{ "generated": "2026-06-14", "reversibility": { "generated": "2026-06-13", "kernel_invariant": { "iscii": { "domain": "[a-z]^1..3", "total": 18278, "exact": 1110, "exact_pct": 6.073, "canonical": 11249, "canonical_pct": 61.544, "rev_or_canonical_pct": 67.617, "irreducible": 5425, "irreducible_pct": 29.68, "crash": 494, "crash_pct": 2.703 }, "unicode": { "domain": "[a-z]^1..3", "total": 18278, "exact": 1110, "exact_pct": 6.073, "canonical": 11249, "canonical_pct": 61.544, "rev_or_canonical_pct": 67.617, "irreducible": 5425, "irreducible_pct": 29.68, "crash": 494, "crash_pct": 2.703 } }, "corpus_unicode": { "corpus_files": [ "corp_hi.txt", "corp_bn.txt", "corp_te.txt" ], "words": 5807, "word_exact": 5776, "word_exact_pct": 99.47, "by_class": { "deva_letter": { "chars": 6942, "preserved": 6942, "pct": 100.0 }, "deva_matra_sign": { "chars": 4382, "preserved": 4377, "pct": 99.89 }, "deva_digit": { "chars": 85, "preserved": 85, "pct": 100.0 }, "deva_punct": { "chars": 172, "preserved": 172, "pct": 100.0 }, "ascii": { "chars": 1229, "preserved": 1145, "pct": 93.17 }, "foreign": { "chars": 19599, "preserved": 19599, "pct": 100.0 } } }, "notes": { "kernel": "Exhaustive [a-z]^1..N fed as substrate input to acii2rmn; canonical=stable fixpoint (phi2==phi); crash-isolated. VARIANT = the kernel currently in bindings/c (repo). Both substrate libraries classify identically.", "corpus": "Codepoint-classified round trip on the UTF-8 substrate (no uni2acii/acii2uni).", "honesty": "Numbers name the kernel variant; gaps flagged, not faked." } }, "roman": { "domain": "[a-z]^1..3", "total": 18278, "raw_kernel_rev_or_canonical_pct": 67.617, "raw_kernel_crash_pct": 2.703, "heuristic_coverage": 18278, "heuristic_coverage_pct": 100.0, "heuristic_kernel_stable": 16735, "heuristic_kernel_stable_pct": 91.56, "max_phoneme_len": 5, "token_inventory": 83 }, "perso": { "arabic_letters": 200, "base_map": 42, "base_covered": 42, "heuristic_filled": 158, "total_covered": 200, "coverage_pct": 100.0, "note": "BASE unchanged where it suffices; heuristic fills the remainder; short-vowel residue documented." }, "scripts": { "all_scripts_enumerated": 187, "families_enumerated": 9, "family_summary": [ { "family": "Brahmic", "scripts": 72, "letters": 4698, "hub_projection_pct": 74.3, "examples": [ "AHOM", "BALINESE", "BATAK", "BENGALI", "BHAIKSUKI", "BRAHMI", "BUGINESE", "BUHID" ] }, { "family": "Other / specialised", "scripts": 41, "letters": 3159, "hub_projection_pct": 14.9, "examples": [ "BOPOMOFO FINAL", "CANADIAN", "CIRCLED", "CIRCLED LATIN", "COMBINING CYRILLIC", "COMBINING DEVANAGARI", "COMBINING GLAGOLITIC", "COMBINING GRANTHA" ] }, { "family": "Semitic/abjad", "scripts": 26, "letters": 1456, "hub_projection_pct": 10.6, "examples": [ "AVESTAN", "CHORASMIAN", "ELYMAIC", "GARAY", "HANIFI ROHINGYA", "HATRAN", "HEBREW", "IMPERIAL ARAMAIC" ] }, { "family": "Alphabetic (L\u2192R/abc)", "scripts": 23, "letters": 4028, "hub_projection_pct": 11.6, "examples": [ "ADLAM", "ARMENIAN", "CAUCASIAN ALBANIAN", "COPTIC", "CYRILLIC", "DESERET", "ELBASAN", "GEORGIAN" ] }, { "family": "CJK / ideographic / syllabic E-Asia", "scripts": 11, "letters": 103807, "hub_projection_pct": 0.2, "examples": [ "BAMUM", "BOPOMOFO", "CJK COMPATIBILITY", "CJK UNIFIED", "HANGUL", "HIRAGANA", "KATAKANA", "LISU" ] }, { "family": "African", "scripts": 6, "letters": 1326, "hub_projection_pct": 7.5, "examples": [ "BASSA VAH", "ETHIOPIC", "MENDE KIKAKUI", "NKO", "TIFINAGH", "VAI" ] }, { "family": "Ancient / logo-syllabic", "scripts": 6, "letters": 514, "hub_projection_pct": 18.7, "examples": [ "CARIAN", "CYPRIOT", "LINEAR B", "LYCIAN", "LYDIAN", "MEROITIC CURSIVE" ] }, { "family": "Perso-Arabic", "scripts": 1, "letters": 514, "hub_projection_pct": 0.6, "examples": [ "ARABIC" ] }, { "family": "American", "scripts": 1, "letters": 172, "hub_projection_pct": 17.4, "examples": [ "CHEROKEE" ] } ], "deep_families": [ { "family": "Brahmic (deep)", "scripts": 72, "avg_coverage_pct": 73.1, "full_coverage_scripts": 14, "scripts_detail": [ { "script": "AHOM", "letters": 68, "mapped": 62, "coverage_pct": 91.2 }, { "script": "BALINESE", "letters": 55, "mapped": 17, "coverage_pct": 30.9 }, { "script": "BATAK", "letters": 38, "mapped": 18, "coverage_pct": 47.4 }, { "script": "BENGALI", "letters": 53, "mapped": 49, "coverage_pct": 92.5 }, { "script": "BHAIKSUKI", "letters": 92, "mapped": 92, "coverage_pct": 100.0 }, { "script": "BRAHMI", "letters": 108, "mapped": 96, "coverage_pct": 88.9 }, { "script": "BUGINESE", "letters": 23, "mapped": 19, "coverage_pct": 82.6 }, { "script": "BUHID", "letters": 18, "mapped": 17, "coverage_pct": 94.4 }, { "script": "CHAKMA", "letters": 76, "mapped": 8, "coverage_pct": 10.5 }, { "script": "CHAM", "letters": 52, "mapped": 33, "coverage_pct": 63.5 }, { "script": "DEVANAGARI", "letters": 79, "mapped": 78, "coverage_pct": 98.7 }, { "script": "DIVES AKURU", "letters": 84, "mapped": 84, "coverage_pct": 100.0 }, { "script": "DOGRA", "letters": 88, "mapped": 88, "coverage_pct": 100.0 }, { "script": "GRANTHA", "letters": 100, "mapped": 92, "coverage_pct": 92.0 }, { "script": "GUJARATI", "letters": 49, "mapped": 49, "coverage_pct": 100.0 }, { "script": "GUNJALA GONDI", "letters": 80, "mapped": 76, "coverage_pct": 95.0 }, { "script": "GURMUKHI", "letters": 48, "mapped": 46, "coverage_pct": 95.8 }, { "script": "GURUNG KHEMA", "letters": 60, "mapped": 60, "coverage_pct": 100.0 }, { "script": "HANUNOO", "letters": 18, "mapped": 17, "coverage_pct": 94.4 }, { "script": "JAVANESE", "letters": 47, "mapped": 26, "coverage_pct": 55.3 }, { "script": "KAITHI", "letters": 90, "mapped": 90, "coverage_pct": 100.0 }, { "script": "KANNADA", "letters": 53, "mapped": 50, "coverage_pct": 94.3 }, { "script": "KAWI", "letters": 94, "mapped": 90, "coverage_pct": 95.7 }, { "script": "KAYAH LI", "letters": 28, "mapped": 25, "coverage_pct": 89.3 }, { "script": "KHAROSHTHI", "letters": 74, "mapped": 66, "coverage_pct": 89.2 }, { "script": "KHMER", "letters": 35, "mapped": 15, "coverage_pct": 42.9 }, { "script": "KHOJKI", "letters": 90, "mapped": 88, "coverage_pct": 97.8 }, { "script": "KHUDAWADI", "letters": 94, "mapped": 94, "coverage_pct": 100.0 }, { "script": "KIRAT RAI", "letters": 64, "mapped": 64, "coverage_pct": 100.0 }, { "script": "LAO", "letters": 43, "mapped": 1, "coverage_pct": 2.3 }, { "script": "LEPCHA", "letters": 39, "mapped": 28, "coverage_pct": 71.8 }, { "script": "LIMBU", "letters": 39, "mapped": 34, "coverage_pct": 87.2 }, { "script": "MAHAJANI", "letters": 70, "mapped": 70, "coverage_pct": 100.0 }, { "script": "MALAYALAM", "letters": 66, "mapped": 51, "coverage_pct": 77.3 }, { "script": "MARCHEN", "letters": 60, "mapped": 50, "coverage_pct": 83.3 }, { "script": "MASARAM GONDI", "letters": 94, "mapped": 88, "coverage_pct": 93.6 }, { "script": "MEETEI MAYEK", "letters": 47, "mapped": 15, "coverage_pct": 31.9 }, { "script": "MODI", "letters": 96, "mapped": 96, "coverage_pct": 100.0 }, { "script": "MONGOLIAN", "letters": 132, "mapped": 26, "coverage_pct": 19.7 }, { "script": "MULTANI", "letters": 74, "mapped": 74, "coverage_pct": 100.0 }, { "script": "MYANMAR", "letters": 115, "mapped": 45, "coverage_pct": 39.1 }, { "script": "NAG MUNDARI", "letters": 54, "mapped": 10, "coverage_pct": 18.5 }, { "script": "NANDINAGARI", "letters": 94, "mapped": 94, "coverage_pct": 100.0 }, { "script": "NEW TAI LUE", "letters": 51, "mapped": 0, "coverage_pct": 0.0 }, { "script": "NEWA", "letters": 108, "mapped": 94, "coverage_pct": 87.0 }, { "script": "OL ONAL", "letters": 60, "mapped": 12, "coverage_pct": 20.0 }, { "script": "ORIYA", "letters": 52, "mapped": 51, "coverage_pct": 98.1 }, { "script": "PAU CIN HAU", "letters": 74, "mapped": 52, "coverage_pct": 70.3 }, { "script": "PHAGS-PA", "letters": 48, "mapped": 36, "coverage_pct": 75.0 }, { "script": "REJANG", "letters": 23, "mapped": 18, "coverage_pct": 78.3 }, { "script": "SAURASHTRA", "letters": 50, "mapped": 48, "coverage_pct": 96.0 }, { "script": "SHARADA", "letters": 96, "mapped": 96, "coverage_pct": 100.0 }, { "script": "SIDDHAM", "letters": 102, "mapped": 94, "coverage_pct": 92.2 }, { "script": "SINHALA", "letters": 59, "mapped": 0, "coverage_pct": 0.0 }, { "script": "SOYOMBO", "letters": 82, "mapped": 72, "coverage_pct": 87.8 }, { "script": "SUNDANESE", "letters": 37, "mapped": 28, "coverage_pct": 75.7 }, { "script": "SUNUWAR", "letters": 66, "mapped": 4, "coverage_pct": 6.1 }, { "script": "SYLOTI NAGRI", "letters": 32, "mapped": 5, "coverage_pct": 15.6 }, { "script": "TAGALOG", "letters": 19, "mapped": 17, "coverage_pct": 89.5 }, { "script": "TAGBANWA", "letters": 16, "mapped": 15, "coverage_pct": 93.8 }, { "script": "TAI LE", "letters": 35, "mapped": 23, "coverage_pct": 65.7 }, { "script": "TAI THAM", "letters": 53, "mapped": 14, "coverage_pct": 26.4 }, { "script": "TAI VIET", "letters": 48, "mapped": 0, "coverage_pct": 0.0 }, { "script": "TAKRI", "letters": 88, "mapped": 86, "coverage_pct": 97.7 }, { "script": "TAMIL", "letters": 35, "mapped": 33, "coverage_pct": 94.3 }, { "script": "TANGSA", "letters": 158, "mapped": 50, "coverage_pct": 31.6 }, { "script": "TELUGU", "letters": 56, "mapped": 50, "coverage_pct": 89.3 }, { "script": "TIBETAN", "letters": 45, "mapped": 35, "coverage_pct": 77.8 }, { "script": "TIRHUTA", "letters": 94, "mapped": 94, "coverage_pct": 100.0 }, { "script": "TOTO", "letters": 60, "mapped": 42, "coverage_pct": 70.0 }, { "script": "WANCHO", "letters": 88, "mapped": 62, "coverage_pct": 70.5 }, { "script": "ZANABAZAR SQUARE", "letters": 82, "mapped": 70, "coverage_pct": 85.4 } ] }, { "family": "Perso-Arabic (deep)", "letters": 200, "mapped": 200, "coverage_pct": 100.0, "note": "shared Arabic block; total via perso2deva_heur (126); 41 languages live in the demo" }, { "family": "NW-Semitic (deep)", "method": "attested consonant correspondence (not name-projection)", "scripts_detail": [ { "script": "HEBREW", "letters": 27, "mapped": 27, "coverage_pct": 100.0 }, { "script": "SYRIAC", "letters": 35, "mapped": 22, "coverage_pct": 62.9 } ], "derivable": [ "Aramaic", "Phoenician", "Samaritan", "Mandaic", "Nabataean", "Palmyrene" ], "note": "Hebrew/Syriac by attested consonant correspondence to the Devanagari skeleton; the rest share the abjad skeleton" } ], "families": [ { "family": "Brahmi-derived", "scripts": 72, "avg_coverage_pct": 73.1, "full_coverage_scripts": 14, "scripts_detail": [ { "script": "AHOM", "letters": 68, "mapped": 62, "coverage_pct": 91.2 }, { "script": "BALINESE", "letters": 55, "mapped": 17, "coverage_pct": 30.9 }, { "script": "BATAK", "letters": 38, "mapped": 18, "coverage_pct": 47.4 }, { "script": "BENGALI", "letters": 53, "mapped": 49, "coverage_pct": 92.5 }, { "script": "BHAIKSUKI", "letters": 92, "mapped": 92, "coverage_pct": 100.0 }, { "script": "BRAHMI", "letters": 108, "mapped": 96, "coverage_pct": 88.9 }, { "script": "BUGINESE", "letters": 23, "mapped": 19, "coverage_pct": 82.6 }, { "script": "BUHID", "letters": 18, "mapped": 17, "coverage_pct": 94.4 }, { "script": "CHAKMA", "letters": 76, "mapped": 8, "coverage_pct": 10.5 }, { "script": "CHAM", "letters": 52, "mapped": 33, "coverage_pct": 63.5 }, { "script": "DEVANAGARI", "letters": 79, "mapped": 78, "coverage_pct": 98.7 }, { "script": "DIVES AKURU", "letters": 84, "mapped": 84, "coverage_pct": 100.0 }, { "script": "DOGRA", "letters": 88, "mapped": 88, "coverage_pct": 100.0 }, { "script": "GRANTHA", "letters": 100, "mapped": 92, "coverage_pct": 92.0 }, { "script": "GUJARATI", "letters": 49, "mapped": 49, "coverage_pct": 100.0 }, { "script": "GUNJALA GONDI", "letters": 80, "mapped": 76, "coverage_pct": 95.0 }, { "script": "GURMUKHI", "letters": 48, "mapped": 46, "coverage_pct": 95.8 }, { "script": "GURUNG KHEMA", "letters": 60, "mapped": 60, "coverage_pct": 100.0 }, { "script": "HANUNOO", "letters": 18, "mapped": 17, "coverage_pct": 94.4 }, { "script": "JAVANESE", "letters": 47, "mapped": 26, "coverage_pct": 55.3 }, { "script": "KAITHI", "letters": 90, "mapped": 90, "coverage_pct": 100.0 }, { "script": "KANNADA", "letters": 53, "mapped": 50, "coverage_pct": 94.3 }, { "script": "KAWI", "letters": 94, "mapped": 90, "coverage_pct": 95.7 }, { "script": "KAYAH LI", "letters": 28, "mapped": 25, "coverage_pct": 89.3 }, { "script": "KHAROSHTHI", "letters": 74, "mapped": 66, "coverage_pct": 89.2 }, { "script": "KHMER", "letters": 35, "mapped": 15, "coverage_pct": 42.9 }, { "script": "KHOJKI", "letters": 90, "mapped": 88, "coverage_pct": 97.8 }, { "script": "KHUDAWADI", "letters": 94, "mapped": 94, "coverage_pct": 100.0 }, { "script": "KIRAT RAI", "letters": 64, "mapped": 64, "coverage_pct": 100.0 }, { "script": "LAO", "letters": 43, "mapped": 1, "coverage_pct": 2.3 }, { "script": "LEPCHA", "letters": 39, "mapped": 28, "coverage_pct": 71.8 }, { "script": "LIMBU", "letters": 39, "mapped": 34, "coverage_pct": 87.2 }, { "script": "MAHAJANI", "letters": 70, "mapped": 70, "coverage_pct": 100.0 }, { "script": "MALAYALAM", "letters": 66, "mapped": 51, "coverage_pct": 77.3 }, { "script": "MARCHEN", "letters": 60, "mapped": 50, "coverage_pct": 83.3 }, { "script": "MASARAM GONDI", "letters": 94, "mapped": 88, "coverage_pct": 93.6 }, { "script": "MEETEI MAYEK", "letters": 47, "mapped": 15, "coverage_pct": 31.9 }, { "script": "MODI", "letters": 96, "mapped": 96, "coverage_pct": 100.0 }, { "script": "MONGOLIAN", "letters": 132, "mapped": 26, "coverage_pct": 19.7 }, { "script": "MULTANI", "letters": 74, "mapped": 74, "coverage_pct": 100.0 }, { "script": "MYANMAR", "letters": 115, "mapped": 45, "coverage_pct": 39.1 }, { "script": "NAG MUNDARI", "letters": 54, "mapped": 10, "coverage_pct": 18.5 }, { "script": "NANDINAGARI", "letters": 94, "mapped": 94, "coverage_pct": 100.0 }, { "script": "NEW TAI LUE", "letters": 51, "mapped": 0, "coverage_pct": 0.0 }, { "script": "NEWA", "letters": 108, "mapped": 94, "coverage_pct": 87.0 }, { "script": "OL ONAL", "letters": 60, "mapped": 12, "coverage_pct": 20.0 }, { "script": "ORIYA", "letters": 52, "mapped": 51, "coverage_pct": 98.1 }, { "script": "PAU CIN HAU", "letters": 74, "mapped": 52, "coverage_pct": 70.3 }, { "script": "PHAGS-PA", "letters": 48, "mapped": 36, "coverage_pct": 75.0 }, { "script": "REJANG", "letters": 23, "mapped": 18, "coverage_pct": 78.3 }, { "script": "SAURASHTRA", "letters": 50, "mapped": 48, "coverage_pct": 96.0 }, { "script": "SHARADA", "letters": 96, "mapped": 96, "coverage_pct": 100.0 }, { "script": "SIDDHAM", "letters": 102, "mapped": 94, "coverage_pct": 92.2 }, { "script": "SINHALA", "letters": 59, "mapped": 0, "coverage_pct": 0.0 }, { "script": "SOYOMBO", "letters": 82, "mapped": 72, "coverage_pct": 87.8 }, { "script": "SUNDANESE", "letters": 37, "mapped": 28, "coverage_pct": 75.7 }, { "script": "SUNUWAR", "letters": 66, "mapped": 4, "coverage_pct": 6.1 }, { "script": "SYLOTI NAGRI", "letters": 32, "mapped": 5, "coverage_pct": 15.6 }, { "script": "TAGALOG", "letters": 19, "mapped": 17, "coverage_pct": 89.5 }, { "script": "TAGBANWA", "letters": 16, "mapped": 15, "coverage_pct": 93.8 }, { "script": "TAI LE", "letters": 35, "mapped": 23, "coverage_pct": 65.7 }, { "script": "TAI THAM", "letters": 53, "mapped": 14, "coverage_pct": 26.4 }, { "script": "TAI VIET", "letters": 48, "mapped": 0, "coverage_pct": 0.0 }, { "script": "TAKRI", "letters": 88, "mapped": 86, "coverage_pct": 97.7 }, { "script": "TAMIL", "letters": 35, "mapped": 33, "coverage_pct": 94.3 }, { "script": "TANGSA", "letters": 158, "mapped": 50, "coverage_pct": 31.6 }, { "script": "TELUGU", "letters": 56, "mapped": 50, "coverage_pct": 89.3 }, { "script": "TIBETAN", "letters": 45, "mapped": 35, "coverage_pct": 77.8 }, { "script": "TIRHUTA", "letters": 94, "mapped": 94, "coverage_pct": 100.0 }, { "script": "TOTO", "letters": 60, "mapped": 42, "coverage_pct": 70.0 }, { "script": "WANCHO", "letters": 88, "mapped": 62, "coverage_pct": 70.5 }, { "script": "ZANABAZAR SQUARE", "letters": 82, "mapped": 70, "coverage_pct": 85.4 } ] }, { "family": "Perso-Arabic", "letters": 200, "mapped": 200, "coverage_pct": 100.0, "note": "shared Arabic block; total via perso2deva_heur (126); 41 languages live in the demo" }, { "family": "NW-Semitic (deep)", "method": "attested consonant correspondence (not name-projection)", "scripts_detail": [ { "script": "HEBREW", "letters": 27, "mapped": 27, "coverage_pct": 100.0 }, { "script": "SYRIAC", "letters": 35, "mapped": 22, "coverage_pct": 62.9 } ], "derivable": [ "Aramaic", "Phoenician", "Samaritan", "Mandaic", "Nabataean", "Palmyrene" ], "note": "Hebrew/Syriac by attested consonant correspondence to the Devanagari skeleton; the rest share the abjad skeleton" } ], "total_scripts_tabled": 74, "note": "Every script Unicode encodes is enumerated and family-classified. The Brahmic core, the Perso-Arabic block and NW-Semitic are projected to the Devanagari hub and measured; every other script is a pending Layer-1 table row at ZERO kernel cost. N tables + 1 kernel, never N^2." }, "langspec": { "generator": "langspec/tools/langspec_gen.py", "demonstrations": [ { "name": "hindi", "constructs": 10, "verify": { "compiled": "OK", "ran": "OK", "src": "hindi.c", "constructs": 5 } }, { "name": "bhojpuri", "constructs": 10, "verify": { "compiled": "OK", "ran": "OK", "src": "bhojpuri.c", "constructs": 5 } }, { "name": "zephyr_idiolect", "constructs": 10, "verify": { "compiled": "OK", "ran": "OK", "src": "zephyr_idiolect.c", "constructs": 5 } } ], "claim": "Anyone defines keywords (language, dialect, or personal idiolect) as a CSV and gets a localized standard document plus a compile-verified C program \u2014 with zero change to the kernel, the toolchain, or the existing programming-language standards.", "software_engineering_process": [ "Localization is a PRESENTATION + ONTOLOGY problem, not a compiler-reinvention problem.", "Keywords are TRANSLATED; execution SEMANTICS are PRESERVED (construct id -> C/Python/VHDL/...).", "Surface keywords map to reversible canonical Romenagri ASCII-7 identifiers (C-identifier-legal).", "Native .uhin -> hincc -> Romenagri identifiers -> UNMODIFIED GCC/LLVM; debuggers/SCM/IDEs/libs untouched.", "Adaptation cost of the linguistic layer across any future execution paradigm = 0." ], "standard_register_process": [ "1. IDIOLECT \u2014 an individual publishes a CSV; langspec_gen yields a personal standard + verified C.", "2. COMMUNITY REGISTER \u2014 a language community converges on a shared CSV; conflicts are reviewed by speakers.", "3. CANONICALISATION \u2014 the construct->keyword table is frozen and versioned in langspec/output/.", "4. STANDARDS-BODY ADOPTION \u2014 the frozen register is submitted to Unicode UTC / W3C i18n / ISO-IEC JTC1 / IEEE SA as a localization of an existing standard (not a new language).", "5. The kernel and toolchain never change at any step \u2014 only registry rows are added." ], "scaling": "7,168+ languages x 600+ scripts: each new language adds CSV rows, each new script adds one projection table; the fixed kernel and the unmodified ecosystem absorb the entire AGI stack (L0-L9) at zero engineering-overhead cost." }, "fltr": { "filter": "fltr_ur_hi / fltr_hi_ur (Urdu<->Devanagari spoke)", "T1_urdu_deva_urdu": { "desc": "abjad->abugida->abjad; lossy by orthography (short-vowel underspecification)", "by_class": { "arabic": { "orig": 3018, "kept": 0, "pct": 0.0 }, "ascii": { "orig": 432, "kept": 0, "pct": 0.0 }, "other": { "orig": 45, "kept": 0, "pct": 0.0 }, "devanagari": { "orig": 4, "kept": 0, "pct": 0.0 }, "telugu": { "orig": 5, "kept": 0, "pct": 0.0 } } }, "T2_note": "hub (Deva->kernel->Deva) faithfulness is measured by 120 corpus test on the Devanagari side", "T3_full_chain": { "desc": "Urdu->Deva->[kernel]->Deva->Urdu; kernel should add ~0 loss over T1", "by_class": { "arabic": { "orig": 3018, "kept": 0, "pct": 0.0 }, "ascii": { "orig": 432, "kept": 0, "pct": 0.0 }, "other": { "orig": 45, "kept": 0, "pct": 0.0 }, "devanagari": { "orig": 4, "kept": 0, "pct": 0.0 }, "telugu": { "orig": 5, "kept": 0, "pct": 0.0 } } }, "deva_hub_kernel_fidelity": { "desc": "Deva vs Deva-after-kernel (isolates kernel's own loss)", "by_class": { "devanagari": { "orig": 3129, "kept": 3129, "pct": 100.0 }, "ascii": { "orig": 480, "kept": 480, "pct": 100.0 }, "other": { "orig": 43, "kept": 43, "pct": 100.0 }, "telugu": { "orig": 5, "kept": 5, "pct": 100.0 } } } } }