Compare Fission against open decompilers under an original_binary oracle. Ranking axis is semantic pass rate only. Boundary honesty, per-function code, compiler pivots, and Ghidra IR parity each have their own route.
Request contract (binary, addr) — did we get that function unit?
Side-by-side decompiled code grid for every corpus function.
Original C next to Fission NIR (semantic) and HIR (readability) — dual-layer diagnostic, not ranking.
Semantic means pivoted by gcc / gcc-m32 and optimization level.
Bare-compile rate, readability proxies (goto / temps / flag soup), and realworld · multi-ISA track pivots — never ranking.
Layered parity (asm / p-code / CFG / discovery) — not ranking.
Fission's readable HIR next to Ghidra's own decompiled output, side by side — not ranking.
Fission's semantic pass rate release over release, plus what changed since the last one.
Same-function · snapshot
Core vs multi boundary rates at a glance. Full status table on Same-function.
Diagnostics · language · ISA (non-ranking)
Multi-corpus pivots from the latest envelope. Full tables on Quality EXT. Ranking remains semantic on core C PE only.
Diagnostics · readability (non-ranking)
Source similarity, AST tree-edit proxies, and readability proxies are shown here for investigation only — they never rank tools. Per-function code detail remains on Functions.