nx_dr_drbench_cli.nx
buildroot/runtime/nx_dr_drbench_cli.nx
about
nx_dr_drbench_cli.nx -- callable MCP/API surface for the DRBench+DEER scorer (DR-0).
Scores a deep-research report from per-item judge flags and emits the field-standard
scorecard JSON. Insight recall is the BINDING constraint (13-36% across the field).
nx_dr_drbench <insight_csv> <distractor_csv> <factuality_csv> <completeness_csv> <evidence_csv>
each = comma-separated 0/1 flags:
insight[i]=1 if groundtruth insight i is covered by the report
distractor[i]=1 if injected distractor i leaked into the report (bad)
factuality[i]=1 if cited claim i is supported
completeness[i]=1 if request sub-question i is answered (DEER)
evidence[i]=1 if claim i (cited OR uncited, back-tracked) has valid evidence (DEER)
e.g. nx_dr_drbench 1,1,0,0,1 0,0,1,0 1,1,1,0 1,1,1,1,0,0 1,0,1
-> {"insight_recall_permil":600,"distractor_avoidance_permil":750,...,"overall_permil":686}
license_tier: ORIGINAL No hw writes (Rule 26). expect_exit: 0
dependencies 3 imports · 0 importers
imports: nx_dr_drbench.nxnx_itoa_lib.nxnx_syscalls.nx
imported by: nobody (leaf or entry point)
call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown
structs
| none |
consts
| 17 | const K_MAGIC_1024: i64 = 1024 |
functions
| 19 | func db_puts(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } sys_write(1, s, n); return 0 } |
| 20 | func db_q() -> i64 { let b: *u8 = sys_mmap(1); b[0] = 34 as u8; sys_write(1, b, 1); return 0 } |
| 21 | func db_key(name: *u8) -> i64 { db_q(); db_puts(name); db_q(); db_puts(":" as *u8); return 0 } |
| 26 | func db_num(v: i64) -> i64 { nxi_out(v); return 0 } |
| 27 | func db_strlen(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } return n } called by 1: db_parse_csv |
| 29 | func db_parse_csv(s: *u8, out: *i64, maxn: i64) -> i64 |
| 42 | func main(argc: i64, argv: *i64) -> i64 |