      ]
    },
    {
      "id": "C_ifish",
      "rule": "four channel/nucleus groups; raw and DW pairs; empty-nucleus exclusions; no 51-plane global assumption",
      "source_refs": [
        "M101",
        "ED2"
      ]
    },
    {
      "id": "C_psf",
      "rule": "ChrX-36plex PSF comparison separated from ChrX-46plex tracing",
      "source_refs": [
        "SI1",
        "M107"
      ]
    },
    {
      "id": "C_rln",
      "rule": "screenshot comparator cannot be marked rerun RLN",
      "source_refs": [
        "SI9"
      ]
    }
  ],
  "source_unknown_policy": "Unknown is a legitimate source value. Completion may use named authored bridge cards but must retain the unknown and task-choice provenance. Never award source-complete reproduction for those bridged details.",
  "completion_labels": {
    "design_complete": "All declared paper branches have initial states, goals, object actions, output bindings, evaluator conditions and gap labels in this design",
    "task_episode_complete": "Requires future trusted runtime event history satisfying all hard gates; not established by this delivery",
    "paper_experiment_reproduced": "Not claimable from preset-output task"
  },
  "not_accepted_as_evidence": [
    "Rendered geometry",
    "agent plan text alone",
    "agent writes passed=true",
    "agent edits saved manifest to invent event IDs",
    "reused specimen ID with renamed carrier",
    "published image/result pasted as newly acquired output"
  ],
  "required_field_records": {
    "tub_confocal": [
      "TUB_F1"
    ],
    "tub_widefield": [
      "TUB_F1"
    ],
    "if_widefield": [
      "IF_F1"
    ],
    "gap_widefield": [
      "GAP_F1",
      "GAP_F2"
    ],
    "tis20": [
      "TIS_F1",
      "TIS_F2",
      "TIS_F3",
      "TIS_F4",
      "TIS_F5",
      "TIS_F6",
      "TIS_F7",
      "TIS_F8",
      "TIS_F9",
      "TIS_F10"
    ],
    "tis60": [
      "TIS_F1",
      "TIS_F2",
      "TIS_F3",
      "TIS_F4",
      "TIS_F5"
    ],
    "g60": [
      "GFAP_R1C1",
      "GFAP_R1C2",
      "GFAP_R1C3",
      "GFAP_R1C4",
      "GFAP_R2C1",
      "GFAP_R2C2",
      "GFAP_R2C3",
      "GFAP_R2C4",
      "GFAP_R3C1",
      "GFAP_R3C2",
      "GFAP_R3C3",
      "GFAP_R3C4",
      "GFAP_R4C1",
      "GFAP_R4C2",
      "GFAP_R4C3",
      "GFAP_R4C4"
    ],
    "g63": [
      "GFAP_R1C1",
      "GFAP_R1C2",
      "GFAP_R1C3",
      "GFAP_R1C4",
      "GFAP_R2C1",
      "GFAP_R2C2",
      "GFAP_R2C3",
      "GFAP_R2C4",
      "GFAP_R3C1",
      "GFAP_R3C2",
      "GFAP_R3C3",
      "GFAP_R3C4",
      "GFAP_R4C1",
      "GFAP_R4C2",
      "GFAP_R4C3",
      "GFAP_R4C4"
    ],
    "g100": [
      "GFAP_R2C2"
    ],
    "p_sted": [
      "PORE_F1"
    ],
    "p_confocal": [
      "PORE_F1"
    ]
  },
  "source_reported_steps_convention": "Methods numbers labeled number_of_steps_source_literal retain the source wording. Do not assume whether a controller counts planes or intervals; a named task-controller adapter may choose a frame convention, recorded as authored rather than paper fact.",
  "preparation_material_transfer_requirements": {
    "seed": "Stock specimen ancestry moves to the chosen carrier in the work vessel; stock vial and derived specimen are separate objects",
    "section": "Block remains at sectioning station; new section gets a child identity and its own support slide",
    "wash": "Material buffer ancestry updates but specimen and carrier remain unchanged",
    "mount": "Same prepared specimen-bearing carrier becomes supported by holder; do not spawn a duplicate final sample",
    "repeat_or_restart": "Fresh replacement starts a new child specimen ID; its failed sibling remains in quarantine"
  }
}