Lab BDS Build

1library(yamaa)23yamaa_spec("spec.yaml")$run()$output

1import yamaa23yamaa.spec("spec.yaml").run().output

1# Reference solution for the yamaa benchmark adam-adlb-bds (R track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67library(dplyr, warn.conflicts = FALSE)8library(readr)910adsl <- read_csv(11  "/app/input/adsl.csv",12  col_types = cols(TRTSDT = col_date(), .default = col_character())13)14lb <- read_csv(15  "/app/input/lb.csv",16  col_types = cols(17    LBDTC = col_date(),18    LBSTRESN = col_double(),19    .default = col_character()20  )21)2223# One record per collected ALT/AST result; a collected result with no24# numeric value produces no record.25collected <- lb |>26  filter(LBTESTCD %in% c("ALT", "AST"), !is.na(LBSTRESN))2728# ALT and AST hold the collected result and unit; ALTSI holds the ALT29# result times 0.0167 with unit ukat/L, one record for each ALT record.30alt <- collected |>31  filter(LBTESTCD %in% "ALT") |>32  transmute(33    STUDYID, USUBJID,34    PARAMCD = "ALT",35    PARAM = "Alanine Aminotransferase",36    ADT = LBDTC,37    AVAL = LBSTRESN,38    AVALU = LBSTRESU39  )40ast <- collected |>41  filter(LBTESTCD %in% "AST") |>42  transmute(43    STUDYID, USUBJID,44    PARAMCD = "AST",45    PARAM = "Aspartate Aminotransferase",46    ADT = LBDTC,47    AVAL = LBSTRESN,48    AVALU = LBSTRESU49  )50altsi <- collected |>51  filter(LBTESTCD %in% "ALT") |>52  transmute(53    STUDYID, USUBJID,54    PARAMCD = "ALTSI",55    PARAM = "Alanine Aminotransferase (SI)",56    ADT = LBDTC,57    AVAL = LBSTRESN * 0.0167,58    AVALU = "ukat/L"59  )6061base <- bind_rows(alt, ast, altsi) |>62  left_join(63    adsl |> select(STUDYID, USUBJID, TRTSDT, TRT01A),64    by = c("STUDYID", "USUBJID")65  )6667# ABLFL is Y on the latest record on or before the treatment start date68# for each subject and parameter, including the start date itself.69baseline_dates <- base |>70  filter(ADT <= TRTSDT) |>71  group_by(STUDYID, USUBJID, PARAMCD) |>72  summarise(BASE_DT = max(ADT), .groups = "drop")7374with_flag <- base |>75  left_join(baseline_dates, by = c("STUDYID", "USUBJID", "PARAMCD")) |>76  mutate(ABLFL = if_else(!is.na(BASE_DT) & ADT == BASE_DT, "Y", NA_character_))7778# BASE repeats the flagged baseline value; CHG and PCHG follow, with PCHG79# empty when the baseline is zero. Values are not rounded.80baselines <- with_flag |>81  filter(ABLFL %in% "Y") |>82  select(STUDYID, USUBJID, PARAMCD, BASE = AVAL)8384adlb_flagged <- with_flag |>85  select(-BASE_DT) |>86  left_join(baselines, by = c("STUDYID", "USUBJID", "PARAMCD")) |>87  mutate(88    CHG = AVAL - BASE,89    PCHG = if_else(90      is.na(BASE) | BASE == 0,91      NA_real_,92      100 * (AVAL - BASE) / BASE93    )94  )9596# ASEQ numbers each subject's records from 1, ordered by PARAMCD (ALT,97# then ALTSI, then AST) and then by ADT.98param_order <- c("ALT" = 0L, "ALTSI" = 1L, "AST" = 2L)99100adlb <- adlb_flagged |>101  mutate(PARAM_ORDER = param_order[PARAMCD]) |>102  arrange(STUDYID, USUBJID, PARAM_ORDER, ADT) |>103  group_by(STUDYID, USUBJID) |>104  mutate(ASEQ = row_number()) |>105  ungroup() |>106  arrange(PARAM_ORDER, USUBJID, ADT) |>107  select(108    STUDYID, USUBJID, PARAMCD, PARAM, ADT, TRTSDT, TRT01A,109    AVAL, AVALU, ABLFL, BASE, CHG, PCHG, ASEQ110  )111112dir.create("/app/output", showWarnings = FALSE)113write_csv(adlb, "/app/output/adlb.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adlb-bds (Python track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67import csv8from collections import OrderedDict910INPUT_DIR = "/app/input"11OUTPUT_DIR = "/app/output"121314def fmt(value):15    if value is None:16        return ""17    if isinstance(value, str):18        return value19    if value == int(value) and abs(value) < 1e15:20        return str(int(value))21    return repr(value)222324def main():25    adsl = {}26    with open(f"{INPUT_DIR}/adsl.csv", newline="") as f:27        for row in csv.DictReader(f):28            adsl[row["USUBJID"]] = row2930    params = {31        "ALT": ("ALT", "Alanine Aminotransferase"),32        "AST": ("AST", "Aspartate Aminotransferase"),33    }3435    records = []36    with open(f"{INPUT_DIR}/lb.csv", newline="") as f:37        for row in csv.DictReader(f):38            testcd = row["LBTESTCD"]39            if testcd not in params:40                continue41            raw = (row["LBSTRESN"] or "").strip()42            if raw == "":43                continue44            try:45                aval = float(raw)46            except ValueError:47                continue48            subj = adsl.get(row["USUBJID"], {})49            base_rec = {50                "STUDYID": row["STUDYID"],51                "USUBJID": row["USUBJID"],52                "ADT": row["LBDTC"],53                "TRTSDT": subj.get("TRTSDT", ""),54                "TRT01A": subj.get("TRT01A", ""),55            }56            paramcd, param = params[testcd]57            rec = dict(base_rec)58            rec.update({59                "PARAMCD": paramcd,60                "PARAM": param,61                "AVAL": aval,62                "AVAL_TEXT": raw,63                "AVALU": row["LBSTRESU"],64            })65            records.append(rec)66            if testcd == "ALT":67                si = dict(base_rec)68                si.update({69                    "PARAMCD": "ALTSI",70                    "PARAM": "Alanine Aminotransferase (SI)",71                    "AVAL": aval * 0.0167,72                    "AVAL_TEXT": None,73                    "AVALU": "ukat/L",74                })75                records.append(si)7677    # Baseline: latest record on or before TRTSDT per subject + parameter.78    groups = OrderedDict()79    for rec in records:80        groups.setdefault((rec["USUBJID"], rec["PARAMCD"]), []).append(rec)8182    for key, grecs in groups.items():83        baseline = None84        for rec in grecs:85            if rec["TRTSDT"] and rec["ADT"] <= rec["TRTSDT"]:86                if baseline is None or rec["ADT"] > baseline["ADT"]:87                    baseline = rec88        for rec in grecs:89            rec["ABLFL"] = "Y" if rec is baseline else ""90            if baseline is None:91                rec["BASE"] = None92                rec["CHG"] = None93                rec["PCHG"] = None94            else:95                base = baseline["AVAL"]96                chg = rec["AVAL"] - base97                rec["BASE"] = base98                rec["CHG"] = chg99                rec["PCHG"] = None if base == 0 else 100.0 * chg / base100101    records.sort(key=lambda r: (r["USUBJID"], r["PARAMCD"], r["ADT"]))102103    by_subject = OrderedDict()104    for rec in records:105        by_subject.setdefault(rec["USUBJID"], []).append(rec)106    for subj, srecs in by_subject.items():107        for i, rec in enumerate(srecs, start=1):108            rec["ASEQ"] = i109110    columns = ["STUDYID", "USUBJID", "PARAMCD", "PARAM", "ADT", "TRTSDT",111               "TRT01A", "AVAL", "AVALU", "ABLFL", "BASE", "CHG", "PCHG",112               "ASEQ"]113    with open(f"{OUTPUT_DIR}/adlb.csv", "w", newline="") as f:114        writer = csv.writer(f)115        writer.writerow(columns)116        for rec in records:117            writer.writerow([118                rec["STUDYID"],119                rec["USUBJID"],120                rec["PARAMCD"],121                rec["PARAM"],122                rec["ADT"],123                rec["TRTSDT"],124                rec["TRT01A"],125                rec["AVAL_TEXT"] if rec["AVAL_TEXT"] is not None126                else fmt(rec["AVAL"]),127                rec["AVALU"],128                rec["ABLFL"],129                fmt(rec["BASE"]),130                fmt(rec["CHG"]),131                fmt(rec["PCHG"]),132                rec["ASEQ"],133            ])134135136if __name__ == "__main__":137    main()

1/* opensas reference solution for the yamaa benchmark adam-adlb-bds.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data adsl;6  length STUDYID USUBJID TRTSDT TRT01A $1024;7  infile "/app/input/adsl.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. TRTSDT : $char1024. TRT01A : $char1024.;9run;1011data lb;12  length STUDYID USUBJID LBTESTCD LBTEST LBDTC LBSTRESU $1024;13  infile "/app/input/lb.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input STUDYID : $char1024. USUBJID : $char1024. LBTESTCD : $char1024. LBTEST : $char1024. LBDTC : $char1024. LBSTRESN LBSTRESU : $char1024.;15run;1617data parameters;18  length PARAMCD $16 PARAM $40 ADT $10 AVALU $1024;19  set lb;20  if LBTESTCD in ('ALT','AST') and not missing(LBSTRESN);21  ADT=LBDTC; PARAMCD=LBTESTCD; AVAL=LBSTRESN; AVALU=LBSTRESU;22  if LBTESTCD='ALT' then PARAM='Alanine Aminotransferase'; else PARAM='Aspartate Aminotransferase';23  output;24  if LBTESTCD='ALT' then do; PARAMCD='ALTSI'; PARAM='Alanine Aminotransferase (SI)'; AVAL=LBSTRESN*0.0167; AVALU='ukat/L'; output; end;25run;26proc sql; create table joined as select a.*,b.TRTSDT,b.TRT01A from parameters as a left join adsl as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID; quit;27proc sort data=joined(where=(ADT ne '' and TRTSDT ne '' and ADT<=TRTSDT)) out=candidates; by STUDYID USUBJID PARAMCD descending ADT; run;28data baseline; set candidates; by STUDYID USUBJID PARAMCD; if first.PARAMCD; BASE=AVAL; keep STUDYID USUBJID PARAMCD ADT BASE; rename ADT=BASEDT; run;29proc sql; create table combined as select a.*,b.BASE,b.BASEDT from joined as a left join baseline as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID and a.PARAMCD=b.PARAMCD; quit;30proc sort data=combined; by STUDYID USUBJID PARAMCD ADT; run;31data result;32  length ABLFL $1;33  set combined; by STUDYID USUBJID;34  if first.USUBJID then ASEQ=0; ASEQ+1;35  if ADT=BASEDT and not missing(BASEDT) then ABLFL='Y';36  CHG=AVAL-BASE;37  if not missing(BASE) and BASE ne 0 then PCHG=100*CHG/BASE;38run;3940proc sql;41  create table final as select STUDYID, USUBJID, PARAMCD, PARAM, ADT, TRTSDT, TRT01A, AVAL, AVALU, ABLFL, BASE, CHG, PCHG, ASEQ from result;42quit;43proc export data=final outfile="/app/output/adlb.csv" dbms=csv replace; run;