code wiki / _hdl_build / nx_referee_v2_test.nx
nx_referee_v2_test.nx
buildroot/runtime/_hdl_build/nx_referee_v2_test.nx
about
nx_referee_v2_test.nx -- the research-grounded gates catch the exact failures the literature documents.
thin suite (TPR 1000 / TNR 100): accepts correct but catches almost no wrong solutions -> NOT trustworthy
(mirrors CodeContests+ 4000 problems with TPR<=0.1; a single score would hide it)
aggregate [ACC,ACC,WA] -> WA (first non-ACC)
gaming: competitor can overwrite the oracle (SWE-Lancer assert 1==1) -> NOT gaming-proof
leakage: judge in the same model family as a competitor -> NOT independent
holdout: test seen in training -> NOT clean
the FULL rigor gate passes only when every condition holds.
Exit 0 on 8/8. license_tier: ORIGINAL
dependencies 3 imports · 0 importers
imports: nx_referee_v2.nxnx_referee.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
| none |
functions
| 15 | func vt_puts(s: *u8) -> i64 { var n: i64 = 0; while s[n] != (0 as u8) { n = n + 1 } sys_write(1, s, n); return 0 } |
| 16 | func vt_num(v: i64) -> i64 { let bb: *u8 = sys_mmap(28); var m: i64=v; if m<0 {m=0-m; sys_write(1,"-" as *u8,1)}; let t: *u8 = sys_mmap(28); var k: i64=0; if m==0 {t[0]=48;k=1}; while m>0 {t[k]=48+(m%10); m=m/10; k=k+1}; var i: i64=0; while i<k {bb[i]=t[k-1-i]; i=i+1}; sys_write(1, bb, k); return 0 } |
| 18 | func main() -> i64 |