oruko_eval.ipynb

{"cells": [{"cell_type": "markdown", "metadata": {"id": "yxGv0qFlYy4u"}, "source": ["# ORÚKỌ-mini: identity preservation of African names in `Qwen/Qwen3.5-4B`\n", "\n", "This notebook runs the whole evaluation end to end:\n", "\n", "1. writes the dataset (Yorùbá, Igbo, Hausa names + high-resource controls + minimal pairs)\n", "2. loads **Qwen/Qwen3.5-4B** (non-thinking mode, greedy decoding)\n", "3. runs 363 prompts across 6 tasks, plus a tokenizer and a log-probability probe\n", "4. scores every output for identity preservation, draws the figures, fills the README\n", "5. publishes everything to a **public Hugging Face Storage Bucket**\n", "\n"]}, {"cell_type": "code", "metadata": {"id": "m1mTWz-SYy4v", "executionInfo": {"status": "ok", "timestamp": 1792105829578, "user_tz": -60, "elapsed": 4518, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}}, "execution_count": 1, "outputs": [], "source": ["!pip -q install -U transformers accelerate \"huggingface_hub>=1.5.0\" matplotlib pandas scipy\n", "# If loading Qwen3.5 fails with an unknown architecture, install transformers from main instead:\n", "# !pip -q install -U \"git+https://github.com/huggingface/transformers.git@main\""]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/"}, "id": "LFnL9TdRYy4w", "executionInfo": {"status": "ok", "timestamp": 1792105831142, "user_tz": -60, "elapsed": 1561, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "9a57573f-600b-467d-9e3c-eb76bca2c3c7"}, "execution_count": 2, "outputs": [{"output_type": "stream", "name": "stderr", "text": ["Note: Environment variable`HF_TOKEN` is set and is the current active token independently from the token you've just configured.\n", "WARNING:huggingface_hub._login:Note: Environment variable`HF_TOKEN` is set and is the current active token independently from the token you've just configured.\n"]}, {"output_type": "stream", "name": "stdout", "text": ["Logged in as coderGit\n"]}], "source": ["import os\n", "MODEL_ID = \"Qwen/Qwen3.5-4B\"\n", "BUCKET_NAME = \"oruko-mini\"          # bucket will be https://huggingface.co/buckets/<you>/oruko-mini\n", "COPY_MODEL_TO_BUCKET = False         # server-side copy of the exact weights evaluated (no download/upload)\n", "BATCH_SIZE = 8                      # lower to 4 on a T4 if you hit out-of-memory\n", "WORKDIR = \"/content/oruko-mini\"\n", "\n", "os.makedirs(WORKDIR, exist_ok=True)\n", "os.chdir(WORKDIR)\n", "try:\n", "    from google.colab import userdata\n", "    os.environ[\"HF_TOKEN\"] = userdata.get(\"HF_TOKEN\")\n", "except Exception as e:\n", "    print(\"Colab secret not found; make sure HF_TOKEN is set in the environment.\", e)\n", "\n", "from huggingface_hub import login, whoami\n", "login(token=os.environ.get(\"HF_TOKEN\"))\n", "HF_USER = whoami()[\"name\"]\n", "print(\"Logged in as\", HF_USER)"]}, {"cell_type": "code", "source": ["%%writefile oruko_core.py\n", "\"\"\"\n", "ORÚKỌ-mini: does a small frontier LLM preserve the identity of African names?\n", "\n", "Core module: dataset, prompt construction, and identity-preservation scoring.\n", "Everything here is pure Python (no GPU needed) so it can be unit-tested anywhere.\n", "\"\"\"\n", "import csv, json, os, re, difflib, unicodedata as ud\n", "from collections import defaultdict\n", "\n", "# ---------------------------------------------------------------------------\n", "# 1. Unicode helpers\n", "# ---------------------------------------------------------------------------\n", "# Letters that are *distinct letters*, not accented forms. NFD does not\n", "# decompose them, so a naive \"strip combining marks\" keeps them, while an\n", "# ASCII fold turns them into a different letter (or drops them).\n", "SPECIAL_FOLD = {\n", "    \"ɓ\": \"b\", \"Ɓ\": \"B\", \"ɗ\": \"d\", \"Ɗ\": \"D\", \"ƙ\": \"k\", \"Ƙ\": \"K\",\n", "    \"ƴ\": \"y\", \"Ƴ\": \"Y\",                                  # Hausa hooked letters\n", "    \"ø\": \"o\", \"Ø\": \"O\", \"ł\": \"l\", \"Ł\": \"L\", \"đ\": \"d\", \"Đ\": \"D\",  # controls\n", "}\n", "HOOKED = set(\"ɓƁɗƊƙƘƴƳ\")\n", "TONE_MARKS = {\"\\u0300\", \"\\u0301\", \"\\u0302\", \"\\u0304\", \"\\u030c\"}  # grave, acute, circumflex, macron, caron\n", "MARK_NAMES = {\n", "    \"\\u0300\": \"grave(low)\", \"\\u0301\": \"acute(high)\", \"\\u0323\": \"dot-below\",\n", "    \"\\u0307\": \"dot-above\", \"\\u0302\": \"circumflex\", \"\\u0304\": \"macron\",\n", "    \"\\u030c\": \"caron\", \"\\u0308\": \"diaeresis\", \"\\u0327\": \"cedilla\",\n", "    \"\\u031b\": \"horn\", \"\\u0303\": \"tilde\", \"\\u030a\": \"ring\",\n", "}\n", "\n", "\n", "def nfc(s):\n", "    return ud.normalize(\"NFC\", s)\n", "\n", "\n", "def units(s):\n", "    \"\"\"Split a string into [base_char, [combining marks]] units (via NFD).\"\"\"\n", "    out = []\n", "    for ch in ud.normalize(\"NFD\", s):\n", "        if ud.combining(ch) and out:\n", "            out[-1][1].append(ch)\n", "        else:\n", "            out.append([ch, []])\n", "    return out\n", "\n", "\n", "def strip_marks(s):\n", "    \"\"\"Remove every diacritic and fold special letters -> plain letters.\n", "    This is what most 'text normalisation' pipelines effectively do.\"\"\"\n", "    return nfc(\"\".join(SPECIAL_FOLD.get(b, b) for b, _ in units(s)))\n", "\n", "\n", "def ascii_ignore(s):\n", "    \"\"\"What `s.encode('ascii','ignore')` does: special letters vanish entirely.\"\"\"\n", "    return nfc(s).encode(\"ascii\", \"ignore\").decode()\n", "\n", "\n", "def features(unit, ignore_tone=False):\n", "    \"\"\"Identity-bearing features of one character unit.\"\"\"\n", "    b, marks = unit\n", "    f = {m for m in marks if not (ignore_tone and m in TONE_MARKS)}\n", "    if b in SPECIAL_FOLD:\n", "        f.add(\"SPECIAL:\" + b.lower())\n", "    return f\n", "\n", "\n", "def n_identity_marks(s, ignore_tone=False):\n", "    return sum(len(features(u, ignore_tone)) for u in units(s))\n", "\n", "\n", "def diacritic_density(s, ignore_tone=False):\n", "    letters = [u for u in units(s) if u[0].isalpha()]\n", "    return n_identity_marks(s, ignore_tone) / max(1, len(letters))\n", "\n", "\n", "def yoruba_tone_pattern(s):\n", "    \"\"\"Tone per vowel: acute=H, grave=L, unmarked=M (standard Yorùbá orthography).\"\"\"\n", "    pat = []\n", "    for b, marks in units(s):\n", "        if b.lower() in \"aeiou\":\n", "            pat.append(\"H\" if \"\\u0301\" in marks else \"L\" if \"\\u0300\" in marks else \"M\")\n", "    return \"-\".join(pat)\n", "\n", "\n", "# ---------------------------------------------------------------------------\n", "# 2. Dataset\n", "# ---------------------------------------------------------------------------\n", "# conf: \"high\" = standard, widely attested spelling; \"check\" = please confirm\n", "# with a native speaker before release (bearer-defined correctness).\n", "# kw: list of keyword groups; a meaning answer is correct if every group has\n", "# at least one keyword present (case-insensitive substring).\n", "NAMES = [\n", "    # ---- Yorùbá (tone marks + under-dots) ----\n", "    (\"yo01\", \"yoruba\", \"Olúwadámilárè\", \"God has vindicated me\", [[\"god\", \"lord\"], [\"justif\", \"vindicat\", \"right\", \"innocen\"]], \"check\"),\n", "    (\"yo02\", \"yoruba\", \"Adéṣọlá\", \"the crown makes wealth/honour\", [[\"crown\", \"royal\"], [\"wealth\", \"honou\", \"honor\", \"riches\", \"prosper\"]], \"high\"),\n", "    (\"yo03\", \"yoruba\", \"Ọláyínká\", \"wealth/honour surrounds me\", [[\"wealth\", \"honou\", \"honor\", \"riches\"], [\"surround\", \"encircl\", \"around\"]], \"high\"),\n", "    (\"yo04\", \"yoruba\", \"Fọláṣadé\", \"uses wealth/honour as a crown\", [[\"wealth\", \"honou\", \"honor\", \"riches\"], [\"crown\"]], \"high\"),\n", "    (\"yo05\", \"yoruba\", \"Olúwaṣeun\", \"thank God / God has done it\", [[\"god\", \"lord\"], [\"thank\", \"done\", \"did\"]], \"high\"),\n", "    (\"yo06\", \"yoruba\", \"Ayọ̀délé\", \"joy has come home\", [[\"joy\", \"happiness\"], [\"home\", \"arriv\", \"come\", \"came\"]], \"high\"),\n", "    (\"yo07\", \"yoruba\", \"Ìyábọ̀dé\", \"mother has returned\", [[\"mother\"], [\"return\", \"come back\", \"came back\", \"has come\"]], \"high\"),\n", "    (\"yo08\", \"yoruba\", \"Bọ́lájí\", \"one who wakes up with wealth\", [[\"wealth\", \"honou\", \"honor\", \"riches\"]], \"high\"),\n", "    (\"yo09\", \"yoruba\", \"Fúnmiláyọ̀\", \"give me joy\", [[\"joy\", \"happiness\"], [\"give\"]], \"high\"),\n", "    (\"yo10\", \"yoruba\", \"Adébáyọ̀\", \"the crown meets joy\", [[\"crown\", \"royal\"], [\"joy\", \"happiness\"]], \"high\"),\n", "    (\"yo11\", \"yoruba\", \"Ọlámidé\", \"my wealth/honour has come\", [[\"wealth\", \"honou\", \"honor\", \"riches\"], [\"come\", \"arriv\", \"came\"]], \"high\"),\n", "    (\"yo12\", \"yoruba\", \"Ìfẹ́olúwa\", \"the love of God\", [[\"love\"], [\"god\", \"lord\"]], \"high\"),\n", "    (\"yo13\", \"yoruba\", \"Táíwò\", \"first-born of twins\", [[\"twin\"]], \"high\"),\n", "    (\"yo14\", \"yoruba\", \"Kẹ́hìndé\", \"second-born of twins\", [[\"twin\"]], \"high\"),\n", "    (\"yo15\", \"yoruba\", \"Ìdòwú\", \"child born after twins\", [[\"twin\"]], \"high\"),\n", "    (\"yo16\", \"yoruba\", \"Babátúndé\", \"father has returned\", [[\"father\"], [\"return\", \"come back\", \"came back\", \"again\"]], \"high\"),\n", "    (\"yo17\", \"yoruba\", \"Olúṣẹ́gun\", \"God has conquered\", [[\"god\", \"lord\"], [\"conquer\", \"victor\", \"triumph\", \"win\", \"defeat\", \"overcome\"]], \"high\"),\n", "    (\"yo18\", \"yoruba\", \"Àbẹ̀ní\", \"we asked for her and got her\", [[\"ask\", \"beg\", \"pray\", \"request\"]], \"check\"),\n", "    (\"yo19\", \"yoruba\", \"Yéwándé\", \"mother has returned\", [[\"mother\"], [\"return\", \"come back\", \"came back\", \"has come\"]], \"high\"),\n", "    (\"yo20\", \"yoruba\", \"Adéwálé\", \"the crown has come home\", [[\"crown\", \"royal\"], [\"home\", \"come\", \"arriv\", \"came\"]], \"high\"),\n", "    # ---- Igbo (dotted vowels ị ọ ụ; tone usually unwritten in names) ----\n", "    (\"ig01\", \"igbo\", \"Chiọma\", \"good God / good personal spirit\", [[\"god\", \"chi\", \"spirit\", \"destiny\"], [\"good\"]], \"high\"),\n", "    (\"ig02\", \"igbo\", \"Ụchenna\", \"the father's will / thought\", [[\"father\"], [\"will\", \"thought\", \"mind\", \"wish\", \"intention\"]], \"high\"),\n", "    (\"ig03\", \"igbo\", \"Obiọma\", \"kind heart\", [[\"heart\"], [\"good\", \"kind\"]], \"high\"),\n", "    (\"ig04\", \"igbo\", \"Ifeọma\", \"good thing\", [[\"good\", \"beautiful\"], [\"thing\"]], \"high\"),\n", "    (\"ig05\", \"igbo\", \"Chinọnso\", \"God is near\", [[\"god\"], [\"near\", \"close\"]], \"check\"),\n", "    (\"ig06\", \"igbo\", \"Ụzọamaka\", \"the road/path is good\", [[\"road\", \"path\", \"way\", \"journey\"], [\"good\", \"beautiful\"]], \"check\"),\n", "    (\"ig07\", \"igbo\", \"Ifeanyịchukwu\", \"nothing is impossible for God\", [[\"god\"], [\"impossible\", \"nothing\", \"beyond\", \"too difficult\", \"too hard\", \"defeat\"]], \"check\"),\n", "    (\"ig08\", \"igbo\", \"Ngọzi\", \"blessing\", [[\"bless\"]], \"check\"),\n", "    (\"ig09\", \"igbo\", \"Chukwuebuka\", \"God is great (no dotted letters: over-correction control)\", [[\"god\"], [\"great\", \"big\", \"mighty\"]], \"high\"),\n", "    (\"ig10\", \"igbo\", \"Nkechi\", \"God's own (no dotted letters: over-correction control)\", [[\"god\"], [\"own\", \"belong\", \"gift\", \"creat\"]], \"high\"),\n", "    # ---- Hausa (hooked letters ɓ ɗ ƙ ƴ are distinct letters) ----\n", "    (\"ha01\", \"hausa\", \"Ɗanjuma\", \"boy born on Friday\", [[\"friday\", \"juma\"]], \"high\"),\n", "    (\"ha02\", \"hausa\", \"Ɗanladi\", \"boy born on Sunday\", [[\"sunday\", \"lahadi\"]], \"high\"),\n", "    (\"ha03\", \"hausa\", \"Ɗanazumi\", \"boy born during the fast (Ramadan)\", [[\"fast\", \"ramadan\", \"azumi\"]], \"check\"),\n", "    (\"ha04\", \"hausa\", \"Ladi\", \"girl born on Sunday (no hooked letters: control)\", [[\"sunday\", \"lahadi\"]], \"high\"),\n", "    # ---- High-resource controls (diacritics in well-served languages) ----\n", "    (\"ct01\", \"control\", \"José\", \"Spanish\", None, \"high\"),\n", "    (\"ct02\", \"control\", \"François\", \"French\", None, \"high\"),\n", "    (\"ct03\", \"control\", \"Søren\", \"Danish\", None, \"high\"),\n", "    (\"ct04\", \"control\", \"Björn\", \"Swedish\", None, \"high\"),\n", "    (\"ct05\", \"control\", \"Łukasz\", \"Polish\", None, \"high\"),\n", "    (\"ct06\", \"control\", \"Jürgen\", \"German\", None, \"high\"),\n", "    (\"ct07\", \"control\", \"Dvořák\", \"Czech\", None, \"high\"),\n", "    (\"ct08\", \"control\", \"Nguyễn Thị Hương\", \"Vietnamese (tonal, like Yorùbá)\", None, \"high\"),\n", "    (\"ct09\", \"control\", \"Zhāng Wěi\", \"Mandarin pinyin (tonal)\", None, \"high\"),\n", "]\n", "CONTROL_LANG = {\"ct01\": \"Spanish\", \"ct02\": \"French\", \"ct03\": \"Danish\", \"ct04\": \"Swedish\",\n", "                \"ct05\": \"Polish\", \"ct06\": \"German\", \"ct07\": \"Czech\", \"ct08\": \"Vietnamese\",\n", "                \"ct09\": \"Mandarin (pinyin)\"}\n", "NO_RESTORE = {\"ct09\"}  # pinyin restoration is genuinely ambiguous\n", "\n", "# Minimal pairs: the same bare letters, different words.\n", "MINIMAL_PAIRS = [\n", "    # set, lang, word, gloss, keywords\n", "    (\"oko\", \"yoruba\", \"oko\", \"farm\", [\"farm\"]),\n", "    (\"oko\", \"yoruba\", \"ọkọ\", \"husband\", [\"husband\"]),\n", "    (\"oko\", \"yoruba\", \"ọkọ̀\", \"vehicle / boat\", [\"vehicle\", \"car\", \"boat\", \"canoe\", \"ship\", \"lorry\", \"transport\", \"bus\"]),\n", "    (\"oko\", \"yoruba\", \"ọkọ́\", \"hoe\", [\"hoe\"]),\n", "    (\"igba\", \"yoruba\", \"igbá\", \"calabash\", [\"calabash\", \"gourd\"]),\n", "    (\"igba\", \"yoruba\", \"ìgbà\", \"time / period\", [\"time\", \"period\", \"season\", \"era\"]),\n", "    (\"igba\", \"yoruba\", \"igba\", \"two hundred\", [\"200\", \"two hundred\", \"two-hundred\"]),\n", "    (\"igba\", \"yoruba\", \"ìgbá\", \"garden egg\", [\"garden egg\", \"eggplant\", \"aubergine\", \"garden-egg\"]),\n", "    (\"akwa\", \"igbo\", \"àkwà\", \"bed\", [\"bed\", \"bridge\"]),\n", "    (\"akwa\", \"igbo\", \"ákwà\", \"cloth\", [\"cloth\", \"fabric\", \"garment\", \"wrapper\", \"clothes\", \"clothing\"]),\n", "    (\"akwa\", \"igbo\", \"ákwá\", \"cry\", [\"cry\", \"weep\", \"tears\"]),\n", "    (\"akwa\", \"igbo\", \"àkwá\", \"egg\", [\"egg\"]),\n", "    (\"kara\", \"hausa\", \"kara\", \"stalk / reed\", [\"stalk\", \"reed\", \"stem\", \"cane\"]),\n", "    (\"kara\", \"hausa\", \"ƙara\", \"increase / complaint\", [\"increase\", \"add\", \"complain\", \"more\", \"noise\", \"sound\", \"lawsuit\"]),\n", "    (\"dari\", \"hausa\", \"dari\", \"cold (weather)\", [\"cold\", \"chill\", \"harmattan\"]),\n", "    (\"dari\", \"hausa\", \"ɗari\", \"hundred\", [\"hundred\", \"100\"]),\n", "]\n", "\n", "LANG_LABEL = {\"yoruba\": \"Yoruba\", \"igbo\": \"Igbo\", \"hausa\": \"Hausa\"}\n", "IGNORE_TONE = {\"yoruba\": False, \"igbo\": True, \"hausa\": True, \"control\": False}\n", "\n", "\n", "def name_rows():\n", "    rows = []\n", "    for nid, lang, name, meaning, kw, conf in NAMES:\n", "        name = nfc(name)\n", "        it = IGNORE_TONE[lang]\n", "        rows.append(dict(\n", "            id=nid, lang=lang,\n", "            sublang=CONTROL_LANG.get(nid, LANG_LABEL.get(lang, \"\")),\n", "            name=name, stripped=strip_marks(name), ascii_ignore=ascii_ignore(name),\n", "            n_marks=n_identity_marks(name, it), density=round(diacritic_density(name, it), 3),\n", "            has_hooked=any(c in HOOKED for c in name),\n", "            n_letters=sum(1 for u in units(name) if u[0].isalpha()),\n", "            tone_pattern=yoruba_tone_pattern(name) if lang == \"yoruba\" else \"\",\n", "            meaning=meaning if lang != \"control\" else \"\",\n", "            meaning_keywords=json.dumps(kw) if kw else \"\",\n", "            confidence=conf,\n", "        ))\n", "    return rows\n", "\n", "\n", "def pair_rows():\n", "    return [dict(set=s, lang=l, word=nfc(w), bare=strip_marks(w), gloss=g, keywords=json.dumps(k))\n", "            for s, l, w, g, k in MINIMAL_PAIRS]\n", "\n", "\n", "# ---------------------------------------------------------------------------\n", "# 3. Prompts\n", "# ---------------------------------------------------------------------------\n", "COPY_TEMPLATES = {\n", "    \"c1_repeat\": \"Repeat the following name exactly as written, with no changes: {name}\",\n", "    \"c2_welcome\": \"Write a one-sentence welcome message for a new colleague named {name}.\",\n", "    \"c3_json\": 'Create a JSON object for a hospital patient record with the fields \"full_name\" and \"greeting\". The patient\\'s name is {name}. Output only the JSON.',\n", "    \"c4_translate\": 'Translate into French: \"{name} will arrive tomorrow morning.\"',\n", "    \"c5_upper\": \"Write the name {name} in capital letters, as it would appear on a boarding pass.\",\n", "}\n", "\n", "\n", "def restore_prompt(r):\n", "    bare = r[\"stripped\"]\n", "    if r[\"lang\"] == \"yoruba\":\n", "        return (f'The Yoruba name \"{bare}\" has been written without its tone marks and under-dots. '\n", "                \"Rewrite it in correct standard Yoruba orthography with all diacritics (tone marks and under-dots). \"\n", "                \"Reply with only the name.\")\n", "    if r[\"lang\"] == \"igbo\":\n", "        return (f'The Igbo name \"{bare}\" has been written without diacritics. Rewrite it in standard Igbo orthography, '\n", "                \"adding the dotted vowels (ị, ọ, ụ) or ṅ wherever they belong; if none belong, leave it unchanged. \"\n", "                \"Tone marks are not needed. Reply with only the name.\")\n", "    if r[\"lang\"] == \"hausa\":\n", "        return (f'The Hausa name \"{bare}\" has been written in plain ASCII. Rewrite it in standard Hausa (Boko) orthography, '\n", "                \"using the hooked letters (ɓ, ɗ, ƙ, ƴ) wherever they belong; if none belong, leave it unchanged. \"\n", "                \"Reply with only the name.\")\n", "    return (f'The {r[\"sublang\"]} name \"{bare}\" has been written without its accents or special letters. '\n", "            f'Rewrite it in correct {r[\"sublang\"]} spelling with all diacritics. Reply with only the name.')\n", "\n", "\n", "def tone_prompt(r):\n", "    return (\"In standard Yoruba orthography, an acute accent (´) marks a high tone (H), a grave accent (`) marks a low tone (L), \"\n", "            \"and a vowel with no tone mark has a mid tone (M). Under-dots (ẹ, ọ, ṣ) are not tone marks. \"\n", "            f'Give the tone of each vowel in the name \"{r[\"name\"]}\", in order, as a sequence like H-M-L. '\n", "            \"Reply with only the sequence.\")\n", "\n", "\n", "def meaning_prompt(r, form):\n", "    return (f'What does the {r[\"sublang\"]} name \"{r[form]}\" mean in English? Answer in one short sentence.')\n", "\n", "\n", "def build_prompts():\n", "    names, pairs = name_rows(), pair_rows()\n", "    P = []\n", "    for r in names:\n", "        for tid, t in COPY_TEMPLATES.items():\n", "            P.append(dict(task=\"copy\", template=tid, item=r[\"id\"], prompt=t.format(name=r[\"name\"])))\n", "        if r[\"id\"] not in NO_RESTORE:\n", "            P.append(dict(task=\"restore\", template=\"restore\", item=r[\"id\"], prompt=restore_prompt(r)))\n", "        if r[\"lang\"] == \"yoruba\":\n", "            P.append(dict(task=\"tone\", template=\"tone\", item=r[\"id\"], prompt=tone_prompt(r)))\n", "        if r[\"lang\"] != \"control\":\n", "            for form in (\"name\", \"stripped\"):\n", "                if form == \"stripped\" and r[\"stripped\"] == r[\"name\"]:\n", "                    continue\n", "                P.append(dict(task=\"meaning\", template=f\"meaning_{form}\", item=r[\"id\"], prompt=meaning_prompt(r, form)))\n", "    for i, p in enumerate(pairs):\n", "        L = LANG_LABEL[p[\"lang\"]]\n", "        P.append(dict(task=\"pair\", template=\"pair_marked\", item=f'{p[\"set\"]}:{i}',\n", "                      prompt=f'What does the {L} word \"{p[\"word\"]}\" mean in English? Answer with just the English word or a short phrase.'))\n", "    for s in sorted({p[\"set\"] for p in pairs}):\n", "        L = LANG_LABEL[next(p[\"lang\"] for p in pairs if p[\"set\"] == s)]\n", "        P.append(dict(task=\"ambiguity\", template=\"ambiguity\", item=s,\n", "                      prompt=(f'In {L}, the word \"{s}\" is written here without tone marks or diacritics, so it could stand for '\n", "                              \"several different words. List every distinct word it could be, giving each one's correct spelling \"\n", "                              \"with diacritics and its English meaning.\")))\n", "    for i, p in enumerate(P):\n", "        p[\"pid\"] = f\"p{i:04d}\"\n", "    return names, pairs, P\n", "\n", "\n", "def write_dataset(outdir=\"data\"):\n", "    os.makedirs(outdir, exist_ok=True)\n", "    names, pairs, prompts = build_prompts()\n", "    for fn, rows in ((\"names.csv\", names), (\"minimal_pairs.csv\", pairs)):\n", "        with open(os.path.join(outdir, fn), \"w\", newline=\"\", encoding=\"utf-8\") as f:\n", "            w = csv.DictWriter(f, fieldnames=list(rows[0].keys()))\n", "            w.writeheader(); w.writerows(rows)\n", "    with open(os.path.join(outdir, \"prompts.jsonl\"), \"w\", encoding=\"utf-8\") as f:\n", "        for p in prompts:\n", "            f.write(json.dumps(p, ensure_ascii=False) + \"\\n\")\n", "    return names, pairs, prompts\n", "\n", "\n", "# ---------------------------------------------------------------------------\n", "# 4. Scoring\n", "# ---------------------------------------------------------------------------\n", "def _folded_with_map(text):\n", "    us = units(text)\n", "    folded, idx = [], []\n", "    for i, (b, _) in enumerate(us):\n", "        f = SPECIAL_FOLD.get(b, b).lower()\n", "        if len(f) != 1:            # rare case-mapping expansions; keep alignment\n", "            f = f[:1] or \" \"\n", "        folded.append(f); idx.append(i)\n", "    return \"\".join(folded), idx, us\n", "\n", "\n", "def _search(hay, needle):\n", "    m = re.search(r\"(?<!\\w)\" + re.escape(needle) + r\"(?!\\w)\", hay)\n", "    return m.start() if m else -1\n", "\n", "\n", "def compare_aligned(gold, pred, ignore_tone=False):\n", "    \"\"\"Position-by-position comparison of identity features (same base letters).\"\"\"\n", "    gu, pu = units(gold), units(pred)\n", "    tp = fp = fn = 0\n", "    lost = defaultdict(int)\n", "    for g, p in zip(gu, pu):\n", "        G, Pf = features(g, ignore_tone), features(p, ignore_tone)\n", "        tp += len(G & Pf); fp += len(Pf - G); fn += len(G - Pf)\n", "        for m in G - Pf:\n", "            lost[MARK_NAMES.get(m, m)] += 1\n", "    return tp, fp, fn, dict(lost)\n", "\n", "\n", "def classify(gold, pred, ignore_tone=False):\n", "    tp, fp, fn, lost = compare_aligned(gold, pred, ignore_tone)\n", "    total = tp + fn\n", "    if fn == 0 and fp == 0:\n", "        status = \"exact\"\n", "    elif tp == 0 and fp == 0 and total > 0:\n", "        status = \"fully_stripped\"\n", "    elif fp > 0 and fn == 0:\n", "        status = \"spurious_marks\"\n", "    elif fp > 0:\n", "        status = \"wrong_marks\"\n", "    else:\n", "        status = \"partially_stripped\"\n", "    return dict(status=status, tp=tp, fp=fp, fn=fn, lost=lost)\n", "\n", "\n", "def find_rendition(output, gold):\n", "    \"\"\"Locate the model's rendering of `gold` inside free text.\"\"\"\n", "    out = nfc(output)\n", "    hay, idx, us = _folded_with_map(out)\n", "    needle = strip_marks(gold).lower()\n", "    pos = _search(hay, needle)\n", "    if pos >= 0:\n", "        a, b = idx[pos], idx[pos + len(needle) - 1] + 1\n", "        return nfc(\"\".join(c + \"\".join(m) for c, m in us[a:b])), \"found\"\n", "    # catastrophic loss: a special letter vanished entirely (ASCII 'ignore')\n", "    if any(c in SPECIAL_FOLD for c in nfc(gold)):\n", "        dropped = \"\".join(c for c in nfc(gold) if c not in SPECIAL_FOLD)\n", "        if _search(hay, strip_marks(dropped).lower()) >= 0:\n", "            return None, \"catastrophic\"\n", "    # otherwise: name was altered (misspelt/transliterated) or missing\n", "    words = hay.split()\n", "    k = max(1, len(needle.split()))\n", "    best = 0.0\n", "    for i in range(len(words)):\n", "        cand = \" \".join(words[i:i + k]).strip(\".,;:!?\\\"'()[]{}\")\n", "        best = max(best, difflib.SequenceMatcher(None, needle, cand).ratio())\n", "    return None, (\"altered\" if best >= 0.6 else \"missing\")\n", "\n", "\n", "def appears_exact(gold, output, ignore_tone=False):\n", "    \"\"\"True if ANY occurrence of gold's letters in output carries exactly gold's marks.\"\"\"\n", "    hay, idx, us = _folded_with_map(nfc(output))\n", "    needle = strip_marks(gold).lower()\n", "    for m in re.finditer(r\"(?<!\\w)\" + re.escape(needle) + r\"(?!\\w)\", hay):\n", "        a, b = idx[m.start()], idx[m.start() + len(needle) - 1] + 1\n", "        rend = nfc(\"\".join(c + \"\".join(mk) for c, mk in us[a:b]))\n", "        if classify(gold, rend, ignore_tone)[\"status\"] == \"exact\":\n", "            return True\n", "    return False\n", "\n", "\n", "def score_copy(gold, output, ignore_tone=False):\n", "    rend, how = find_rendition(output, gold)\n", "    if rend is None:\n", "        n = n_identity_marks(gold, ignore_tone)\n", "        return dict(status=how, rendition=\"\", tp=0, fp=0, fn=n, lost={})\n", "    d = classify(gold, rend, ignore_tone)\n", "    d[\"rendition\"] = rend\n", "    return d\n", "\n", "\n", "def clean_short_answer(text):\n", "    text = re.sub(r\"<think>.*?</think>\", \"\", text, flags=re.S).strip()\n", "    line = next((l for l in text.splitlines() if l.strip()), \"\")\n", "    line = re.sub(r\"^\\s*(answer|name)\\s*:\\s*\", \"\", line, flags=re.I)\n", "    return nfc(line.strip().strip(\"*`\\\"'“”‘’\").rstrip(\".\").strip())\n", "\n", "\n", "def score_restore(gold, output, ignore_tone=False):\n", "    ans = clean_short_answer(output)\n", "    if strip_marks(ans).lower() == strip_marks(gold).lower():\n", "        d = classify(gold, ans, ignore_tone)\n", "    else:\n", "        rend, how = find_rendition(output, gold)\n", "        if rend is None:\n", "            n = n_identity_marks(gold, ignore_tone)\n", "            d = dict(status=\"base_changed\", tp=0, fp=0, fn=n, lost={})\n", "        else:\n", "            d = classify(gold, rend, ignore_tone); ans = rend\n", "    d[\"answer\"] = ans\n", "    return d\n", "\n", "\n", "def score_tone(gold_pattern, output):\n", "    text = re.sub(r\"<think>.*?</think>\", \"\", output, flags=re.S).upper()\n", "    pred = re.findall(r\"(?<![A-Z])[HML](?![A-Z])\", text)\n", "    gold = gold_pattern.split(\"-\")\n", "    same_len = len(pred) == len(gold)\n", "    pos_acc = sum(a == b for a, b in zip(gold, pred)) / len(gold)\n", "    return dict(pred=\"-\".join(pred), exact=pred == gold, same_length=same_len, position_acc=round(pos_acc, 3))\n", "\n", "\n", "def score_meaning(kw_groups, output):\n", "    t = output.lower()\n", "    if isinstance(kw_groups, str):\n", "        kw_groups = json.loads(kw_groups)\n", "    return all(any(k in t for k in grp) for grp in kw_groups)\n", "\n", "\n", "def score_ambiguity(set_pairs, output):\n", "    t = output.lower()\n", "    senses = sum(any(k in t for k in json.loads(p[\"keywords\"])) for p in set_pairs)\n", "    # Tone is what distinguishes these words, so never ignore it here, and only\n", "    # count words that carry marks (a bare form like \"kara\" is trivially \"spelled\").\n", "    marked = [p for p in set_pairs if n_identity_marks(p[\"word\"]) > 0]\n", "    spelled = sum(appears_exact(p[\"word\"], output) for p in marked)\n", "    return dict(senses_recalled=senses, correctly_spelled=spelled, n_senses=len(set_pairs), n_marked=len(marked))\n", "\n", "\n", "# ---------------------------------------------------------------------------\n", "# 5. Score everything\n", "# ---------------------------------------------------------------------------\n", "def score_all(names, pairs, prompts, gens):\n", "    \"\"\"gens: dict pid -> model output text. Returns dict of row lists.\"\"\"\n", "    N = {r[\"id\"]: r for r in names}\n", "    res = defaultdict(list)\n", "    for p in prompts:\n", "        out = gens.get(p[\"pid\"])\n", "        if out is None:\n", "            continue\n", "        base = dict(pid=p[\"pid\"], task=p[\"task\"], template=p[\"template\"], item=p[\"item\"], output=out)\n", "        if p[\"task\"] in (\"copy\", \"restore\", \"tone\", \"meaning\"):\n", "            r = N[p[\"item\"]]\n", "            base.update(lang=r[\"lang\"], sublang=r[\"sublang\"], name=r[\"name\"], density=r[\"density\"],\n", "                        n_marks=r[\"n_marks\"], has_hooked=r[\"has_hooked\"], n_letters=r[\"n_letters\"])\n", "            it = IGNORE_TONE[r[\"lang\"]]\n", "            if p[\"task\"] == \"copy\":\n", "                d = score_copy(r[\"name\"], out, it)\n", "            elif p[\"task\"] == \"restore\":\n", "                d = score_restore(r[\"name\"], out, it)\n", "            elif p[\"task\"] == \"tone\":\n", "                d = score_tone(r[\"tone_pattern\"], out); d[\"gold\"] = r[\"tone_pattern\"]\n", "            else:\n", "                d = dict(correct=score_meaning(r[\"meaning_keywords\"], out), form=p[\"template\"].split(\"_\")[1])\n", "            if \"lost\" in d:\n", "                d[\"lost\"] = json.dumps(d[\"lost\"])\n", "            base.update(d)\n", "        elif p[\"task\"] == \"pair\":\n", "            s, i = p[\"item\"].split(\":\")\n", "            pr = pairs[int(i)]\n", "            base.update(lang=pr[\"lang\"], set=s, word=pr[\"word\"], gloss=pr[\"gloss\"],\n", "                        correct=any(k in out.lower() for k in json.loads(pr[\"keywords\"])))\n", "        elif p[\"task\"] == \"ambiguity\":\n", "            sp = [x for x in pairs if x[\"set\"] == p[\"item\"]]\n", "            base.update(lang=sp[0][\"lang\"], **score_ambiguity(sp, out))\n", "        res[p[\"task\"]].append(base)\n", "    return dict(res)\n", "\n", "\n", "def summarize(res):\n", "    \"\"\"Headline numbers per language group.\"\"\"\n", "    S = defaultdict(dict)\n", "    def frac(rows, f):\n", "        return round(sum(1 for r in rows if f(r)) / len(rows), 3) if rows else None\n", "    groups = [\"yoruba\", \"igbo\", \"hausa\", \"control\"]\n", "    for g in groups:\n", "        c = [r for r in res.get(\"copy\", []) if r[\"lang\"] == g and r[\"n_marks\"] > 0]\n", "        S[g][\"copy_exact\"] = frac(c, lambda r: r[\"status\"] == \"exact\")\n", "        S[g][\"copy_fully_stripped\"] = frac(c, lambda r: r[\"status\"] == \"fully_stripped\")\n", "        S[g][\"copy_catastrophic\"] = frac(c, lambda r: r[\"status\"] == \"catastrophic\")\n", "        tp = sum(r[\"tp\"] for r in c); fn = sum(r[\"fn\"] for r in c)\n", "        S[g][\"copy_mark_recall\"] = round(tp / (tp + fn), 3) if tp + fn else None\n", "        rs = [r for r in res.get(\"restore\", []) if r[\"lang\"] == g]\n", "        rs_m = [r for r in rs if r[\"n_marks\"] > 0]\n", "        S[g][\"restore_exact\"] = frac(rs_m, lambda r: r[\"status\"] == \"exact\")\n", "        tp = sum(r[\"tp\"] for r in rs_m); fp = sum(r[\"fp\"] for r in rs_m); fn = sum(r[\"fn\"] for r in rs_m)\n", "        S[g][\"restore_mark_f1\"] = round(2 * tp / (2 * tp + fp + fn), 3) if (tp + fp + fn) else None\n", "        ctrl = [r for r in rs if r[\"n_marks\"] == 0]\n", "        S[g][\"restore_overcorrection\"] = frac(ctrl, lambda r: r[\"fp\"] > 0 or r[\"status\"] == \"base_changed\")\n", "        if g != \"control\":\n", "            m = res.get(\"meaning\", [])\n", "            S[g][\"meaning_acc_marked\"] = frac([r for r in m if r[\"lang\"] == g and r[\"form\"] == \"name\"], lambda r: r[\"correct\"])\n", "            S[g][\"meaning_acc_stripped\"] = frac([r for r in m if r[\"lang\"] == g and r[\"form\"] == \"stripped\"], lambda r: r[\"correct\"])\n", "            pr = [r for r in res.get(\"pair\", []) if r[\"lang\"] == g]\n", "            S[g][\"minimal_pair_acc\"] = frac(pr, lambda r: r[\"correct\"])\n", "            amb = [r for r in res.get(\"ambiguity\", []) if r[\"lang\"] == g]\n", "            if amb:\n", "                S[g][\"ambiguity_sense_recall\"] = round(sum(r[\"senses_recalled\"] for r in amb) / sum(r[\"n_senses\"] for r in amb), 3)\n", "                S[g][\"ambiguity_spelled_correctly\"] = round(sum(r[\"correctly_spelled\"] for r in amb) / max(1, sum(r[\"n_marked\"] for r in amb)), 3)\n", "    t = res.get(\"tone\", [])\n", "    if t:\n", "        S[\"yoruba\"][\"tone_read_exact\"] = frac(t, lambda r: r[\"exact\"])\n", "        S[\"yoruba\"][\"tone_read_position_acc\"] = round(sum(r[\"position_acc\"] for r in t) / len(t), 3)\n", "    return {k: dict(v) for k, v in S.items()}\n", "\n", "\n", "def copy_by_template(res):\n", "    out = defaultdict(dict)\n", "    for r in res.get(\"copy\", []):\n", "        if r[\"n_marks\"] == 0:\n", "            continue\n", "        out[r[\"lang\"]].setdefault(r[\"template\"], []).append(r[\"status\"] == \"exact\")\n", "    return {g: {t: round(sum(v) / len(v), 3) for t, v in d.items()} for g, d in out.items()}\n", "\n", "\n", "def lost_mark_counts(res, task=\"copy\"):\n", "    tot = defaultdict(lambda: defaultdict(int))\n", "    for r in res.get(task, []):\n", "        for k, v in json.loads(r.get(\"lost\") or \"{}\").items():\n", "            tot[r[\"lang\"]][k] += v\n", "    return {g: dict(d) for g, d in tot.items()}"], "metadata": {"colab": {"base_uri": "https://localhost:8080/"}, "id": "vGhGoF12tNgV", "executionInfo": {"status": "ok", "timestamp": 1792105831181, "user_tz": -60, "elapsed": 17, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "b38b90fd-2e80-44cc-a669-4114d6219dc8"}, "execution_count": 3, "outputs": [{"output_type": "stream", "name": "stdout", "text": ["Overwriting oruko_core.py\n"]}]}, {"cell_type": "code", "source": ["%%writefile README.md\n", "# ORÚKỌ-mini: Does a small frontier LLM keep an African name intact?\n", "\n", "*A small-scale, text-only probe of the identity-preservation gap described in the ORÚKỌ benchmark proposal, run on **Qwen/Qwen3.5-4B**.*\n", "\n", "Author: Flora Oladipupo · October 2026\n", "\n", "---\n", "\n", "## TL;DR\n", "\n", "Qwen3.5-4B claims 201 languages. When a correctly written Yorùbá, Igbo or Hausa name sits in the prompt, it mostly copies it, but asked to write that name in capital letters it drops the marks for most of them (*Ɗanjuma* → *DANJUMA*, *Chiọma* → *CHIOMA*). When it has to supply the marks itself, it fails almost completely: it cannot restore a stripped Yorùbá or Hausa name, cannot read Yorùbá tone marks even when the rule is in the prompt, gives the same answer to words that differ only by tone, and rates the stripped spelling as more likely than the correct one for every African name tested. The same model restores *François* and *Nguyễn Thị Hương* perfectly. Standard benchmarks miss this, because they either strip diacritics before scoring or never test names at all.\n", "\n", "The practical consequence: in any pipeline where the name arrives without its marks (a form, an ASR transcript, a database that already folded it) or must be transformed (uppercased, normalised for TTS), the model will not recover the person's real name, and nothing in a typical eval will flag it.\n", "\n", "<!-- RESULTS:START -->\n", "*Results are inserted here automatically by `oruko_eval.ipynb` after the run.*\n", "<!-- RESULTS:END -->\n", "\n", "---\n", "\n", "## 1. The blind spot (Question 1)\n", "\n", "**Identity erasure of African personal names.** In Yorùbá and Igbo, tone and diacritics are part of the word; in Hausa, ɓ, ɗ and ƙ are separate letters, not decorations. *Ọkọ* is a husband, *ọkọ̀* a vehicle, *ọkọ́* a hoe and *oko* a farm; Igbo *àkwà*, *ákwà*, *ákwá* and *àkwá* are bed, cloth, cry and egg. Strip the marks and the word, and any name built on it, becomes a different word or no word at all. *Ɗanjuma* folded to ASCII becomes *Danjuma*, and with a naive `encode(\"ascii\", \"ignore\")` it becomes *anjuma*, a name that starts mid-word.\n", "\n", "**Why I care.** I am a Nigerian AI engineer. I have watched names like mine, and those of nearly everyone I grew up with, come back from forms, airline systems, banking apps, voice assistants and now chatbots as a flattened approximation. I have experienced it over and over again as an African with an African name seen people mispronounce my name without remorse even when corrected so since I have the will power to correct this in AI system since I have limited control over correcting humans on such mistake hence why I am utilizing my skillset as an AI Engineer to correct this.  Each instance looks like a small courtesy failure. Together they mean the systems that increasingly mediate identity (KYC, health records, government services, voice agents) are quietly recording people under names that are not theirs.\n", "\n", "**Why benchmarks miss it.**\n", "\n", "* **Normalisation hides it.** WER/CER pipelines, BLEU tokenisers and many pre-training filters lowercase and strip diacritics before scoring or deduplicating. The error is removed before anyone measures it.\n", "* **Names are treated as copy tokens.** Benchmarks assume a model either knows a language or not, and treat names as opaque strings to pass through.\n", "* **Language coverage is mistaken for identity fidelity.** Model cards list Yorùbá, Igbo and Hausa as \"supported\" based on QA or translation scores. A model can translate a Yorùbá sentence acceptably and still drop every tone mark in the speaker's name.\n", "* **It looks like success.** The output is fluent, grammatical and in the right language. Only a speaker of the language notices the name is wrong, and those speakers are rarely the ones writing evals.\n", "\n", "This probe takes the full ORÚKỌ framing (orthographic, phonemic, tonal, prosodic and semantic fidelity in speech systems) and asks the text-only, LLM-sized version of the question: **before any audio is produced, does the language model at the front of the pipeline even keep the name?**\n", "\n", "---\n", "\n", "## 2. How it was evaluated (Question 2)\n", "\n", "### Model choice: `Qwen/Qwen3.5-4B`\n", "\n", "* **Current frontier family.** Released February 2026 as the small member of the Qwen3.5 family; its post-training recipe, tokenizer (≈248k vocabulary) and multilingual data are shared with the much larger flagship models, so its failure modes are a reasonable proxy for theirs.\n", "* **Explicitly multilingual.** The model card claims support for 201 languages and dialects. That makes it the right test: if the gap appears here, it is not a \"language not supported\" story.\n", "* **Practical.** ≈4B language-model parameters (≈5B with the vision encoder, inside the 0.6–6B limit), Apache-2.0, ungated, and loadable on a free Colab GPU.\n", "* Run in **non-thinking (instruct) mode** with **greedy decoding**, so every number is reproducible.\n", "\n", "### Dataset (`data/`)\n", "\n", "| File | Contents |\n", "|---|---|\n", "| `names.csv` | 43 names: 20 Yorùbá, 10 Igbo, 4 Hausa, 9 high-resource controls (Spanish, French, Danish, Swedish, Polish, German, Czech, Vietnamese, Mandarin pinyin). Each row has the correct orthography (Unicode NFC), the stripped form, the ASCII-ignore form, number of identity-bearing marks, diacritic density, hooked-letter flag, Yorùbá tone pattern, meaning gloss and keywords, and a `confidence` flag. |\n", "| `minimal_pairs.csv` | 16 words in 5 sets that collapse to the same bare letters (`oko`, `igba`, `akwa`, `kara`, `dari`). |\n", "| `prompts.jsonl` | All 363 prompts, generated deterministically from the two files above. |\n", "\n", "Controls include names with no marks at all (*Chukwuebuka*, *Nkechi*, *Ladi*) to catch over-correction, and tonal high-resource scripts (Vietnamese, pinyin) so that \"tones are hard\" cannot explain the result on its own. Igbo and Hausa are scored on dotted vowels and hooked letters only, since their standard orthographies do not write tone in names.\n", "\n", "**Correctness is bearer-defined.** Rows marked `check` should be confirmed by a native speaker (ideally a bearer of the name) before the dataset is cited.\n", "\n", "### Tasks (the Identity Preservation Framework, adapted to text)\n", "\n", "| Task | IPF component | What it asks | Metric |\n", "|---|---|---|---|\n", "| **Copy under transformation** | Orthographic | Use the name in 5 realistic contexts: repeat it, welcome message, JSON patient record, French translation, UPPERCASE boarding pass. | Exact identity match; mark recall; outcome class: exact / partially stripped / fully stripped / wrong marks / **catastrophic** (letter dropped entirely) / altered / missing |\n", "| **Restoration** | Orthographic | Given the stripped name, restore full orthography. | Exact match; mark-level F1; over-correction on names that have no marks |\n", "| **Tone reading** (Yorùbá) | Tonal | The prompt states the rule (acute = H, grave = L, unmarked = M) and asks for the tone sequence. Pure symbol reading, no world knowledge needed. | Exact sequence; per-position accuracy |\n", "| **Meaning** | Semantic | Meaning of the name, asked once with marks and once stripped. | Keyword-group accuracy |\n", "| **Minimal pairs + ambiguity** | Semantic shift | Meaning of each marked word; then, for each bare form, list every word it could be. | Accuracy; senses recalled; senses spelled correctly |\n", "| **Likelihood preference** | Model prior | Log-probability of \"The name is X.\" with the correct vs the stripped form. | % of names where the model prefers the stripped form |\n", "| **Tokenizer fragmentation** | Representation | Tokens for the correct vs stripped form. | Token ratio per language |\n", "\n", "The analysis also relates outcomes to linguistic features (diacritic density, length, hooked letters), as proposed in ORÚKỌ.\n", "\n", "### Reproducing\n", "\n", "Open `oruko_eval.ipynb` in Google Colab (GPU runtime; L4 or A100 preferred, T4 works), add an `HF_TOKEN` secret with write access, and run all cells. The run takes roughly 15–30 minutes, writes everything under `results/` and `figures/`, fills in the results section of this README, and syncs the folder to the bucket.\n", "\n", "---\n", "\n", "## 3. A path forward (Question 3)\n", "\n", "These are proposals, not tested results.\n", "\n", "**1. Stop deleting the signal in data curation.**\n", "* Audit pre-training and SFT pipelines for normalisation that strips combining marks or folds special letters (NFKD + ASCII fold, `unidecode`, aggressive dedup keys). Deduplicate on NFC text, never on folded text.\n", "* Up-weight *properly marked* Yorùbá, Igbo and Hausa text (edited news, literature, Bible translations, curated parallel corpora) relative to the large volume of unmarked social-media text, and tag diacritic completeness as a quality signal.\n", "* Build a **bearer-verified name lexicon**: name, orthography, tone pattern, meaning, regional variants, consented by native speakers. This is small, cheap, and the highest-value data for this gap.\n", "\n", "**2. Fine-tune for preservation, not just knowledge.**\n", "* *Case-mapping data:* the clearest failure in this run is uppercasing. Text where marked names appear in capitals (ID cards, boarding passes, headlines, legal documents) is rare in training data, so the model never learned that *ọ* uppercases to *Ọ* and *ɗ* to *Ɗ*. A few thousand synthetic examples should close most of this.\n", "* *Copy-under-transformation SFT:* synthetic tasks that embed marked names in forms, JSON, translation, case changes and summaries, with targets that keep every mark. This attacks the most common real-world failure directly.\n", "* *Contrastive minimal pairs:* train the model to answer *ọkọ / ọkọ̀ / ọkọ́ / oko* differently, and to say \"this is ambiguous without tone marks\" when given a bare form instead of guessing.\n", "\n", "\n", "**3. Small architectural and inference-time changes.**\n", "* **Tokenizer:** add merges for tone-marked syllables and NFC-precomposed under-dot vowels so a marked name is not 2–3x more fragmented than its stripped form. Fragmentation makes the stripped form the cheaper path at every decoding step.\n", "* **Protected spans:** when a user supplies a name, treat it as an immutable span (a copy/pointer mechanism or constrained decoding) that must be reproduced byte-for-byte, the way some systems already protect URLs and code identifiers.\n", "* **Output guardrail:** a cheap post-hoc check that any name present in the input reappears with identical identity features, and flags or repairs it if not.\n", "\n", "**4. Make it a standard metric.** Report identity preservation on model cards next to multilingual QA scores, and never compute WER/CER/BLEU on diacritic-stripped text for these languages. The broader ORÚKỌ benchmark extends the same measurements to TTS and ASR, where the stripped name finally becomes a mispronounced one.\n", "\n", "---\n", "\n", "## Bucket contents\n", "\n", "```\n", "README.md                 this file (answers to Q1 and Q3, method and results for Q2)\n", "oruko_eval.ipynb          Colab notebook that runs the whole evaluation\n", "oruko_core.py             dataset, prompts and scoring (pure Python, unit-testable)\n", "data/                     names.csv, minimal_pairs.csv, prompts.jsonl\n", "results/                  generations.jsonl, per-task score CSVs, summary.json, tokenizer and log-prob tables\n", "figures/                  charts used in this README\n", "model/                    server-side copy of Qwen/Qwen3.5-4B weights (exact model evaluated)\n", "```\n", "\n", "## Limitations\n", "\n", "* Small dataset (43 names, 16 words); results are indicative, not a leaderboard. Phase 1 covers only three Nigerian languages.\n", "* Meaning is scored with keyword groups, which can miss valid paraphrases; the raw generations are released for manual review.\n", "* The likelihood probe compares summed log-probabilities, so shorter (stripped) spellings get some advantage from length alone. The controls are scored the same way and prefer the stripped form far less often, which suggests length is not the whole story, but a per-token comparison would make this cleaner.\n", "* Copy-task loss is concentrated in a single context (uppercase). More transformation contexts (title case, slug/URL, transliteration into other scripts, summarisation) would show how general that is.\n", "* Greedy decoding with one prompt per template; sampling and prompt variation would give confidence intervals.\n", "* Text only. The full ORÚKỌ framework measures pronunciation, tone contours and prosody in audio.\n", "\n", "## References\n", "\n", "1. Olatunji et al. *AfriNames: Most ASR models \"butcher\" African Names.* arXiv:2306.00253, 2023.\n", "2. Intron Innovation. *Afri-names: Read Speech Dataset of Numbers and African Named Entities.* Hugging Face, 2025. CC BY-NC-SA 4.0.\n", "3. Qwen Team. *Qwen3.5: Towards Native Multimodal Agents.* 2026.\n", "\n", "Dataset and code released under CC BY 4.0 (data) and Apache-2.0 (code). Model weights remain under their original Apache-2.0 license."], "metadata": {"colab": {"base_uri": "https://localhost:8080/"}, "id": "-MpNRrnOtrLM", "executionInfo": {"status": "ok", "timestamp": 1792105831191, "user_tz": -60, "elapsed": 9, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "c2cce4d9-da22-4d72-f57d-6dd0426600c9"}, "execution_count": 4, "outputs": [{"output_type": "stream", "name": "stdout", "text": ["Overwriting README.md\n"]}]}, {"cell_type": "markdown", "metadata": {"id": "t8LMDjqsYy4x"}, "source": ["## 1. Dataset"]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/", "height": 1000}, "id": "50Y22p1kYy4y", "executionInfo": {"status": "ok", "timestamp": 1792105831467, "user_tz": -60, "elapsed": 275, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "42f39065-e390-4bb3-b3bd-df3c218cd69a"}, "execution_count": 5, "outputs": [{"output_type": "stream", "name": "stdout", "text": ["43 names | 16 minimal-pair words | 363 prompts\n", "{'copy': 215, 'meaning': 65, 'restore': 42, 'tone': 20, 'pair': 16, 'ambiguity': 5}\n"]}, {"output_type": "execute_result", "data": {"text/plain": ["      id     lang              name          stripped  ascii_ignore  n_marks  \\\n", "0   yo01   yoruba     Olúwadámilárè     Oluwadamilare     Olwadmilr        4   \n", "1   yo02   yoruba           Adéṣọlá           Adesola           Adl        4   \n", "2   yo03   yoruba          Ọláyínká          Olayinka          lynk        4   \n", "3   yo04   yoruba          Fọláṣadé          Folasade          Flad        4   \n", "4   yo05   yoruba         Olúwaṣeun         Oluwaseun       Olwaeun        2   \n", "5   yo06   yoruba          Ayọ̀délé           Ayodele          Aydl        4   \n", "6   yo07   yoruba          Ìyábọ̀dé           Iyabode           ybd        5   \n", "7   yo08   yoruba           Bọ́lájí            Bolaji           Blj        4   \n", "8   yo09   yoruba        Fúnmiláyọ̀         Funmilayo        Fnmily        4   \n", "9   yo10   yoruba          Adébáyọ̀           Adebayo          Adby        4   \n", "10  yo11   yoruba           Ọlámidé           Olamide          lmid        3   \n", "11  yo12   yoruba         Ìfẹ́olúwa          Ifeoluwa         folwa        4   \n", "12  yo13   yoruba             Táíwò             Taiwo            Tw        3   \n", "13  yo14   yoruba          Kẹ́hìndé           Kehinde          Khnd        4   \n", "14  yo15   yoruba             Ìdòwú             Idowu            dw        3   \n", "15  yo16   yoruba         Babátúndé         Babatunde        Babtnd        3   \n", "16  yo17   yoruba         Olúṣẹ́gun          Olusegun         Olgun        4   \n", "17  yo18   yoruba            Àbẹ̀ní             Abeni            bn        4   \n", "18  yo19   yoruba           Yéwándé           Yewande          Ywnd        3   \n", "19  yo20   yoruba           Adéwálé           Adewale          Adwl        3   \n", "20  ig01     igbo            Chiọma            Chioma         Chima        1   \n", "21  ig02     igbo           Ụchenna           Uchenna        chenna        1   \n", "22  ig03     igbo            Obiọma            Obioma         Obima        1   \n", "23  ig04     igbo            Ifeọma            Ifeoma         Ifema        1   \n", "24  ig05     igbo          Chinọnso          Chinonso       Chinnso        1   \n", "25  ig06     igbo          Ụzọamaka          Uzoamaka        zamaka        2   \n", "26  ig07     igbo     Ifeanyịchukwu     Ifeanyichukwu  Ifeanychukwu        1   \n", "27  ig08     igbo             Ngọzi             Ngozi          Ngzi        1   \n", "28  ig09     igbo       Chukwuebuka       Chukwuebuka   Chukwuebuka        0   \n", "29  ig10     igbo            Nkechi            Nkechi        Nkechi        0   \n", "30  ha01    hausa           Ɗanjuma           Danjuma        anjuma        1   \n", "31  ha02    hausa           Ɗanladi           Danladi        anladi        1   \n", "32  ha03    hausa          Ɗanazumi          Danazumi       anazumi        1   \n", "33  ha04    hausa              Ladi              Ladi          Ladi        0   \n", "34  ct01  control              José              Jose           Jos        1   \n", "35  ct02  control          François          Francois       Franois        1   \n", "36  ct03  control             Søren             Soren          Sren        1   \n", "37  ct04  control             Björn             Bjorn          Bjrn        1   \n", "38  ct05  control            Łukasz            Lukasz         ukasz        1   \n", "39  ct06  control            Jürgen            Jurgen         Jrgen        1   \n", "40  ct07  control            Dvořák            Dvorak          Dvok        2   \n", "41  ct08  control  Nguyễn Thị Hương  Nguyen Thi Huong  Nguyn Th Hng        5   \n", "42  ct09  control         Zhāng Wěi         Zhang Wei       Zhng Wi        2   \n", "\n", "     tone_pattern                                            meaning  \\\n", "0   M-H-M-H-M-H-L                              God has vindicated me   \n", "1         M-H-M-H                      the crown makes wealth/honour   \n", "2         M-H-H-H                         wealth/honour surrounds me   \n", "3         M-H-M-H                      uses wealth/honour as a crown   \n", "4       M-H-M-M-M                        thank God / God has done it   \n", "5         M-L-H-H                                  joy has come home   \n", "6         L-H-L-H                                mother has returned   \n", "7           H-H-H                       one who wakes up with wealth   \n", "8         H-M-H-L                                        give me joy   \n", "9         M-H-H-L                                the crown meets joy   \n", "10        M-H-M-H                          my wealth/honour has come   \n", "11      L-H-M-H-M                                    the love of God   \n", "12          H-H-L                                first-born of twins   \n", "13          H-L-H                               second-born of twins   \n", "14          L-L-H                             child born after twins   \n", "15        M-H-H-H                                father has returned   \n", "16        M-H-H-M                                  God has conquered   \n", "17          L-L-H                       we asked for her and got her   \n", "18          H-H-H                                mother has returned   \n", "19        M-H-H-H                            the crown has come home   \n", "20                                   good God / good personal spirit   \n", "21                                       the father's will / thought   \n", "22                                                        kind heart   \n", "23                                                        good thing   \n", "24                                                       God is near   \n", "25                                             the road/path is good   \n", "26                                     nothing is impossible for God   \n", "27                                                          blessing   \n", "28                 God is great (no dotted letters: over-correcti...   \n", "29                 God's own (no dotted letters: over-correction ...   \n", "30                                                boy born on Friday   \n", "31                                                boy born on Sunday   \n", "32                                boy born during the fast (Ramadan)   \n", "33                  girl born on Sunday (no hooked letters: control)   \n", "34                                                                     \n", "35                                                                     \n", "36                                                                     \n", "37                                                                     \n", "38                                                                     \n", "39                                                                     \n", "40                                                                     \n", "41                                                                     \n", "42                                                                     \n", "\n", "   confidence  \n", "0       check  \n", "1        high  \n", "2        high  \n", "3        high  \n", "4        high  \n", "5        high  \n", "6        high  \n", "7        high  \n", "8        high  \n", "9        high  \n", "10       high  \n", "11       high  \n", "12       high  \n", "13       high  \n", "14       high  \n", "15       high  \n", "16       high  \n", "17      check  \n", "18       high  \n", "19       high  \n", "20       high  \n", "21       high  \n", "22       high  \n", "23       high  \n", "24      check  \n", "25      check  \n", "26      check  \n", "27      check  \n", "28       high  \n", "29       high  \n", "30       high  \n", "31       high  \n", "32      check  \n", "33       high  \n", "34       high  \n", "35       high  \n", "36       high  \n", "37       high  \n", "38       high  \n", "39       high  \n", "40       high  \n", "41       high  \n", "42       high  "], "text/html": ["<div>\n", "<style scoped>\n", "    .dataframe tbody tr th:only-of-type {\n", "        vertical-align: middle;\n", "    }\n", "\n", "    .dataframe tbody tr th {\n", "        vertical-align: top;\n", "    }\n", "\n", "    .dataframe thead th {\n", "        text-align: right;\n", "    }\n", "</style>\n", "<table border=\"1\" class=\"dataframe\">\n", "  <thead>\n", "    <tr style=\"text-align: right;\">\n", "      <th></th>\n", "      <th>id</th>\n", "      <th>lang</th>\n", "      <th>name</th>\n", "      <th>stripped</th>\n", "      <th>ascii_ignore</th>\n", "      <th>n_marks</th>\n", "      <th>tone_pattern</th>\n", "      <th>meaning</th>\n", "      <th>confidence</th>\n", "    </tr>\n", "  </thead>\n", "  <tbody>\n", "    <tr>\n", "      <th>0</th>\n", "      <td>yo01</td>\n", "      <td>yoruba</td>\n", "      <td>Olúwadámilárè</td>\n", "      <td>Oluwadamilare</td>\n", "      <td>Olwadmilr</td>\n", "      <td>4</td>\n", "      <td>M-H-M-H-M-H-L</td>\n", "      <td>God has vindicated me</td>\n", "      <td>check</td>\n", "    </tr>\n", "    <tr>\n", "      <th>1</th>\n", "      <td>yo02</td>\n", "      <td>yoruba</td>\n", "      <td>Adéṣọlá</td>\n", "      <td>Adesola</td>\n", "      <td>Adl</td>\n", "      <td>4</td>\n", "      <td>M-H-M-H</td>\n", "      <td>the crown makes wealth/honour</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>2</th>\n", "      <td>yo03</td>\n", "      <td>yoruba</td>\n", "      <td>Ọláyínká</td>\n", "      <td>Olayinka</td>\n", "      <td>lynk</td>\n", "      <td>4</td>\n", "      <td>M-H-H-H</td>\n", "      <td>wealth/honour surrounds me</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>3</th>\n", "      <td>yo04</td>\n", "      <td>yoruba</td>\n", "      <td>Fọláṣadé</td>\n", "      <td>Folasade</td>\n", "      <td>Flad</td>\n", "      <td>4</td>\n", "      <td>M-H-M-H</td>\n", "      <td>uses wealth/honour as a crown</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>4</th>\n", "      <td>yo05</td>\n", "      <td>yoruba</td>\n", "      <td>Olúwaṣeun</td>\n", "      <td>Oluwaseun</td>\n", "      <td>Olwaeun</td>\n", "      <td>2</td>\n", "      <td>M-H-M-M-M</td>\n", "      <td>thank God / God has done it</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>5</th>\n", "      <td>yo06</td>\n", "      <td>yoruba</td>\n", "      <td>Ayọ̀délé</td>\n", "      <td>Ayodele</td>\n", "      <td>Aydl</td>\n", "      <td>4</td>\n", "      <td>M-L-H-H</td>\n", "      <td>joy has come home</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>6</th>\n", "      <td>yo07</td>\n", "      <td>yoruba</td>\n", "      <td>Ìyábọ̀dé</td>\n", "      <td>Iyabode</td>\n", "      <td>ybd</td>\n", "      <td>5</td>\n", "      <td>L-H-L-H</td>\n", "      <td>mother has returned</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>7</th>\n", "      <td>yo08</td>\n", "      <td>yoruba</td>\n", "      <td>Bọ́lájí</td>\n", "      <td>Bolaji</td>\n", "      <td>Blj</td>\n", "      <td>4</td>\n", "      <td>H-H-H</td>\n", "      <td>one who wakes up with wealth</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>8</th>\n", "      <td>yo09</td>\n", "      <td>yoruba</td>\n", "      <td>Fúnmiláyọ̀</td>\n", "      <td>Funmilayo</td>\n", "      <td>Fnmily</td>\n", "      <td>4</td>\n", "      <td>H-M-H-L</td>\n", "      <td>give me joy</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>9</th>\n", "      <td>yo10</td>\n", "      <td>yoruba</td>\n", "      <td>Adébáyọ̀</td>\n", "      <td>Adebayo</td>\n", "      <td>Adby</td>\n", "      <td>4</td>\n", "      <td>M-H-H-L</td>\n", "      <td>the crown meets joy</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>10</th>\n", "      <td>yo11</td>\n", "      <td>yoruba</td>\n", "      <td>Ọlámidé</td>\n", "      <td>Olamide</td>\n", "      <td>lmid</td>\n", "      <td>3</td>\n", "      <td>M-H-M-H</td>\n", "      <td>my wealth/honour has come</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>11</th>\n", "      <td>yo12</td>\n", "      <td>yoruba</td>\n", "      <td>Ìfẹ́olúwa</td>\n", "      <td>Ifeoluwa</td>\n", "      <td>folwa</td>\n", "      <td>4</td>\n", "      <td>L-H-M-H-M</td>\n", "      <td>the love of God</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>12</th>\n", "      <td>yo13</td>\n", "      <td>yoruba</td>\n", "      <td>Táíwò</td>\n", "      <td>Taiwo</td>\n", "      <td>Tw</td>\n", "      <td>3</td>\n", "      <td>H-H-L</td>\n", "      <td>first-born of twins</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>13</th>\n", "      <td>yo14</td>\n", "      <td>yoruba</td>\n", "      <td>Kẹ́hìndé</td>\n", "      <td>Kehinde</td>\n", "      <td>Khnd</td>\n", "      <td>4</td>\n", "      <td>H-L-H</td>\n", "      <td>second-born of twins</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>14</th>\n", "      <td>yo15</td>\n", "      <td>yoruba</td>\n", "      <td>Ìdòwú</td>\n", "      <td>Idowu</td>\n", "      <td>dw</td>\n", "      <td>3</td>\n", "      <td>L-L-H</td>\n", "      <td>child born after twins</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>15</th>\n", "      <td>yo16</td>\n", "      <td>yoruba</td>\n", "      <td>Babátúndé</td>\n", "      <td>Babatunde</td>\n", "      <td>Babtnd</td>\n", "      <td>3</td>\n", "      <td>M-H-H-H</td>\n", "      <td>father has returned</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>16</th>\n", "      <td>yo17</td>\n", "      <td>yoruba</td>\n", "      <td>Olúṣẹ́gun</td>\n", "      <td>Olusegun</td>\n", "      <td>Olgun</td>\n", "      <td>4</td>\n", "      <td>M-H-H-M</td>\n", "      <td>God has conquered</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>17</th>\n", "      <td>yo18</td>\n", "      <td>yoruba</td>\n", "      <td>Àbẹ̀ní</td>\n", "      <td>Abeni</td>\n", "      <td>bn</td>\n", "      <td>4</td>\n", "      <td>L-L-H</td>\n", "      <td>we asked for her and got her</td>\n", "      <td>check</td>\n", "    </tr>\n", "    <tr>\n", "      <th>18</th>\n", "      <td>yo19</td>\n", "      <td>yoruba</td>\n", "      <td>Yéwándé</td>\n", "      <td>Yewande</td>\n", "      <td>Ywnd</td>\n", "      <td>3</td>\n", "      <td>H-H-H</td>\n", "      <td>mother has returned</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>19</th>\n", "      <td>yo20</td>\n", "      <td>yoruba</td>\n", "      <td>Adéwálé</td>\n", "      <td>Adewale</td>\n", "      <td>Adwl</td>\n", "      <td>3</td>\n", "      <td>M-H-H-H</td>\n", "      <td>the crown has come home</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>20</th>\n", "      <td>ig01</td>\n", "      <td>igbo</td>\n", "      <td>Chiọma</td>\n", "      <td>Chioma</td>\n", "      <td>Chima</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td>good God / good personal spirit</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>21</th>\n", "      <td>ig02</td>\n", "      <td>igbo</td>\n", "      <td>Ụchenna</td>\n", "      <td>Uchenna</td>\n", "      <td>chenna</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td>the father's will / thought</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>22</th>\n", "      <td>ig03</td>\n", "      <td>igbo</td>\n", "      <td>Obiọma</td>\n", "      <td>Obioma</td>\n", "      <td>Obima</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td>kind heart</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>23</th>\n", "      <td>ig04</td>\n", "      <td>igbo</td>\n", "      <td>Ifeọma</td>\n", "      <td>Ifeoma</td>\n", "      <td>Ifema</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td>good thing</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>24</th>\n", "      <td>ig05</td>\n", "      <td>igbo</td>\n", "      <td>Chinọnso</td>\n", "      <td>Chinonso</td>\n", "      <td>Chinnso</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td>God is near</td>\n", "      <td>check</td>\n", "    </tr>\n", "    <tr>\n", "      <th>25</th>\n", "      <td>ig06</td>\n", "      <td>igbo</td>\n", "      <td>Ụzọamaka</td>\n", "      <td>Uzoamaka</td>\n", "      <td>zamaka</td>\n", "      <td>2</td>\n", "      <td></td>\n", "      <td>the road/path is good</td>\n", "      <td>check</td>\n", "    </tr>\n", "    <tr>\n", "      <th>26</th>\n", "      <td>ig07</td>\n", "      <td>igbo</td>\n", "      <td>Ifeanyịchukwu</td>\n", "      <td>Ifeanyichukwu</td>\n", "      <td>Ifeanychukwu</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td>nothing is impossible for God</td>\n", "      <td>check</td>\n", "    </tr>\n", "    <tr>\n", "      <th>27</th>\n", "      <td>ig08</td>\n", "      <td>igbo</td>\n", "      <td>Ngọzi</td>\n", "      <td>Ngozi</td>\n", "      <td>Ngzi</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td>blessing</td>\n", "      <td>check</td>\n", "    </tr>\n", "    <tr>\n", "      <th>28</th>\n", "      <td>ig09</td>\n", "      <td>igbo</td>\n", "      <td>Chukwuebuka</td>\n", "      <td>Chukwuebuka</td>\n", "      <td>Chukwuebuka</td>\n", "      <td>0</td>\n", "      <td></td>\n", "      <td>God is great (no dotted letters: over-correcti...</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>29</th>\n", "      <td>ig10</td>\n", "      <td>igbo</td>\n", "      <td>Nkechi</td>\n", "      <td>Nkechi</td>\n", "      <td>Nkechi</td>\n", "      <td>0</td>\n", "      <td></td>\n", "      <td>God's own (no dotted letters: over-correction ...</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>30</th>\n", "      <td>ha01</td>\n", "      <td>hausa</td>\n", "      <td>Ɗanjuma</td>\n", "      <td>Danjuma</td>\n", "      <td>anjuma</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td>boy born on Friday</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>31</th>\n", "      <td>ha02</td>\n", "      <td>hausa</td>\n", "      <td>Ɗanladi</td>\n", "      <td>Danladi</td>\n", "      <td>anladi</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td>boy born on Sunday</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>32</th>\n", "      <td>ha03</td>\n", "      <td>hausa</td>\n", "      <td>Ɗanazumi</td>\n", "      <td>Danazumi</td>\n", "      <td>anazumi</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td>boy born during the fast (Ramadan)</td>\n", "      <td>check</td>\n", "    </tr>\n", "    <tr>\n", "      <th>33</th>\n", "      <td>ha04</td>\n", "      <td>hausa</td>\n", "      <td>Ladi</td>\n", "      <td>Ladi</td>\n", "      <td>Ladi</td>\n", "      <td>0</td>\n", "      <td></td>\n", "      <td>girl born on Sunday (no hooked letters: control)</td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>34</th>\n", "      <td>ct01</td>\n", "      <td>control</td>\n", "      <td>José</td>\n", "      <td>Jose</td>\n", "      <td>Jos</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td></td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>35</th>\n", "      <td>ct02</td>\n", "      <td>control</td>\n", "      <td>François</td>\n", "      <td>Francois</td>\n", "      <td>Franois</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td></td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>36</th>\n", "      <td>ct03</td>\n", "      <td>control</td>\n", "      <td>Søren</td>\n", "      <td>Soren</td>\n", "      <td>Sren</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td></td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>37</th>\n", "      <td>ct04</td>\n", "      <td>control</td>\n", "      <td>Björn</td>\n", "      <td>Bjorn</td>\n", "      <td>Bjrn</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td></td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>38</th>\n", "      <td>ct05</td>\n", "      <td>control</td>\n", "      <td>Łukasz</td>\n", "      <td>Lukasz</td>\n", "      <td>ukasz</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td></td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>39</th>\n", "      <td>ct06</td>\n", "      <td>control</td>\n", "      <td>Jürgen</td>\n", "      <td>Jurgen</td>\n", "      <td>Jrgen</td>\n", "      <td>1</td>\n", "      <td></td>\n", "      <td></td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>40</th>\n", "      <td>ct07</td>\n", "      <td>control</td>\n", "      <td>Dvořák</td>\n", "      <td>Dvorak</td>\n", "      <td>Dvok</td>\n", "      <td>2</td>\n", "      <td></td>\n", "      <td></td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>41</th>\n", "      <td>ct08</td>\n", "      <td>control</td>\n", "      <td>Nguyễn Thị Hương</td>\n", "      <td>Nguyen Thi Huong</td>\n", "      <td>Nguyn Th Hng</td>\n", "      <td>5</td>\n", "      <td></td>\n", "      <td></td>\n", "      <td>high</td>\n", "    </tr>\n", "    <tr>\n", "      <th>42</th>\n", "      <td>ct09</td>\n", "      <td>control</td>\n", "      <td>Zhāng Wěi</td>\n", "      <td>Zhang Wei</td>\n", "      <td>Zhng Wi</td>\n", "      <td>2</td>\n", "      <td></td>\n", "      <td></td>\n", "      <td>high</td>\n", "    </tr>\n", "  </tbody>\n", "</table>\n", "</div>"]}, "metadata": {}, "execution_count": 5}], "source": ["import importlib, json\n", "import pandas as pd\n", "import oruko_core as oc\n", "importlib.reload(oc)\n", "\n", "names, pairs, prompts = oc.write_dataset(\"data\")\n", "print(len(names), \"names |\", len(pairs), \"minimal-pair words |\", len(prompts), \"prompts\")\n", "print(pd.Series([p[\"task\"] for p in prompts]).value_counts().to_dict())\n", "pd.DataFrame(names)[[\"id\", \"lang\", \"name\", \"stripped\", \"ascii_ignore\", \"n_marks\", \"tone_pattern\", \"meaning\", \"confidence\"]]"]}, {"cell_type": "markdown", "metadata": {"id": "udVXvkCiYy4y"}, "source": ["## 2. Load the model"]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/", "height": 84, "referenced_widgets": ["df82e6ba303345d3abc0e6153b167b49", "541bbb7d39074dc08a17f783b0e46a69", "1a8781a695f64c21894c986b5494fd58", "a9f2a9894cae44c0b3028d5d135ad144", "1f2c6d7259f442da814bbe4ecdece838", "044e17399c674a128d34f72e2225e8f8", "e9897add835f480c8c9c60d3565775a4", "55901d4388d94f7ca3faf43b1126051e", "e78a379c0ae14d4699ca65025ff3b89e", "1df95a82d8f34472a21f15ee11b225d7", "09eb42db9ed64a818e40ec652e0e6040"]}, "id": "oDCxLC2wYy4y", "executionInfo": {"status": "ok", "timestamp": 1792105843815, "user_tz": -60, "elapsed": 12347, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "2d27ee82-953e-439e-b998-60a7ad3efdb2"}, "execution_count": 6, "outputs": [{"output_type": "display_data", "data": {"text/plain": ["Loading weights:   0%|          | 0/723 [00:00<?, ?it/s]"], "application/vnd.jupyter.widget-view+json": {"version_major": 2, "version_minor": 0, "model_id": "df82e6ba303345d3abc0e6153b167b49"}}, "metadata": {}}, {"output_type": "stream", "name": "stdout", "text": ["Loaded with AutoModelForImageTextToText in torch.bfloat16\n", "Parameters (incl. vision encoder): 4.54B\n"]}], "source": ["import re, time, torch, unicodedata as ud\n", "from transformers import AutoTokenizer\n", "import transformers\n", "\n", "tok = AutoTokenizer.from_pretrained(MODEL_ID)\n", "if getattr(tok, \"chat_template\", None) is None:\n", "    from transformers import AutoProcessor\n", "    tok.chat_template = AutoProcessor.from_pretrained(MODEL_ID).chat_template\n", "tok.padding_side = \"left\"\n", "if tok.pad_token is None:\n", "    tok.pad_token = tok.eos_token\n", "\n", "dtype = torch.bfloat16 if (torch.cuda.is_available() and torch.cuda.is_bf16_supported()) else torch.float16\n", "model = None\n", "for cls_name in [\"AutoModelForImageTextToText\", \"AutoModelForMultimodalLM\", \"AutoModelForCausalLM\"]:\n", "    cls = getattr(transformers, cls_name, None)\n", "    if cls is None:\n", "        continue\n", "    try:\n", "        model = cls.from_pretrained(MODEL_ID, dtype=dtype, device_map=\"auto\")\n", "        print(\"Loaded with\", cls_name, \"in\", dtype)\n", "        break\n", "    except Exception as e:\n", "        print(cls_name, \"failed:\", str(e)[:300])\n", "assert model is not None, \"Could not load the model; try installing transformers from main (see first cell).\"\n", "model.eval()\n", "DEVICE = next(model.parameters()).device\n", "print(\"Parameters (incl. vision encoder): %.2fB\" % (sum(p.numel() for p in model.parameters()) / 1e9))"]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/"}, "id": "m7ekMpcbYy4y", "executionInfo": {"status": "ok", "timestamp": 1792105845038, "user_tz": -60, "elapsed": 1216, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "40800aa5-385b-4dff-d992-438a07772bfb"}, "execution_count": 7, "outputs": [{"output_type": "stream", "name": "stdout", "text": ["<|im_start|>user\n", "Repeat this name: Ọláyínká<|im_end|>\n", "<|im_start|>assistant\n", "<think>\n", "\n", "</think>\n", "\n", "\n"]}, {"output_type": "stream", "name": "stderr", "text": ["[transformers] `causal_conv1d_fn` is falling back to its reference PyTorch implementation because `causal_conv1d` is not installed. This is correct but much slower; install `causal_conv1d` for the optimized kernel.\n", "[transformers] `chunk_gated_delta_rule` is falling back to its reference PyTorch implementation because `flash-linear-attention` is not installed. This is correct but much slower; install `flash-linear-attention` for the optimized kernel.\n", "[transformers] `causal_conv1d_update` is falling back to its reference PyTorch implementation because `causal_conv1d` is not installed. This is correct but much slower; install `causal_conv1d` for the optimized kernel.\n", "[transformers] `fused_recurrent_gated_delta_rule` is falling back to its reference PyTorch implementation because `flash-linear-attention` is not installed. This is correct but much slower; install `flash-linear-attention` for the optimized kernel.\n"]}, {"output_type": "stream", "name": "stdout", "text": ["→ ['Ọláyínká']\n"]}], "source": ["def render(prompt):\n", "    \"\"\"Chat-format a single user turn with thinking disabled.\"\"\"\n", "    for content in (prompt, [{\"type\": \"text\", \"text\": prompt}]):\n", "        try:\n", "            return tok.apply_chat_template([{\"role\": \"user\", \"content\": content}], tokenize=False,\n", "                                           add_generation_prompt=True, enable_thinking=False)\n", "        except Exception as e:\n", "            err = e\n", "    raise err\n", "\n", "def clean(text):\n", "    text = re.sub(r\"<think>.*?</think>\", \"\", text, flags=re.S)\n", "    return ud.normalize(\"NFC\", text.replace(\"<think>\", \"\").replace(\"</think>\", \"\").strip())\n", "\n", "MAX_NEW = {\"ambiguity\": 400, \"copy\": 160, \"meaning\": 120, \"restore\": 48, \"tone\": 48, \"pair\": 32}\n", "\n", "@torch.no_grad()\n", "def generate(batch_prompts, max_new_tokens):\n", "    texts = [render(p) for p in batch_prompts]\n", "    enc = tok(texts, return_tensors=\"pt\", padding=True, add_special_tokens=False).to(DEVICE)\n", "    out = model.generate(**enc, max_new_tokens=max_new_tokens, do_sample=False,\n", "                         temperature=None, top_p=None, top_k=None, pad_token_id=tok.pad_token_id)\n", "    return [clean(tok.decode(row, skip_special_tokens=True)) for row in out[:, enc[\"input_ids\"].shape[1]:]]\n", "\n", "print(render(\"Repeat this name: Ọláyínká\"))\n", "print(\"→\", generate([\"Repeat the following name exactly as written, with no changes: Ọláyínká\"], 32))"]}, {"cell_type": "markdown", "metadata": {"id": "Xbwm2qSzYy4y"}, "source": ["## 3. Run all prompts (cached and resumable)"]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/"}, "id": "2kvxqZQwYy4y", "executionInfo": {"status": "ok", "timestamp": 1792105845074, "user_tz": -60, "elapsed": 18, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "038bb062-cc6a-4a33-abc3-dc20be143643"}, "execution_count": 8, "outputs": [{"output_type": "stream", "name": "stdout", "text": ["363 cached, 0 to run\n", "done in 0.0 min\n"]}], "source": ["os.makedirs(\"results\", exist_ok=True)\n", "GEN_PATH = \"results/generations.jsonl\"\n", "gens = {}\n", "if os.path.exists(GEN_PATH):\n", "    with open(GEN_PATH, encoding=\"utf-8\") as f:\n", "        for line in f:\n", "            r = json.loads(line); gens[r[\"pid\"]] = r[\"output\"]\n", "todo = [p for p in prompts if p[\"pid\"] not in gens]\n", "todo.sort(key=lambda p: (p[\"task\"], len(p[\"prompt\"])))\n", "print(f\"{len(gens)} cached, {len(todo)} to run\")\n", "\n", "t0 = time.time()\n", "with open(GEN_PATH, \"a\", encoding=\"utf-8\") as f:\n", "    for i in range(0, len(todo), BATCH_SIZE):\n", "        batch = [p for p in todo[i:i + BATCH_SIZE]]\n", "        mx = max(MAX_NEW[p[\"task\"]] for p in batch)\n", "        outs = generate([p[\"prompt\"] for p in batch], mx)\n", "        for p, o in zip(batch, outs):\n", "            gens[p[\"pid\"]] = o\n", "            f.write(json.dumps(dict(pid=p[\"pid\"], task=p[\"task\"], template=p[\"template\"], item=p[\"item\"],\n", "                                    prompt=p[\"prompt\"], output=o, model=MODEL_ID), ensure_ascii=False) + \"\\n\")\n", "        f.flush()\n", "        if (i // BATCH_SIZE) % 5 == 0:\n", "            print(f\"{i + len(batch)}/{len(todo)}  ({time.time() - t0:.0f}s)  e.g. {batch[0]['task']}: {outs[0][:80]!r}\")\n", "print(\"done in %.1f min\" % ((time.time() - t0) / 60))"]}, {"cell_type": "markdown", "metadata": {"id": "qeyYYE52Yy4y"}, "source": ["## 4. Score"]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/", "height": 488}, "id": "gf7iTVnDYy4y", "executionInfo": {"status": "ok", "timestamp": 1792105845114, "user_tz": -60, "elapsed": 38, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "302f6b5a-31a5-4932-aa7b-af0a7a6f14de"}, "execution_count": 9, "outputs": [{"output_type": "execute_result", "data": {"text/plain": ["                             yoruba   igbo  hausa  control\n", "copy_exact                    0.860  0.800  0.867    0.956\n", "copy_fully_stripped           0.040  0.175  0.133    0.044\n", "copy_catastrophic             0.000  0.000  0.000    0.000\n", "copy_mark_recall              0.923  0.822  0.867    0.907\n", "restore_exact                 0.000  0.125  0.000    1.000\n", "restore_mark_f1               0.453  0.200  0.000    1.000\n", "restore_overcorrection          NaN  0.000  0.000      NaN\n", "meaning_acc_marked            0.050  0.000  0.500      NaN\n", "meaning_acc_stripped          0.000  0.000  0.667      NaN\n", "minimal_pair_acc              0.250  0.000  0.250      NaN\n", "ambiguity_sense_recall        0.125  0.000  0.250      NaN\n", "ambiguity_spelled_correctly   0.167  1.000  0.000      NaN\n", "tone_read_exact               0.050    NaN    NaN      NaN\n", "tone_read_position_acc        0.349    NaN    NaN      NaN"], "text/html": ["<div>\n", "<style scoped>\n", "    .dataframe tbody tr th:only-of-type {\n", "        vertical-align: middle;\n", "    }\n", "\n", "    .dataframe tbody tr th {\n", "        vertical-align: top;\n", "    }\n", "\n", "    .dataframe thead th {\n", "        text-align: right;\n", "    }\n", "</style>\n", "<table border=\"1\" class=\"dataframe\">\n", "  <thead>\n", "    <tr style=\"text-align: right;\">\n", "      <th></th>\n", "      <th>yoruba</th>\n", "      <th>igbo</th>\n", "      <th>hausa</th>\n", "      <th>control</th>\n", "    </tr>\n", "  </thead>\n", "  <tbody>\n", "    <tr>\n", "      <th>copy_exact</th>\n", "      <td>0.860</td>\n", "      <td>0.800</td>\n", "      <td>0.867</td>\n", "      <td>0.956</td>\n", "    </tr>\n", "    <tr>\n", "      <th>copy_fully_stripped</th>\n", "      <td>0.040</td>\n", "      <td>0.175</td>\n", "      <td>0.133</td>\n", "      <td>0.044</td>\n", "    </tr>\n", "    <tr>\n", "      <th>copy_catastrophic</th>\n", "      <td>0.000</td>\n", "      <td>0.000</td>\n", "      <td>0.000</td>\n", "      <td>0.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>copy_mark_recall</th>\n", "      <td>0.923</td>\n", "      <td>0.822</td>\n", "      <td>0.867</td>\n", "      <td>0.907</td>\n", "    </tr>\n", "    <tr>\n", "      <th>restore_exact</th>\n", "      <td>0.000</td>\n", "      <td>0.125</td>\n", "      <td>0.000</td>\n", "      <td>1.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>restore_mark_f1</th>\n", "      <td>0.453</td>\n", "      <td>0.200</td>\n", "      <td>0.000</td>\n", "      <td>1.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>restore_overcorrection</th>\n", "      <td>NaN</td>\n", "      <td>0.000</td>\n", "      <td>0.000</td>\n", "      <td>NaN</td>\n", "    </tr>\n", "    <tr>\n", "      <th>meaning_acc_marked</th>\n", "      <td>0.050</td>\n", "      <td>0.000</td>\n", "      <td>0.500</td>\n", "      <td>NaN</td>\n", "    </tr>\n", "    <tr>\n", "      <th>meaning_acc_stripped</th>\n", "      <td>0.000</td>\n", "      <td>0.000</td>\n", "      <td>0.667</td>\n", "      <td>NaN</td>\n", "    </tr>\n", "    <tr>\n", "      <th>minimal_pair_acc</th>\n", "      <td>0.250</td>\n", "      <td>0.000</td>\n", "      <td>0.250</td>\n", "      <td>NaN</td>\n", "    </tr>\n", "    <tr>\n", "      <th>ambiguity_sense_recall</th>\n", "      <td>0.125</td>\n", "      <td>0.000</td>\n", "      <td>0.250</td>\n", "      <td>NaN</td>\n", "    </tr>\n", "    <tr>\n", "      <th>ambiguity_spelled_correctly</th>\n", "      <td>0.167</td>\n", "      <td>1.000</td>\n", "      <td>0.000</td>\n", "      <td>NaN</td>\n", "    </tr>\n", "    <tr>\n", "      <th>tone_read_exact</th>\n", "      <td>0.050</td>\n", "      <td>NaN</td>\n", "      <td>NaN</td>\n", "      <td>NaN</td>\n", "    </tr>\n", "    <tr>\n", "      <th>tone_read_position_acc</th>\n", "      <td>0.349</td>\n", "      <td>NaN</td>\n", "      <td>NaN</td>\n", "      <td>NaN</td>\n", "    </tr>\n", "  </tbody>\n", "</table>\n", "</div>"]}, "metadata": {}, "execution_count": 9}], "source": ["importlib.reload(oc)\n", "res = oc.score_all(names, pairs, prompts, gens)\n", "for task, rows in res.items():\n", "    pd.DataFrame(rows).to_csv(f\"results/scores_{task}.csv\", index=False)\n", "summary = oc.summarize(res)\n", "by_template = oc.copy_by_template(res)\n", "lost_copy = oc.lost_mark_counts(res, \"copy\")\n", "pd.DataFrame(summary)"]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/", "height": 1000}, "id": "bkqWtoFiYy4y", "executionInfo": {"status": "ok", "timestamp": 1792105845145, "user_tz": -60, "elapsed": 28, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "e60ecfb6-e77d-45a9-a567-9007ace608ab"}, "execution_count": 10, "outputs": [{"output_type": "execute_result", "data": {"text/plain": ["        lang              name status         rendition\n", "2     yoruba     Olúwadámilárè  exact     Olúwadámilárè\n", "7     yoruba           Adéṣọlá  exact           Adéṣọlá\n", "12    yoruba          Ọláyínká  exact          Ọláyínká\n", "17    yoruba          Fọláṣadé  exact          Fọláṣadé\n", "22    yoruba         Olúwaṣeun  exact         Olúwaṣeun\n", "27    yoruba          Ayọ̀délé  exact          Ayọ̀délé\n", "32    yoruba          Ìyábọ̀dé  exact          Ìyábọ̀dé\n", "37    yoruba           Bọ́lájí  exact           Bọ́lájí\n", "42    yoruba        Fúnmiláyọ̀  exact        Fúnmiláyọ̀\n", "47    yoruba          Adébáyọ̀  exact          Adébáyọ̀\n", "52    yoruba           Ọlámidé  exact           Ọlámidé\n", "57    yoruba         Ìfẹ́olúwa  exact         Ìfẹ́olúwa\n", "62    yoruba             Táíwò  exact             Táíwò\n", "67    yoruba          Kẹ́hìndé  exact          Kẹ́hìndé\n", "72    yoruba             Ìdòwú  exact             Ìdòwú\n", "77    yoruba         Babátúndé  exact         Babátúndé\n", "82    yoruba         Olúṣẹ́gun  exact         Olúṣẹ́gun\n", "87    yoruba            Àbẹ̀ní  exact            Àbẹ̀ní\n", "92    yoruba           Yéwándé  exact           Yéwándé\n", "97    yoruba           Adéwálé  exact           Adéwálé\n", "102     igbo            Chiọma  exact            Chiọma\n", "107     igbo           Ụchenna  exact           Ụchenna\n", "112     igbo            Obiọma  exact            Obiọma\n", "117     igbo            Ifeọma  exact            Ifeọma\n", "122     igbo          Chinọnso  exact          Chinọnso\n", "127     igbo          Ụzọamaka  exact          Ụzọamaka\n", "132     igbo     Ifeanyịchukwu  exact     Ifeanyịchukwu\n", "137     igbo             Ngọzi  exact             Ngọzi\n", "142     igbo       Chukwuebuka  exact       Chukwuebuka\n", "147     igbo            Nkechi  exact            Nkechi\n", "152    hausa           Ɗanjuma  exact           Ɗanjuma\n", "157    hausa           Ɗanladi  exact           Ɗanladi\n", "162    hausa          Ɗanazumi  exact          Ɗanazumi\n", "167    hausa              Ladi  exact              Ladi\n", "172  control              José  exact              José\n", "177  control          François  exact          François\n", "182  control             Søren  exact             Søren\n", "187  control             Björn  exact             Björn\n", "192  control            Łukasz  exact            Łukasz\n", "197  control            Jürgen  exact            Jürgen\n", "202  control            Dvořák  exact            Dvořák\n", "207  control  Nguyễn Thị Hương  exact  Nguyễn Thị Hương\n", "212  control         Zhāng Wěi  exact         Zhāng Wěi"], "text/html": ["<div>\n", "<style scoped>\n", "    .dataframe tbody tr th:only-of-type {\n", "        vertical-align: middle;\n", "    }\n", "\n", "    .dataframe tbody tr th {\n", "        vertical-align: top;\n", "    }\n", "\n", "    .dataframe thead th {\n", "        text-align: right;\n", "    }\n", "</style>\n", "<table border=\"1\" class=\"dataframe\">\n", "  <thead>\n", "    <tr style=\"text-align: right;\">\n", "      <th></th>\n", "      <th>lang</th>\n", "      <th>name</th>\n", "      <th>status</th>\n", "      <th>rendition</th>\n", "    </tr>\n", "  </thead>\n", "  <tbody>\n", "    <tr>\n", "      <th>2</th>\n", "      <td>yoruba</td>\n", "      <td>Olúwadámilárè</td>\n", "      <td>exact</td>\n", "      <td>Olúwadámilárè</td>\n", "    </tr>\n", "    <tr>\n", "      <th>7</th>\n", "      <td>yoruba</td>\n", "      <td>Adéṣọlá</td>\n", "      <td>exact</td>\n", "      <td>Adéṣọlá</td>\n", "    </tr>\n", "    <tr>\n", "      <th>12</th>\n", "      <td>yoruba</td>\n", "      <td>Ọláyínká</td>\n", "      <td>exact</td>\n", "      <td>Ọláyínká</td>\n", "    </tr>\n", "    <tr>\n", "      <th>17</th>\n", "      <td>yoruba</td>\n", "      <td>Fọláṣadé</td>\n", "      <td>exact</td>\n", "      <td>Fọláṣadé</td>\n", "    </tr>\n", "    <tr>\n", "      <th>22</th>\n", "      <td>yoruba</td>\n", "      <td>Olúwaṣeun</td>\n", "      <td>exact</td>\n", "      <td>Olúwaṣeun</td>\n", "    </tr>\n", "    <tr>\n", "      <th>27</th>\n", "      <td>yoruba</td>\n", "      <td>Ayọ̀délé</td>\n", "      <td>exact</td>\n", "      <td>Ayọ̀délé</td>\n", "    </tr>\n", "    <tr>\n", "      <th>32</th>\n", "      <td>yoruba</td>\n", "      <td>Ìyábọ̀dé</td>\n", "      <td>exact</td>\n", "      <td>Ìyábọ̀dé</td>\n", "    </tr>\n", "    <tr>\n", "      <th>37</th>\n", "      <td>yoruba</td>\n", "      <td>Bọ́lájí</td>\n", "      <td>exact</td>\n", "      <td>Bọ́lájí</td>\n", "    </tr>\n", "    <tr>\n", "      <th>42</th>\n", "      <td>yoruba</td>\n", "      <td>Fúnmiláyọ̀</td>\n", "      <td>exact</td>\n", "      <td>Fúnmiláyọ̀</td>\n", "    </tr>\n", "    <tr>\n", "      <th>47</th>\n", "      <td>yoruba</td>\n", "      <td>Adébáyọ̀</td>\n", "      <td>exact</td>\n", "      <td>Adébáyọ̀</td>\n", "    </tr>\n", "    <tr>\n", "      <th>52</th>\n", "      <td>yoruba</td>\n", "      <td>Ọlámidé</td>\n", "      <td>exact</td>\n", "      <td>Ọlámidé</td>\n", "    </tr>\n", "    <tr>\n", "      <th>57</th>\n", "      <td>yoruba</td>\n", "      <td>Ìfẹ́olúwa</td>\n", "      <td>exact</td>\n", "      <td>Ìfẹ́olúwa</td>\n", "    </tr>\n", "    <tr>\n", "      <th>62</th>\n", "      <td>yoruba</td>\n", "      <td>Táíwò</td>\n", "      <td>exact</td>\n", "      <td>Táíwò</td>\n", "    </tr>\n", "    <tr>\n", "      <th>67</th>\n", "      <td>yoruba</td>\n", "      <td>Kẹ́hìndé</td>\n", "      <td>exact</td>\n", "      <td>Kẹ́hìndé</td>\n", "    </tr>\n", "    <tr>\n", "      <th>72</th>\n", "      <td>yoruba</td>\n", "      <td>Ìdòwú</td>\n", "      <td>exact</td>\n", "      <td>Ìdòwú</td>\n", "    </tr>\n", "    <tr>\n", "      <th>77</th>\n", "      <td>yoruba</td>\n", "      <td>Babátúndé</td>\n", "      <td>exact</td>\n", "      <td>Babátúndé</td>\n", "    </tr>\n", "    <tr>\n", "      <th>82</th>\n", "      <td>yoruba</td>\n", "      <td>Olúṣẹ́gun</td>\n", "      <td>exact</td>\n", "      <td>Olúṣẹ́gun</td>\n", "    </tr>\n", "    <tr>\n", "      <th>87</th>\n", "      <td>yoruba</td>\n", "      <td>Àbẹ̀ní</td>\n", "      <td>exact</td>\n", "      <td>Àbẹ̀ní</td>\n", "    </tr>\n", "    <tr>\n", "      <th>92</th>\n", "      <td>yoruba</td>\n", "      <td>Yéwándé</td>\n", "      <td>exact</td>\n", "      <td>Yéwándé</td>\n", "    </tr>\n", "    <tr>\n", "      <th>97</th>\n", "      <td>yoruba</td>\n", "      <td>Adéwálé</td>\n", "      <td>exact</td>\n", "      <td>Adéwálé</td>\n", "    </tr>\n", "    <tr>\n", "      <th>102</th>\n", "      <td>igbo</td>\n", "      <td>Chiọma</td>\n", "      <td>exact</td>\n", "      <td>Chiọma</td>\n", "    </tr>\n", "    <tr>\n", "      <th>107</th>\n", "      <td>igbo</td>\n", "      <td>Ụchenna</td>\n", "      <td>exact</td>\n", "      <td>Ụchenna</td>\n", "    </tr>\n", "    <tr>\n", "      <th>112</th>\n", "      <td>igbo</td>\n", "      <td>Obiọma</td>\n", "      <td>exact</td>\n", "      <td>Obiọma</td>\n", "    </tr>\n", "    <tr>\n", "      <th>117</th>\n", "      <td>igbo</td>\n", "      <td>Ifeọma</td>\n", "      <td>exact</td>\n", "      <td>Ifeọma</td>\n", "    </tr>\n", "    <tr>\n", "      <th>122</th>\n", "      <td>igbo</td>\n", "      <td>Chinọnso</td>\n", "      <td>exact</td>\n", "      <td>Chinọnso</td>\n", "    </tr>\n", "    <tr>\n", "      <th>127</th>\n", "      <td>igbo</td>\n", "      <td>Ụzọamaka</td>\n", "      <td>exact</td>\n", "      <td>Ụzọamaka</td>\n", "    </tr>\n", "    <tr>\n", "      <th>132</th>\n", "      <td>igbo</td>\n", "      <td>Ifeanyịchukwu</td>\n", "      <td>exact</td>\n", "      <td>Ifeanyịchukwu</td>\n", "    </tr>\n", "    <tr>\n", "      <th>137</th>\n", "      <td>igbo</td>\n", "      <td>Ngọzi</td>\n", "      <td>exact</td>\n", "      <td>Ngọzi</td>\n", "    </tr>\n", "    <tr>\n", "      <th>142</th>\n", "      <td>igbo</td>\n", "      <td>Chukwuebuka</td>\n", "      <td>exact</td>\n", "      <td>Chukwuebuka</td>\n", "    </tr>\n", "    <tr>\n", "      <th>147</th>\n", "      <td>igbo</td>\n", "      <td>Nkechi</td>\n", "      <td>exact</td>\n", "      <td>Nkechi</td>\n", "    </tr>\n", "    <tr>\n", "      <th>152</th>\n", "      <td>hausa</td>\n", "      <td>Ɗanjuma</td>\n", "      <td>exact</td>\n", "      <td>Ɗanjuma</td>\n", "    </tr>\n", "    <tr>\n", "      <th>157</th>\n", "      <td>hausa</td>\n", "      <td>Ɗanladi</td>\n", "      <td>exact</td>\n", "      <td>Ɗanladi</td>\n", "    </tr>\n", "    <tr>\n", "      <th>162</th>\n", "      <td>hausa</td>\n", "      <td>Ɗanazumi</td>\n", "      <td>exact</td>\n", "      <td>Ɗanazumi</td>\n", "    </tr>\n", "    <tr>\n", "      <th>167</th>\n", "      <td>hausa</td>\n", "      <td>Ladi</td>\n", "      <td>exact</td>\n", "      <td>Ladi</td>\n", "    </tr>\n", "    <tr>\n", "      <th>172</th>\n", "      <td>control</td>\n", "      <td>José</td>\n", "      <td>exact</td>\n", "      <td>José</td>\n", "    </tr>\n", "    <tr>\n", "      <th>177</th>\n", "      <td>control</td>\n", "      <td>François</td>\n", "      <td>exact</td>\n", "      <td>François</td>\n", "    </tr>\n", "    <tr>\n", "      <th>182</th>\n", "      <td>control</td>\n", "      <td>Søren</td>\n", "      <td>exact</td>\n", "      <td>Søren</td>\n", "    </tr>\n", "    <tr>\n", "      <th>187</th>\n", "      <td>control</td>\n", "      <td>Björn</td>\n", "      <td>exact</td>\n", "      <td>Björn</td>\n", "    </tr>\n", "    <tr>\n", "      <th>192</th>\n", "      <td>control</td>\n", "      <td>Łukasz</td>\n", "      <td>exact</td>\n", "      <td>Łukasz</td>\n", "    </tr>\n", "    <tr>\n", "      <th>197</th>\n", "      <td>control</td>\n", "      <td>Jürgen</td>\n", "      <td>exact</td>\n", "      <td>Jürgen</td>\n", "    </tr>\n", "    <tr>\n", "      <th>202</th>\n", "      <td>control</td>\n", "      <td>Dvořák</td>\n", "      <td>exact</td>\n", "      <td>Dvořák</td>\n", "    </tr>\n", "    <tr>\n", "      <th>207</th>\n", "      <td>control</td>\n", "      <td>Nguyễn Thị Hương</td>\n", "      <td>exact</td>\n", "      <td>Nguyễn Thị Hương</td>\n", "    </tr>\n", "    <tr>\n", "      <th>212</th>\n", "      <td>control</td>\n", "      <td>Zhāng Wěi</td>\n", "      <td>exact</td>\n", "      <td>Zhāng Wěi</td>\n", "    </tr>\n", "  </tbody>\n", "</table>\n", "</div>"]}, "metadata": {}, "execution_count": 10}], "source": ["# Look at real examples: what the model did to each name in the JSON patient-record task\n", "copy_df = pd.DataFrame(res[\"copy\"])\n", "copy_df[copy_df.template == \"c3_json\"][[\"lang\", \"name\", \"status\", \"rendition\"]]"]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/", "height": 551}, "id": "yY1ovk8aYy4y", "executionInfo": {"status": "ok", "timestamp": 1792105845177, "user_tz": -60, "elapsed": 30, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "2025da86-2326-4c2f-b1b9-78bfb038a816"}, "execution_count": 11, "outputs": [{"output_type": "execute_result", "data": {"text/plain": ["      lang  word                 gloss  correct   output\n", "0   yoruba   oko                  farm    False    House\n", "1   yoruba   ọkọ               husband    False      Car\n", "2   yoruba  ọkọ̀        vehicle / boat     True      Car\n", "3   yoruba  ọkọ́                   hoe    False      Car\n", "4   yoruba  igbá              calabash    False     Time\n", "5   yoruba  ìgbà         time / period     True     Time\n", "6   yoruba  igba           two hundred    False     Time\n", "7   yoruba  ìgbá            garden egg    False     Time\n", "8     igbo  àkwà                   bed    False   Beauty\n", "9     igbo  ákwà                 cloth    False    Money\n", "10    igbo  ákwá                   cry    False    Water\n", "11    igbo  àkwá                   egg    False    Water\n", "12   hausa  kara          stalk / reed    False      Add\n", "13   hausa  ƙara  increase / complaint     True      Add\n", "14   hausa  dari        cold (weather)    False  Nothing\n", "15   hausa  ɗari               hundred    False    Seven"], "text/html": ["<div>\n", "<style scoped>\n", "    .dataframe tbody tr th:only-of-type {\n", "        vertical-align: middle;\n", "    }\n", "\n", "    .dataframe tbody tr th {\n", "        vertical-align: top;\n", "    }\n", "\n", "    .dataframe thead th {\n", "        text-align: right;\n", "    }\n", "</style>\n", "<table border=\"1\" class=\"dataframe\">\n", "  <thead>\n", "    <tr style=\"text-align: right;\">\n", "      <th></th>\n", "      <th>lang</th>\n", "      <th>word</th>\n", "      <th>gloss</th>\n", "      <th>correct</th>\n", "      <th>output</th>\n", "    </tr>\n", "  </thead>\n", "  <tbody>\n", "    <tr>\n", "      <th>0</th>\n", "      <td>yoruba</td>\n", "      <td>oko</td>\n", "      <td>farm</td>\n", "      <td>False</td>\n", "      <td>House</td>\n", "    </tr>\n", "    <tr>\n", "      <th>1</th>\n", "      <td>yoruba</td>\n", "      <td>ọkọ</td>\n", "      <td>husband</td>\n", "      <td>False</td>\n", "      <td>Car</td>\n", "    </tr>\n", "    <tr>\n", "      <th>2</th>\n", "      <td>yoruba</td>\n", "      <td>ọkọ̀</td>\n", "      <td>vehicle / boat</td>\n", "      <td>True</td>\n", "      <td>Car</td>\n", "    </tr>\n", "    <tr>\n", "      <th>3</th>\n", "      <td>yoruba</td>\n", "      <td>ọkọ́</td>\n", "      <td>hoe</td>\n", "      <td>False</td>\n", "      <td>Car</td>\n", "    </tr>\n", "    <tr>\n", "      <th>4</th>\n", "      <td>yoruba</td>\n", "      <td>igbá</td>\n", "      <td>calabash</td>\n", "      <td>False</td>\n", "      <td>Time</td>\n", "    </tr>\n", "    <tr>\n", "      <th>5</th>\n", "      <td>yoruba</td>\n", "      <td>ìgbà</td>\n", "      <td>time / period</td>\n", "      <td>True</td>\n", "      <td>Time</td>\n", "    </tr>\n", "    <tr>\n", "      <th>6</th>\n", "      <td>yoruba</td>\n", "      <td>igba</td>\n", "      <td>two hundred</td>\n", "      <td>False</td>\n", "      <td>Time</td>\n", "    </tr>\n", "    <tr>\n", "      <th>7</th>\n", "      <td>yoruba</td>\n", "      <td>ìgbá</td>\n", "      <td>garden egg</td>\n", "      <td>False</td>\n", "      <td>Time</td>\n", "    </tr>\n", "    <tr>\n", "      <th>8</th>\n", "      <td>igbo</td>\n", "      <td>àkwà</td>\n", "      <td>bed</td>\n", "      <td>False</td>\n", "      <td>Beauty</td>\n", "    </tr>\n", "    <tr>\n", "      <th>9</th>\n", "      <td>igbo</td>\n", "      <td>ákwà</td>\n", "      <td>cloth</td>\n", "      <td>False</td>\n", "      <td>Money</td>\n", "    </tr>\n", "    <tr>\n", "      <th>10</th>\n", "      <td>igbo</td>\n", "      <td>ákwá</td>\n", "      <td>cry</td>\n", "      <td>False</td>\n", "      <td>Water</td>\n", "    </tr>\n", "    <tr>\n", "      <th>11</th>\n", "      <td>igbo</td>\n", "      <td>àkwá</td>\n", "      <td>egg</td>\n", "      <td>False</td>\n", "      <td>Water</td>\n", "    </tr>\n", "    <tr>\n", "      <th>12</th>\n", "      <td>hausa</td>\n", "      <td>kara</td>\n", "      <td>stalk / reed</td>\n", "      <td>False</td>\n", "      <td>Add</td>\n", "    </tr>\n", "    <tr>\n", "      <th>13</th>\n", "      <td>hausa</td>\n", "      <td>ƙara</td>\n", "      <td>increase / complaint</td>\n", "      <td>True</td>\n", "      <td>Add</td>\n", "    </tr>\n", "    <tr>\n", "      <th>14</th>\n", "      <td>hausa</td>\n", "      <td>dari</td>\n", "      <td>cold (weather)</td>\n", "      <td>False</td>\n", "      <td>Nothing</td>\n", "    </tr>\n", "    <tr>\n", "      <th>15</th>\n", "      <td>hausa</td>\n", "      <td>ɗari</td>\n", "      <td>hundred</td>\n", "      <td>False</td>\n", "      <td>Seven</td>\n", "    </tr>\n", "  </tbody>\n", "</table>\n", "</div>"]}, "metadata": {}, "execution_count": 11}], "source": ["pd.DataFrame(res[\"pair\"])[[\"lang\", \"word\", \"gloss\", \"correct\", \"output\"]]"]}, {"cell_type": "markdown", "metadata": {"id": "8KZkpJnxYy4y"}, "source": ["## 5. Tokenizer fragmentation and likelihood preference"]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/", "height": 1000}, "id": "tyesLi8PYy4y", "executionInfo": {"status": "ok", "timestamp": 1792105845200, "user_tz": -60, "elapsed": 21, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "ce8354e2-bf3d-4acb-e696-070cdfab0194"}, "execution_count": 12, "outputs": [{"output_type": "stream", "name": "stdout", "text": ["mean tokens(marked)/tokens(stripped): {'control': 1.23, 'hausa': 1.89, 'igbo': 1.31, 'yoruba': 2.13}\n"]}, {"output_type": "execute_result", "data": {"text/plain": ["      id     lang              name          stripped  tokens_marked  \\\n", "0   yo01   yoruba     Olúwadámilárè     Oluwadamilare              7   \n", "1   yo02   yoruba           Adéṣọlá           Adesola              7   \n", "2   yo03   yoruba          Ọláyínká          Olayinka              6   \n", "3   yo04   yoruba          Fọláṣadé          Folasade              8   \n", "4   yo05   yoruba         Olúwaṣeun         Oluwaseun              8   \n", "5   yo06   yoruba          Ayọ̀délé           Ayodele              5   \n", "6   yo07   yoruba          Ìyábọ̀dé           Iyabode              7   \n", "7   yo08   yoruba           Bọ́lájí            Bolaji              6   \n", "8   yo09   yoruba        Fúnmiláyọ̀         Funmilayo              7   \n", "9   yo10   yoruba          Adébáyọ̀           Adebayo              5   \n", "10  yo11   yoruba           Ọlámidé           Olamide              5   \n", "11  yo12   yoruba         Ìfẹ́olúwa          Ifeoluwa              8   \n", "12  yo13   yoruba             Táíwò             Taiwo              4   \n", "13  yo14   yoruba          Kẹ́hìndé           Kehinde              7   \n", "14  yo15   yoruba             Ìdòwú             Idowu              6   \n", "15  yo16   yoruba         Babátúndé         Babatunde              5   \n", "16  yo17   yoruba         Olúṣẹ́gun          Olusegun              8   \n", "17  yo18   yoruba            Àbẹ̀ní             Abeni              5   \n", "18  yo19   yoruba           Yéwándé           Yewande              5   \n", "19  yo20   yoruba           Adéwálé           Adewale              5   \n", "20  ig01     igbo            Chiọma            Chioma              3   \n", "21  ig02     igbo           Ụchenna           Uchenna              4   \n", "22  ig03     igbo            Obiọma            Obioma              4   \n", "23  ig04     igbo            Ifeọma            Ifeoma              4   \n", "24  ig05     igbo          Chinọnso          Chinonso              3   \n", "25  ig06     igbo          Ụzọamaka          Uzoamaka              6   \n", "26  ig07     igbo     Ifeanyịchukwu     Ifeanyichukwu              6   \n", "27  ig08     igbo             Ngọzi             Ngozi              2   \n", "28  ha01    hausa           Ɗanjuma           Danjuma              5   \n", "29  ha02    hausa           Ɗanladi           Danladi              6   \n", "30  ha03    hausa          Ɗanazumi          Danazumi              6   \n", "31  ct01  control              José              Jose              1   \n", "32  ct02  control          François          Francois              1   \n", "33  ct03  control             Søren             Soren              2   \n", "34  ct04  control             Björn             Bjorn              2   \n", "35  ct05  control            Łukasz            Lukasz              3   \n", "36  ct06  control            Jürgen            Jurgen              1   \n", "37  ct07  control            Dvořák            Dvorak              4   \n", "38  ct08  control  Nguyễn Thị Hương  Nguyen Thi Huong              3   \n", "39  ct09  control         Zhāng Wěi         Zhang Wei              6   \n", "\n", "    tokens_stripped  ratio  \n", "0                 5  1.400  \n", "1                 3  2.333  \n", "2                 3  2.000  \n", "3                 3  2.667  \n", "4                 5  1.600  \n", "5                 3  1.667  \n", "6                 4  1.750  \n", "7                 2  3.000  \n", "8                 3  2.333  \n", "9                 3  1.667  \n", "10                2  2.500  \n", "11                4  2.000  \n", "12                2  2.000  \n", "13                2  3.500  \n", "14                3  2.000  \n", "15                3  1.667  \n", "16                3  2.667  \n", "17                2  2.500  \n", "18                3  1.667  \n", "19                3  1.667  \n", "20                2  1.500  \n", "21                3  1.333  \n", "22                3  1.333  \n", "23                3  1.333  \n", "24                2  1.500  \n", "25                4  1.500  \n", "26                6  1.000  \n", "27                2  1.000  \n", "28                3  1.667  \n", "29                3  2.000  \n", "30                3  2.000  \n", "31                1  1.000  \n", "32                1  1.000  \n", "33                2  1.000  \n", "34                2  1.000  \n", "35                2  1.500  \n", "36                2  0.500  \n", "37                3  1.333  \n", "38                4  0.750  \n", "39                2  3.000  "], "text/html": ["<div>\n", "<style scoped>\n", "    .dataframe tbody tr th:only-of-type {\n", "        vertical-align: middle;\n", "    }\n", "\n", "    .dataframe tbody tr th {\n", "        vertical-align: top;\n", "    }\n", "\n", "    .dataframe thead th {\n", "        text-align: right;\n", "    }\n", "</style>\n", "<table border=\"1\" class=\"dataframe\">\n", "  <thead>\n", "    <tr style=\"text-align: right;\">\n", "      <th></th>\n", "      <th>id</th>\n", "      <th>lang</th>\n", "      <th>name</th>\n", "      <th>stripped</th>\n", "      <th>tokens_marked</th>\n", "      <th>tokens_stripped</th>\n", "      <th>ratio</th>\n", "    </tr>\n", "  </thead>\n", "  <tbody>\n", "    <tr>\n", "      <th>0</th>\n", "      <td>yo01</td>\n", "      <td>yoruba</td>\n", "      <td>Olúwadámilárè</td>\n", "      <td>Oluwadamilare</td>\n", "      <td>7</td>\n", "      <td>5</td>\n", "      <td>1.400</td>\n", "    </tr>\n", "    <tr>\n", "      <th>1</th>\n", "      <td>yo02</td>\n", "      <td>yoruba</td>\n", "      <td>Adéṣọlá</td>\n", "      <td>Adesola</td>\n", "      <td>7</td>\n", "      <td>3</td>\n", "      <td>2.333</td>\n", "    </tr>\n", "    <tr>\n", "      <th>2</th>\n", "      <td>yo03</td>\n", "      <td>yoruba</td>\n", "      <td>Ọláyínká</td>\n", "      <td>Olayinka</td>\n", "      <td>6</td>\n", "      <td>3</td>\n", "      <td>2.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>3</th>\n", "      <td>yo04</td>\n", "      <td>yoruba</td>\n", "      <td>Fọláṣadé</td>\n", "      <td>Folasade</td>\n", "      <td>8</td>\n", "      <td>3</td>\n", "      <td>2.667</td>\n", "    </tr>\n", "    <tr>\n", "      <th>4</th>\n", "      <td>yo05</td>\n", "      <td>yoruba</td>\n", "      <td>Olúwaṣeun</td>\n", "      <td>Oluwaseun</td>\n", "      <td>8</td>\n", "      <td>5</td>\n", "      <td>1.600</td>\n", "    </tr>\n", "    <tr>\n", "      <th>5</th>\n", "      <td>yo06</td>\n", "      <td>yoruba</td>\n", "      <td>Ayọ̀délé</td>\n", "      <td>Ayodele</td>\n", "      <td>5</td>\n", "      <td>3</td>\n", "      <td>1.667</td>\n", "    </tr>\n", "    <tr>\n", "      <th>6</th>\n", "      <td>yo07</td>\n", "      <td>yoruba</td>\n", "      <td>Ìyábọ̀dé</td>\n", "      <td>Iyabode</td>\n", "      <td>7</td>\n", "      <td>4</td>\n", "      <td>1.750</td>\n", "    </tr>\n", "    <tr>\n", "      <th>7</th>\n", "      <td>yo08</td>\n", "      <td>yoruba</td>\n", "      <td>Bọ́lájí</td>\n", "      <td>Bolaji</td>\n", "      <td>6</td>\n", "      <td>2</td>\n", "      <td>3.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>8</th>\n", "      <td>yo09</td>\n", "      <td>yoruba</td>\n", "      <td>Fúnmiláyọ̀</td>\n", "      <td>Funmilayo</td>\n", "      <td>7</td>\n", "      <td>3</td>\n", "      <td>2.333</td>\n", "    </tr>\n", "    <tr>\n", "      <th>9</th>\n", "      <td>yo10</td>\n", "      <td>yoruba</td>\n", "      <td>Adébáyọ̀</td>\n", "      <td>Adebayo</td>\n", "      <td>5</td>\n", "      <td>3</td>\n", "      <td>1.667</td>\n", "    </tr>\n", "    <tr>\n", "      <th>10</th>\n", "      <td>yo11</td>\n", "      <td>yoruba</td>\n", "      <td>Ọlámidé</td>\n", "      <td>Olamide</td>\n", "      <td>5</td>\n", "      <td>2</td>\n", "      <td>2.500</td>\n", "    </tr>\n", "    <tr>\n", "      <th>11</th>\n", "      <td>yo12</td>\n", "      <td>yoruba</td>\n", "      <td>Ìfẹ́olúwa</td>\n", "      <td>Ifeoluwa</td>\n", "      <td>8</td>\n", "      <td>4</td>\n", "      <td>2.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>12</th>\n", "      <td>yo13</td>\n", "      <td>yoruba</td>\n", "      <td>Táíwò</td>\n", "      <td>Taiwo</td>\n", "      <td>4</td>\n", "      <td>2</td>\n", "      <td>2.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>13</th>\n", "      <td>yo14</td>\n", "      <td>yoruba</td>\n", "      <td>Kẹ́hìndé</td>\n", "      <td>Kehinde</td>\n", "      <td>7</td>\n", "      <td>2</td>\n", "      <td>3.500</td>\n", "    </tr>\n", "    <tr>\n", "      <th>14</th>\n", "      <td>yo15</td>\n", "      <td>yoruba</td>\n", "      <td>Ìdòwú</td>\n", "      <td>Idowu</td>\n", "      <td>6</td>\n", "      <td>3</td>\n", "      <td>2.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>15</th>\n", "      <td>yo16</td>\n", "      <td>yoruba</td>\n", "      <td>Babátúndé</td>\n", "      <td>Babatunde</td>\n", "      <td>5</td>\n", "      <td>3</td>\n", "      <td>1.667</td>\n", "    </tr>\n", "    <tr>\n", "      <th>16</th>\n", "      <td>yo17</td>\n", "      <td>yoruba</td>\n", "      <td>Olúṣẹ́gun</td>\n", "      <td>Olusegun</td>\n", "      <td>8</td>\n", "      <td>3</td>\n", "      <td>2.667</td>\n", "    </tr>\n", "    <tr>\n", "      <th>17</th>\n", "      <td>yo18</td>\n", "      <td>yoruba</td>\n", "      <td>Àbẹ̀ní</td>\n", "      <td>Abeni</td>\n", "      <td>5</td>\n", "      <td>2</td>\n", "      <td>2.500</td>\n", "    </tr>\n", "    <tr>\n", "      <th>18</th>\n", "      <td>yo19</td>\n", "      <td>yoruba</td>\n", "      <td>Yéwándé</td>\n", "      <td>Yewande</td>\n", "      <td>5</td>\n", "      <td>3</td>\n", "      <td>1.667</td>\n", "    </tr>\n", "    <tr>\n", "      <th>19</th>\n", "      <td>yo20</td>\n", "      <td>yoruba</td>\n", "      <td>Adéwálé</td>\n", "      <td>Adewale</td>\n", "      <td>5</td>\n", "      <td>3</td>\n", "      <td>1.667</td>\n", "    </tr>\n", "    <tr>\n", "      <th>20</th>\n", "      <td>ig01</td>\n", "      <td>igbo</td>\n", "      <td>Chiọma</td>\n", "      <td>Chioma</td>\n", "      <td>3</td>\n", "      <td>2</td>\n", "      <td>1.500</td>\n", "    </tr>\n", "    <tr>\n", "      <th>21</th>\n", "      <td>ig02</td>\n", "      <td>igbo</td>\n", "      <td>Ụchenna</td>\n", "      <td>Uchenna</td>\n", "      <td>4</td>\n", "      <td>3</td>\n", "      <td>1.333</td>\n", "    </tr>\n", "    <tr>\n", "      <th>22</th>\n", "      <td>ig03</td>\n", "      <td>igbo</td>\n", "      <td>Obiọma</td>\n", "      <td>Obioma</td>\n", "      <td>4</td>\n", "      <td>3</td>\n", "      <td>1.333</td>\n", "    </tr>\n", "    <tr>\n", "      <th>23</th>\n", "      <td>ig04</td>\n", "      <td>igbo</td>\n", "      <td>Ifeọma</td>\n", "      <td>Ifeoma</td>\n", "      <td>4</td>\n", "      <td>3</td>\n", "      <td>1.333</td>\n", "    </tr>\n", "    <tr>\n", "      <th>24</th>\n", "      <td>ig05</td>\n", "      <td>igbo</td>\n", "      <td>Chinọnso</td>\n", "      <td>Chinonso</td>\n", "      <td>3</td>\n", "      <td>2</td>\n", "      <td>1.500</td>\n", "    </tr>\n", "    <tr>\n", "      <th>25</th>\n", "      <td>ig06</td>\n", "      <td>igbo</td>\n", "      <td>Ụzọamaka</td>\n", "      <td>Uzoamaka</td>\n", "      <td>6</td>\n", "      <td>4</td>\n", "      <td>1.500</td>\n", "    </tr>\n", "    <tr>\n", "      <th>26</th>\n", "      <td>ig07</td>\n", "      <td>igbo</td>\n", "      <td>Ifeanyịchukwu</td>\n", "      <td>Ifeanyichukwu</td>\n", "      <td>6</td>\n", "      <td>6</td>\n", "      <td>1.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>27</th>\n", "      <td>ig08</td>\n", "      <td>igbo</td>\n", "      <td>Ngọzi</td>\n", "      <td>Ngozi</td>\n", "      <td>2</td>\n", "      <td>2</td>\n", "      <td>1.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>28</th>\n", "      <td>ha01</td>\n", "      <td>hausa</td>\n", "      <td>Ɗanjuma</td>\n", "      <td>Danjuma</td>\n", "      <td>5</td>\n", "      <td>3</td>\n", "      <td>1.667</td>\n", "    </tr>\n", "    <tr>\n", "      <th>29</th>\n", "      <td>ha02</td>\n", "      <td>hausa</td>\n", "      <td>Ɗanladi</td>\n", "      <td>Danladi</td>\n", "      <td>6</td>\n", "      <td>3</td>\n", "      <td>2.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>30</th>\n", "      <td>ha03</td>\n", "      <td>hausa</td>\n", "      <td>Ɗanazumi</td>\n", "      <td>Danazumi</td>\n", "      <td>6</td>\n", "      <td>3</td>\n", "      <td>2.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>31</th>\n", "      <td>ct01</td>\n", "      <td>control</td>\n", "      <td>José</td>\n", "      <td>Jose</td>\n", "      <td>1</td>\n", "      <td>1</td>\n", "      <td>1.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>32</th>\n", "      <td>ct02</td>\n", "      <td>control</td>\n", "      <td>François</td>\n", "      <td>Francois</td>\n", "      <td>1</td>\n", "      <td>1</td>\n", "      <td>1.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>33</th>\n", "      <td>ct03</td>\n", "      <td>control</td>\n", "      <td>Søren</td>\n", "      <td>Soren</td>\n", "      <td>2</td>\n", "      <td>2</td>\n", "      <td>1.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>34</th>\n", "      <td>ct04</td>\n", "      <td>control</td>\n", "      <td>Björn</td>\n", "      <td>Bjorn</td>\n", "      <td>2</td>\n", "      <td>2</td>\n", "      <td>1.000</td>\n", "    </tr>\n", "    <tr>\n", "      <th>35</th>\n", "      <td>ct05</td>\n", "      <td>control</td>\n", "      <td>Łukasz</td>\n", "      <td>Lukasz</td>\n", "      <td>3</td>\n", "      <td>2</td>\n", "      <td>1.500</td>\n", "    </tr>\n", "    <tr>\n", "      <th>36</th>\n", "      <td>ct06</td>\n", "      <td>control</td>\n", "      <td>Jürgen</td>\n", "      <td>Jurgen</td>\n", "      <td>1</td>\n", "      <td>2</td>\n", "      <td>0.500</td>\n", "    </tr>\n", "    <tr>\n", "      <th>37</th>\n", "      <td>ct07</td>\n", "      <td>control</td>\n", "      <td>Dvořák</td>\n", "      <td>Dvorak</td>\n", "      <td>4</td>\n", "      <td>3</td>\n", "      <td>1.333</td>\n", "    </tr>\n", "    <tr>\n", "      <th>38</th>\n", "      <td>ct08</td>\n", "      <td>control</td>\n", "      <td>Nguyễn Thị Hương</td>\n", "      <td>Nguyen Thi Huong</td>\n", "      <td>3</td>\n", "      <td>4</td>\n", "      <td>0.750</td>\n", "    </tr>\n", "    <tr>\n", "      <th>39</th>\n", "      <td>ct09</td>\n", "      <td>control</td>\n", "      <td>Zhāng Wěi</td>\n", "      <td>Zhang Wei</td>\n", "      <td>6</td>\n", "      <td>2</td>\n", "      <td>3.000</td>\n", "    </tr>\n", "  </tbody>\n", "</table>\n", "</div>"]}, "metadata": {}, "execution_count": 12}], "source": ["tok_rows = []\n", "for r in names:\n", "    if r[\"n_marks\"] == 0:\n", "        continue\n", "    a = len(tok(\" \" + r[\"name\"], add_special_tokens=False).input_ids)\n", "    b = len(tok(\" \" + r[\"stripped\"], add_special_tokens=False).input_ids)\n", "    tok_rows.append(dict(id=r[\"id\"], lang=r[\"lang\"], name=r[\"name\"], stripped=r[\"stripped\"],\n", "                         tokens_marked=a, tokens_stripped=b, ratio=round(a / b, 3)))\n", "tok_df = pd.DataFrame(tok_rows)\n", "tok_df.to_csv(\"results/tokenizer.csv\", index=False)\n", "tok_summary = tok_df.groupby(\"lang\")[\"ratio\"].mean().round(2).to_dict()\n", "print(\"mean tokens(marked)/tokens(stripped):\", tok_summary)\n", "tok_df"]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/"}, "id": "-2P5UnwyYy4z", "executionInfo": {"status": "ok", "timestamp": 1792105860661, "user_tz": -60, "elapsed": 15460, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "8420b0fc-1ed0-42fa-d455-4d776e17f2e2"}, "execution_count": 13, "outputs": [{"output_type": "stream", "name": "stdout", "text": ["share of (name, context) pairs where the model prefers the STRIPPED form: {'control': 0.5, 'hausa': 1.0, 'igbo': 1.0, 'yoruba': 1.0}\n"]}], "source": ["@torch.no_grad()\n", "def cont_logprob(prefix, cont):\n", "    p_ids = tok(prefix, add_special_tokens=False).input_ids\n", "    f_ids = tok(prefix + cont, add_special_tokens=False).input_ids\n", "    n = 0\n", "    while n < min(len(p_ids), len(f_ids)) and p_ids[n] == f_ids[n]:\n", "        n += 1\n", "    x = torch.tensor([f_ids], device=DEVICE)\n", "    logits = model(input_ids=x).logits[0].float()\n", "    lp = torch.log_softmax(logits[:-1], -1)\n", "    tl = lp.gather(1, x[0, 1:, None])[:, 0]\n", "    return tl[max(n, 1) - 1:].sum().item()\n", "\n", "lp_rows = []\n", "for r in names:\n", "    if r[\"n_marks\"] == 0:\n", "        continue\n", "    for ctx in (\"The name is\", f\"My {r['sublang']} name is\"):\n", "        m = cont_logprob(ctx, \" \" + r[\"name\"] + \".\")\n", "        s = cont_logprob(ctx, \" \" + r[\"stripped\"] + \".\")\n", "        lp_rows.append(dict(id=r[\"id\"], lang=r[\"lang\"], name=r[\"name\"], context=ctx,\n", "                            logp_marked=round(m, 3), logp_stripped=round(s, 3),\n", "                            prefers_stripped=s > m, delta_nats=round(m - s, 3)))\n", "lp_df = pd.DataFrame(lp_rows)\n", "lp_df.to_csv(\"results/logprob.csv\", index=False)\n", "lp_summary = lp_df.groupby(\"lang\")[\"prefers_stripped\"].mean().round(3).to_dict()\n", "print(\"share of (name, context) pairs where the model prefers the STRIPPED form:\", lp_summary)"]}, {"cell_type": "markdown", "metadata": {"id": "zOlJqHSiYy40"}, "source": ["## 6. Feature analysis: what predicts identity loss?"]}, {"cell_type": "code", "metadata": {"colab": {"base_uri": "https://localhost:8080/"}, "id": "die_ScWjYy40", "executionInfo": {"status": "ok", "timestamp": 1792105860675, "user_tz": -60, "elapsed": 9, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "a92dbd7b-4f2c-473b-cf56-a614ebe75636"}, "execution_count": 14, "outputs": [{"output_type": "stream", "name": "stdout", "text": ["African names: copy-exact vs diacritic density rho=0.09 (p=0.627); vs length rho=0.15 (p=0.431)\n"]}], "source": ["from scipy.stats import spearmanr\n", "per_name = (copy_df[copy_df.n_marks > 0]\n", "            .assign(exact=lambda d: d.status == \"exact\")\n", "            .groupby([\"item\", \"lang\", \"name\", \"density\", \"n_letters\", \"n_marks\", \"has_hooked\"], as_index=False)[\"exact\"].mean())\n", "per_name.to_csv(\"results/per_name_copy.csv\", index=False)\n", "african = per_name[per_name.lang != \"control\"]\n", "rho_d, p_d = spearmanr(african.density, african.exact)\n", "rho_l, p_l = spearmanr(african.n_letters, african.exact)\n", "print(f\"African names: copy-exact vs diacritic density rho={rho_d:.2f} (p={p_d:.3f}); vs length rho={rho_l:.2f} (p={p_l:.3f})\")"]}, {"cell_type": "markdown", "metadata": {"id": "yDgzh0jPYy40"}, "source": ["## 7. Figures"]}, {"cell_type": "code", "source": ["import matplotlib.pyplot as plt\n", "os.makedirs(\"figures\", exist_ok=True)\n", "SURF, INK, INK2, GRID = \"#fcfcfb\", \"#0b0b0b\", \"#52514e\", \"#e4e3df\"\n", "SERIES = [\"#2a78d6\", \"#eb6834\", \"#1baf7a\", \"#eda100\", \"#e87ba4\", \"#008300\", \"#6250d6\", \"#e34948\"]\n", "GROUPS = [\"yoruba\", \"igbo\", \"hausa\", \"control\"]\n", "GLABEL = {\"yoruba\": \"Yorùbá\", \"igbo\": \"Igbo\", \"hausa\": \"Hausa\", \"control\": \"High-resource controls\"}\n", "GCOLOR = dict(zip(GROUPS, SERIES))\n", "plt.rcParams.update({\"figure.facecolor\": SURF, \"axes.facecolor\": SURF, \"savefig.facecolor\": SURF,\n", "                     \"axes.edgecolor\": GRID, \"axes.labelcolor\": INK2, \"xtick.color\": INK2, \"ytick.color\": INK2,\n", "                     \"text.color\": INK, \"font.size\": 11, \"axes.spines.top\": False, \"axes.spines.right\": False,\n", "                     \"axes.grid\": True, \"grid.color\": GRID, \"grid.linewidth\": 0.8, \"axes.axisbelow\": True})\n", "\n", "# Fig 1: outcome distribution in the copy task (100% stacked)\n", "OUTCOMES = [\"exact\", \"partially_stripped\", \"fully_stripped\", \"wrong_marks\", \"spurious_marks\", \"catastrophic\", \"altered\", \"missing\"]\n", "OLABEL = {\"exact\": \"Exact\", \"partially_stripped\": \"Some marks lost\", \"fully_stripped\": \"All marks lost\",\n", "          \"wrong_marks\": \"Wrong marks\", \"spurious_marks\": \"Extra marks\", \"catastrophic\": \"Letter dropped\",\n", "          \"altered\": \"Respelled\", \"missing\": \"Name missing\"}\n", "OCOLOR = {\"exact\": \"#2a78d6\", \"partially_stripped\": \"#eb6834\", \"fully_stripped\": \"#e34948\", \"wrong_marks\": \"#e87ba4\",\n", "          \"spurious_marks\": \"#6250d6\", \"catastrophic\": \"#eda100\", \"altered\": \"#1baf7a\", \"missing\": \"#008300\"}\n", "cd = copy_df[copy_df.n_marks > 0]\n", "fig, ax = plt.subplots(figsize=(10, 3.8))\n", "for gi, g in enumerate(GROUPS):\n", "    sub = cd[cd.lang == g]; left = 0\n", "    for oi, o in enumerate(OUTCOMES):\n", "        w = (sub.status == o).mean() if len(sub) else 0\n", "        if w > 0:\n", "            ax.barh(gi, w, left=left, color=OCOLOR[o], edgecolor=SURF, linewidth=2, height=0.6,\n", "                    label=OLABEL[o] if gi == 0 or o not in ax.get_legend_handles_labels()[1] else None)\n", "            if w >= 0.08:\n", "                ax.text(left + w / 2, gi, f\"{w:.0%}\", ha=\"center\", va=\"center\", color=\"white\", fontsize=10, weight=\"bold\")\n", "            left += w\n", "ax.set_yticks(range(len(GROUPS)), [GLABEL[g] for g in GROUPS]); ax.invert_yaxis()\n", "ax.set_xlim(0, 1); ax.xaxis.set_major_formatter(lambda v, _: f\"{v:.0%}\"); ax.grid(axis=\"y\", visible=False)\n", "h, l = ax.get_legend_handles_labels(); seen = {}\n", "for hh, ll in zip(h, l): seen.setdefault(ll, hh)\n", "ax.legend(seen.values(), seen.keys(), ncol=4, loc=\"upper center\", bbox_to_anchor=(0.5, -0.12), frameon=False, fontsize=9)\n", "ax.set_title(\"What happens to a name when the model uses it (5 everyday tasks)\", loc=\"left\", fontsize=13, weight=\"bold\")\n", "plt.tight_layout(); plt.savefig(\"figures/fig1_copy_outcomes.png\", dpi=180, bbox_inches=\"tight\"); plt.show()\n", "\n", "# Fig 2: headline metrics per language group (grouped bars)\n", "METRICS = [(\"copy_exact\", \"Name kept exactly\\n(copy tasks)\"), (\"restore_exact\", \"Diacritics restored\\nexactly\"),\n", "           (\"meaning_acc_marked\", \"Meaning correct\\n(marked name)\"), (\"meaning_acc_stripped\", \"Meaning correct\\n(stripped name)\")]\n", "fig, ax = plt.subplots(figsize=(10, 4.2)); W = 0.2\n", "for gi, g in enumerate(GROUPS):\n", "    vals = [summary.get(g, {}).get(k) for k, _ in METRICS]\n", "    xs = [i + (gi - 1.5) * W for i in range(len(METRICS))]\n", "    for x, v in zip(xs, vals):\n", "        if v is None:\n", "            continue\n", "        ax.bar(x, v, W * 0.9, color=GCOLOR[g], label=GLABEL[g] if x == xs[0] or GLABEL[g] not in ax.get_legend_handles_labels()[1] else None)\n", "        ax.text(x, v + 0.015, f\"{v:.0%}\", ha=\"center\", va=\"bottom\", fontsize=8, color=INK2)\n", "ax.set_xticks(range(len(METRICS)), [m for _, m in METRICS]); ax.set_ylim(0, 1.1)\n", "ax.yaxis.set_major_formatter(lambda v, _: f\"{v:.0%}\"); ax.grid(axis=\"x\", visible=False)\n", "h, l = ax.get_legend_handles_labels(); seen = {}\n", "for hh, ll in zip(h, l): seen.setdefault(ll, hh)\n", "ax.legend(seen.values(), seen.keys(), ncol=4, loc=\"upper center\", bbox_to_anchor=(0.5, -0.18), frameon=False)\n", "ax.set_title(\"Identity preservation by language: Qwen3.5-4B\", loc=\"left\", fontsize=13, weight=\"bold\")\n", "plt.tight_layout(); plt.savefig(\"figures/fig2_headline.png\", dpi=180, bbox_inches=\"tight\"); plt.show()\n", "\n", "# Fig 3: copy exact rate by context\n", "TMPL = [(\"c1_repeat\", \"Repeat\"), (\"c2_welcome\", \"Welcome msg\"), (\"c3_json\", \"JSON record\"), (\"c4_translate\", \"Translate\"), (\"c5_upper\", \"UPPERCASE\")]\n", "fig, ax = plt.subplots(figsize=(10, 3.8))\n", "for gi, g in enumerate(GROUPS):\n", "    xs = [i + (gi - 1.5) * W for i in range(len(TMPL))]\n", "    vals = [by_template.get(g, {}).get(t) for t, _ in TMPL]\n", "    ax.bar([x for x, v in zip(xs, vals) if v is not None], [v for v in vals if v is not None], W * 0.9, color=GCOLOR[g], label=GLABEL[g])\n", "ax.set_xticks(range(len(TMPL)), [t for _, t in TMPL]); ax.set_ylim(0, 1.05)\n", "ax.yaxis.set_major_formatter(lambda v, _: f\"{v:.0%}\"); ax.grid(axis=\"x\", visible=False)\n", "ax.legend(ncol=4, loc=\"upper center\", bbox_to_anchor=(0.5, -0.12), frameon=False)\n", "ax.set_title(\"Share of names kept exactly, by context\", loc=\"left\", fontsize=13, weight=\"bold\")\n", "plt.tight_layout(); plt.savefig(\"figures/fig3_copy_by_context.png\", dpi=180, bbox_inches=\"tight\"); plt.show()\n", "\n", "# Fig 4: diacritic density vs copy-exact rate per name\n", "fig, ax = plt.subplots(figsize=(8, 4.5))\n", "for g in GROUPS:\n", "    s = per_name[per_name.lang == g]\n", "    ax.scatter(s.density, s.exact, s=60, color=GCOLOR[g], edgecolor=SURF, linewidth=2, label=GLABEL[g], zorder=3)\n", "ax.set_xlabel(\"Identity marks per letter\"); ax.set_ylabel(\"Copy tasks with name kept exactly\")\n", "ax.yaxis.set_major_formatter(lambda v, _: f\"{v:.0%}\"); ax.set_ylim(-0.05, 1.05)\n", "ax.legend(frameon=False, fontsize=9, loc=\"best\")\n", "ax.set_title(f\"Diacritic density vs. names kept exactly (African names: Spearman ρ = {rho_d:.2f})\", loc=\"left\", fontsize=13, weight=\"bold\")\n", "plt.tight_layout(); plt.savefig(\"figures/fig4_density_vs_preservation.png\", dpi=180, bbox_inches=\"tight\"); plt.show()\n", "\n", "# Fig 5 and 6: tokenizer and likelihood (two separate charts, never a dual axis)\n", "for fname, data, title, fmt, ref in [\n", "    (\"fig5_token_ratio.png\", tok_summary, \"Tokens for the correct name ÷ tokens for the stripped name\", lambda v: f\"{v:.2f}×\", 1.0),\n", "    (\"fig6_prefers_stripped.png\", lp_summary, \"How often the model finds the stripped name MORE likely\", lambda v: f\"{v:.0%}\", None)]:\n", "    fig, ax = plt.subplots(figsize=(8, 3))\n", "    gs = [g for g in GROUPS if g in data]\n", "    ax.barh(range(len(gs)), [data[g] for g in gs], color=[GCOLOR[g] for g in gs], height=0.55)\n", "    for i, g in enumerate(gs):\n", "        ax.text(data[g], i, \"  \" + fmt(data[g]), va=\"center\", color=INK2, fontsize=10)\n", "    if ref:\n", "        ax.axvline(ref, color=INK2, linewidth=1, linestyle=\"--\")\n", "    ax.set_yticks(range(len(gs)), [GLABEL[g] for g in gs]); ax.invert_yaxis(); ax.grid(axis=\"y\", visible=False)\n", "    if fmt(0.5).endswith(\"%\"):\n", "        ax.set_xlim(0, 1.15); ax.xaxis.set_major_formatter(lambda v, _: f\"{v:.0%}\")\n", "    else:\n", "        ax.set_xlim(0, max(data.values()) * 1.25)\n", "    ax.set_title(title, loc=\"left\", fontsize=12, weight=\"bold\")\n", "    plt.tight_layout(); plt.savefig(\"figures/\" + fname, dpi=180, bbox_inches=\"tight\"); plt.show()"], "metadata": {"colab": {"base_uri": "https://localhost:8080/", "height": 1000}, "id": "tTlB_4sxt39-", "executionInfo": {"status": "ok", "timestamp": 1792105863135, "user_tz": -60, "elapsed": 2459, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "095a1ea1-70b6-43e8-baef-1c689d967f9d"}, "execution_count": 15, "outputs": [{"output_type": "display_data", "data": {"text/plain": ["<Figure size 1000x380 with 1 Axes>"], "image/png": "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\n"}, "metadata": {}}, {"output_type": "display_data", "data": {"text/plain": ["<Figure size 1000x420 with 1 Axes>"], "image/png": "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\n"}, "metadata": {}}, {"output_type": "display_data", "data": {"text/plain": ["<Figure size 1000x380 with 1 Axes>"], "image/png": "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\n"}, "metadata": {}}, {"output_type": "display_data", "data": {"text/plain": ["<Figure size 800x450 with 1 Axes>"], "image/png": "iVBORw0KGgoAAAANSUhEUgAAA24AAAG1CAYAAACBN6eiAAAAOnRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjExLjIsIGh0dHBzOi8vbWF0cGxvdGxpYi5vcmcvgI3uAAAAAAlwSFlzAAAPYQAAD2EBqD+naQAAm2pJREFUeJzs3XV4U2cbBvC7qbtSL9RosRYp7hR3GDKGbYMPNibIBsOHb8OZ4RuDDcaAIRsMGwx3twoVpO5uSZPvj9JDU6NNkzZt79917Vpz9HmfnBzy5LznPRoSiVgGIiIiIiIiUluiqg6AiIiIiIiISsfCjYiIiIiISM2xcCMiIiIiIlJzLNyIiIiIiIjUHAs3IiIiIiIiNcfCjYiIiIiISM2xcCMiIiIiIlJzLNyIiIiIiIjUHAs3IiIiIiIiNcfCTQ25uLjC1NQMpqZmuHnzZlWHI6hIXFXRpqtXrwr7rFu3bqXsU1Hq+p6TeuNxU3389NPPwns1aNBgle0nPj4eTk51YWpqho8//lgp20xPT8eSJUvRvHkL1KljLbQjKSmpTOtXVtuJiNTZpUuXhHPhjh2/KLQNFm4q0r17D+HNMTU1g6WlFRwcHOHl5Y3Bg4dgw4ZvER8fX9VhKsWkSZOFdi5btryqw6lRmFuqTp4/fy533svKyqrqkCqFOrX7m2++QUpKCkQiEWbMmFHsMoGBgXLxmpqa4dKlSyVuc+HCL7Fu3TqEhIQgJydHVaFTDSYWi/HDDz+ga9ducHBwhJVVHbi4uKJly1Z45513sGHDt1UdIlVD//13DiNGjISLiyvq1LGGl5cXpk+fgbCwsErZnlgsxo8//ojOnbvA3t4BtrZ2aN++PVavXo2MjIwiy3fs2BFt27YFAHz99ddIS0srd4ws3CqJRCJBWloaXrx4gXPnzmHRokVo1qw5jhw5UtWhERFRDRATEyP8iuvr6wt3d/dil9u9e0+RaXv2FJ2W7++//xb+3r59G5KSEpGcnAQzM7MKxUu1Q25uLt56axjmz1+Au3fvIi0tDWKxGAkJCXj69Cn++ec4fvnll6oOk6qZjRs3YsiQITh16hQSEhKQk5ODFy9eYseOHejUqTMePnyo0u1lZ2dj6NChmDdvPu7fv4/09HRkZmbi8eMnWL58Bfr27YeUlJQi+5k0aRIAIDo6Gjt27Ch3u1m4VYL3338fyclJePnyBQ4fPox27doBAFJSUvDee+/j5MmTcsuHhoYgOTkJyclJaNWqVVWEXKyKxKWubVIXzA8RVdTOnbsgFosBAMOHDyt2GalUin379hWZfuTIX0hPTy8yPTc3FzExMcLr1q1bQ0NDo1xxTZw4QTi//fUXf6ysbQ4fPoILFy4AABo1aoSLFy8gJiYaT58G4vTpU5g5cyacnZ2rNkiqVh48eIAFCxYCAHR1dXHkyBFERkbg888/BwAkJCRg0qTJkEqlKtved999h4sX83oqeHp64s6dOwgI8BeuqN27dw+LFy8psq9+/frC0NAQAPDLLzshk8nK1XYWbpXIxMQE3bp1xdGjf6NDh/YA8v4R/fzzz+W6n5R030qXLl2LdL90d6+PESNG4vTp08XuMygoCNOnz0CLFj6wtbWDq6sbhg0bjnPnzsstV3ifO3fuQsuWrWBhYYndu3cXu8ydO3dhamom9yVgzZo1wjL9+/d/Y5vKE2Nptm//CS1btoK1tQ1at26D/fv3v3Gdx48fY8qUKfDy8kKdOtaoW7cu+vfvj4MHDxVZtnD8f//9Nzp06ABbWzu0bt0G27ZtK/Lhu3PnLsaMGQtPzwbCe9WnT1+sXLkSL1++LHX7ZcltSkoKHBwcYWpqBnNzC7x48UJum3/+eVBYvkePniXmISMjA3Xr1hWWDQwMlJv/zz//CPPatGmrUPsUoUjOy/sZKdjumzdvYtu27WjSpAkcHBwxcuTbeP78OQBg7969aN26Dezs7NG+fXvs3bu32JjLc0ypKn9SqVSui23z5i2E7UkkEmzbtg29evVG3bp1X3UF8cbnn89ERESE3HYK35cUFxeHiRP/h7p168HR0QljxoxFaOgzYfnRo0fD27up3DZsbGyFbfz337k3xl7W+J4+fQp7eweYmprBwsISt2/fFuZt3LhR7njNzMwEAPz7779Fugfa2dmjTZu2WLBgIRISEovN5W+//YaBAwfBxcVVOL8sXbpMWL4i7f7mm2+E5T788MMi89u1ay/ML/wDX3EOHjwo/N2rV69ilzl79j8hl25ubsKXjLS0NPz1119yy86cOQsWFpZy07y9m8LU1Axt2+b9AFn4OImMjMT48e+ibt26aNSocbHLFFSWHAPlf/8K7zM5ORmffPIp6tVzhqurG0aNGoWQkNA35lTRbZU33q1bt8rtIzw8HCNHvg17ewd4eXlh06ZNAIDY2NhXn8O6qF/fA++99z4iIyOLxFyez7pYLMbatevQsWMnODg4wsbGFq1atcbEif/D33//DYlEIrd8/foe5fpcX7t2Vfh76NAh8Pb2hq6uLqytrdG6dWssXLgAhw/LnyPLe/5RpN0VPaaKO9aV+T5W5TGfLzs7Gz/88AO6dfOFk1NdWFhYonHjJpg1axbi4uLKtS1l2rJlK3JzcwEAQ4YMRteuXWBgYIB58+bC1NQUAODn54ezZ8+qbHv79x8Q/v7oo4/g5uYKW1tbfP75Z8L03bt3Izk5WW5fBgYG6NixI4C877/37t0vX+MlErGM/yn/vy5dusgMDQ1lhoaGsk8++aTI/PPnzwnzDQ0NZSdOnBDmOTk5CdOvXr0iTO/QoYPcOoX/27Ztm9w+Dh06KLO0tCxx+bS01GL3OXHiBLnlduz4udi4bty4UWo8vXv3emObyhNjSf8tXbqk2HU/+OAD4W97ezu5dfbt2yczNzcvcb+ffTZDbvmC8U+dOrXYdXbu/EVYPiAgoNR2de/evcTtlye3n376qTBtyZLFctscPXq0MG/Lli2l5rDgdhYvXiQ3b9y4ccK8DRvWK9Q+Rf4rb84V+YzY29sJ895///0iyzdu3Fi2YcP6Yrd19OjfCh9Tyspf4eMmJydbNmnS/4RpLVu2lIWHh8kkErEsJSVZ1qtXzxL3Wa9ePdnDhw+EbW/ZskWY17VrV5mPj0+RdVxcXGTPnz+TSSRi2ciRI0rN/enTp0ttS3nj27HjZ2Fe8+bNZWlpqTI/vycyKysrmaGhoczS0lL24MF9YfkTJ46XGp+3t7csISFeWD4tLVXWv3//EpefPXt2udpdMJ/9+vWTSSRiWXh4mMzMzExmaGgos7KyksXHxwn79/N7IpfnrKzMUvMXExMtMzIykhkaGsqaNGlS4nLjx48Xtrt8+TK5uPr06SO37PTp00tsV8uWLYu0q0uXLjJvb2/htZubW4ltL0+OFXn/Cu6zW7dusnbt2hX7+c7ISH/j50yRbZU33o0bf5TLo5eXV5F1Vq9eLWvWrFmR6e3atZPl5GQr/Fn65JNPSo31xInjcvlwcXEp8+daIhHLvvhilrB8gwaesv3798vi4mLLnPOynH8UaXdFjqmSjnVlvo9VecxLJGJZVFSk3HfZwv95eHjInj0LfeN2srIyS21H4f/+++/sG7fZqFEjYfnvvvtWbl7BY2DevHllaqsi26tTp44w/Z9/jgnT7969I9eekydPFtnfV1+tEOavW7euTDHm/8crblWkdevW0NTUFF7fu3fvjeucP39O6G6SlJSI0NBQrFu3Tpi/bNkyoZtMVFQUJk/+QLhJvlevXrh+/RrCw8Nw5sy/GD58eIndXY4ePYaffvoJYWEvkZychHHjxhW7XIsWzZGcnISRI0cK02bOnCnEeOzYsVLbU5EY8z179gxr1qwVXi9YMB/Pnz/DyZMncOrUqWLXiYyMxEcffSRc5VyzZjXCwl7i8uVLQneNLVu24vjx48Wuf+jQIfz9918IDQ3FiBEjhOk//fSz8Pfx4/8I7ZoxYwbCw8Pw7NkznD17BjNnzoSDg0Op7SprbidPniTM3717j3AZPyMjQ7jCpKuri2HD3ip1fwXf4337Xl+tTEtLw4kTJwAA2traGDVqlFLaV15lyTlQvs9IYdeuXcWVK5fx5Mlj2NvbA8g7vubPX4Dvv/8OL1++wNtvv34/Cu67vMeUKvInk8nw6aefYs+e3wEATZs2xT//HIONjQ0AYMWKFbh8+Yow79q1qwgLe4n58+cBAOLi4jBp0uRiu23cvHkT7du3R0hIMK5duwpPT08AefdU5XcF2bNnDx48kP/lMDo6Sng/unXrWmr85Y1v3LhxeOutvOM6MDAQX375JT744EPhCttXX32FRo0aCdvv0aOHEEtychJiY2Nw/fo1tGnTBgAQHByMX37ZWSCer3Du3DkAgJGREX788QeEhobiyZPH2LjxR9jb21W43TY2Nhg0aBAAIDMzU+6zd/jwYeHvkSNHQktLq9T8PXnyRMiNm5tbscskJyfLnZdHjBiJoUOHQltbG0DeiGcFr9yvWbMa8fHyv6o/eHAfyclJcldR8t26dQtubm64c+c2kpISERDgX2rMZc0xUP73r6AbN26gSZPGCA4OwokTx2FkZAQg7/N95syZUmNUdFsViffWrVsYMmQIXrx4jkWLFgnTFy9eDH19fTx8+ACXL1+Cjo4OgLxuXrdu3SqQ1/J9lg4dyrvaZWVlhevXryE2NgZ3797Fzp070bt3b2hqln7svUmrVq2Fv1++DMN7770HZ2cXtG7dBtOmTX/jVbuynH8UaXdF36M3HesVfR+r+pifMuUjIZ4JEyYUyX94eDjmzp1Xpm0pk1QqlTtP1alTR26+paWV8Hdo6JuvMCq6vfycAkBExOurpVFRUXLrF+4NBQBubq/vP378+NEbYyyIhVsV0dbWhpGRofC68KXUN9HQ0ICFhTkmTpwAO7u8f9zi4+Ph5+cHANi/f78woo2NjQ1++WUHGjRoACMjI7Rs2RI//bQdurq6xW77k08+xvDhw2BsbKxI08qsIjHm+/vvv4XL225ubpg1axbMzMzQtm1bTJkypYT9HhC+3Pn6+mLSpEkwNjZGkyZN8MknnwjL/frrr8Wu/8knn6Bz586wsDDHu++OF6aHhIQIf+vp6Qt/x8REIyQkBLq6OvDx8cHChQvw888/ldqusmrYsCE6dcq75B4eHo5///0XAHDy5Ckht/369XvjIAItWjRHkyZ5XT2eP3+OGzduAACOHj0qbKdv376wsrKq1PblK0vOC3vTZ6S4fTRu3BgODg5o3fr1Fw0fHx+MHz8eJiYmQqFQeN/lPaZUkb958+bjt9/yujW3adMGf//9FywsLADk/cNUcECKb775Gg0bNoSxsTG++OIL2NraAgAePnyIu3fvFdm2qakJvvpqBSwtLdGwYUPMm/f6H+ujR48W6UpVXorGt2HDeuFRH1u2bBW6TA4cOBATJ04odZ86Ojpo0KABxo9//aNF/siKUqlU7vM/a9YsjB07FhYW5nBwcMCYMWOUNtR+wR9fdu3aJfxdsHAbPfqdN24nNvZ1gZX/vhd28OAh4QeDli1bws3NFRYW5ujZsweAvOK/pG7AZWFkZIRt27bCzc3tjT+6VTTHpb1/hZmYmGDNmjWwsrJCu3bt4OPjI8wr7RyizG2VJ14LCwuhi9agQQPl5s2bNxd169ZFkyZN4O3tXWTfinyW9PXzzkdicQ6Cg4MRGxsLFxdnDBkyGPv2/VHkx4enTwPL/IMMkNf1rG/fPnLTZDIZAgIC8Msvv2DIkCEYMWKk8G95YWU5/1T0HAeU7z0qy7FekfexovFV9JgPCQkVfvwuLf+nTp1640izWlpacgXom/7r1KlTqdvLyMiQ+4FRS0tbbn5+IQyg2Pt2lbW9Hj26C39v3rwZISGhiIqKkvuxGACys4uOMmxubi78XfDcXRYs3KqIWCxGWtrrAyC/D21pbt++gwkTJsLLyxvW1jZCX+aC/aJjYmIBAP7+AcK0Vq1aCTdClkXr1m3KvGxFVCTGfAX7axf8dR0AvLyaFLtOwV/Gzp49K9d/fObMmcI8P7/ify1u3ry58HfB4rbg0K99+vQW3tPdu/egU6fOsLd3gLd3U3z66dQi95FVxP/+9/qL386deb++FfzSN2bM6DJtp7irbgX7cBf8x6Iy2weULedA+T4jhXl6NhD+LvhLWsHjqmDBVXDf5T2mVJG//HtHbWxscPDgn3LnlOjoaCQmvr4fom/ffnIxFvyF0N+/aGHr7l5f+HIHyH+2MjIyir3HpjwUjc/U1BQ//bQdItHrf8rs7Ozw/fffF9lHdnY2Nmz4Fj169ES9es6wsLCEqakZPv74dWGdPwhHTEyMXDxdu3apUPtK07ZtW3h5eQHI+8X97t17CAoKwqNHjwHkXTlo3LjxG7dT8ItHSTe7Fxw5cuTI11euR4wYWWCZ38vXgAI8PDzKPNJkeXNcnvevsPr168v9+1LaOeRNyrqtisTr5uYmXAUteC4C5M9H+vp6wt/p6Xn7VuSz9PbbbwMAkpNTMHr0GDRq1Bj29g7o27cvtm//qcSCqqw0NDSwZ88ebN26BV27dpU7l+Q7deoUfv7552LWLtv5R5F2V+Q9KsuxXpH3saLxVfSYL9gLrLT8Z2VlITo6+o3bUyYDAwO5Ylksli8cCxaSZfleqej25s+fL/wg4Ofnh+bNm8PTswGuXJHvjVDccVKW83VJWLhVkevXr8udDAt+MS3OrVu30Lt3b/z555948eIFsrOzi11OIsnvBla+A6EgExPVXml7TfEYhS0UOOAL/+pV3g9DYSX9UmNs/PoEXPALY0GOjo44deokxo8fD1dXV4hEIshkMjx//hy7du1C3779lHZj74AB/YWudSdOnERo6DOhm6StrS18fX3LtJ2RI98WrnAePHgQkZGRQjcmBwcHdO/++telymwfULacl/8zIs/Q0ED4u+CxVPDKuLKOKVXkL7/rdXR0NL7++usKx1iQsj9bFVE4vuDgYLmRvhISEhAeXvSZOxMmTMSiRYtw8+ZNJCUlFftlNP/YqOz25Q8PDeRddTt06LDwuixX2wCgTp3X3XkKfoHNFxQUJFxJB4AvvpgtfKl9//33hemhoaG4cuVKecIXlKeXRnlzXJ73r7DCX5pLOoeURVm3VZF4DQyKPxcV3r+yzkdffrkQ69evR6dOHYUffDIyMnDlylV8/vnn+Oorxc8n+UQiEd5++20cOXIYL148x8mTJzBjxgzo6b0uWq5du17suqr6t70i71FZjvWKvo9VecwX7AX2pvy/6f2QSCRFBlkp7b+LFy+Wuj2RSCT0tADyBnopKC7u9WsXF5dSt1WR7Tk6OuK//85i3LhxsLW1hY6ODurWrYuPP/5IrrtlcT+8FTxHFzx3lwULtyogkUjw1VdfCa+dnByF7m4l2bPnd+HenLZt2+LhwwdITEwo8Vk6DRq8vnpw8+bNcv+qWB4FP9PlOaEqI0ZXV1fh78ePH8vNy//FujAPD0/h7379+pZ4uT4wMKDY9cuqQYMG+P7773D37h3Exsbg6tUraNGiBYC8vvYXLpR+cgLKllstLS28//57APKOrUmTJgn/MI0cOVLuXsrSWFiYY8CAAQDyuhROnTpN6AI3ZszoIid+ZbRPmcr7GVEmRY4pZefvq69WCP9Y//DDj3IPbLexsZHLwdmzZ0uMsWARke/p06dyD5Uu+NkyMDAQuqIq+gVL0fiCg4Mxc+YsABDuAcvOzsaECRPlzicJCQk4evSo8HrTpo2IjIxAcnIS1qxZXWw8BbuynD9/odT4K/rFcuTIEUL7Dxw4IIyKq62tLXdPZ2kaNmwoxBEcHFxkfnmupFXkqltZlSfH5X3/qlpVxqvIZ0lTUxMTJryPo0eP4sWL5wgNDcG3375+IHb+PXDKoqOjg7Zt22Lx4kXCfdMASuxyXZbzT3nbre7HVFXHV/DHsLKe/ytTwe6UBUdlFIvFct8FO3cuvdtlRbdnb2+PH374HgEB/oiNjcHDhw8wYsRIofhzdXUttnALDg4S/m7cuPjeYSVh4VaJUlNTce7ceQwcOFC4gVYkEmHt2rVyfWiLo639+uZgPT09GBkZIykpCfPmzUNSUlKR5YcPHy782hMdHY0JEyYgICAA6enpuHPnLsaOHVfiFYnyMjF53SUr/yGEZaGMGAcOHCgUFMHBwVi9ejWSk5Nx7do1YdjdovsdJvzKd/z4Caxbtx4xMTHIzMxEUFAQDh8+gvHj38XPP5f/wYj5Dh06jHfffQ8nTpxAaOgzSKXSIr/QlfRLWUFlze27774rHEMFH7dQ1l/r8xXsLpnfv11DQwNjxoyRW6687Zs+fYbwa9qcOXPKFVNZlfczokzlPaaUdXwU5OPjg99++004DtasWYO1a/P62otEIrzzzusvSNOmTcWNGzeQmZmJ2NhY3L59G6tWrULXrt2K3XZycjLmz1+AhIQE+Pn5yf3w1L9/f6FoMjExkVvv2rXrZSpiFIkvJycH778/AWlpaQDy7mkZMmQIACAgIACzZ78+zrS0tOSKKzMzM2hqauLmzZv47rvvio1n7NixwuvVq1djz549SEhIRGRkJH799Vd8993r7piKtjufvr4+xo7N+4ylpKQgICCvwO/duxcsLS1LW1VgZWUl/BgWGhoqN1S4VCrFH3/8Ibw+cuRIkS+zV65cFuYfPnxYpT/2AeXLcXnfv6pWlfEq8ll6++23sXbtOty7dw8JCYkwMDCAnt7r+8sLD+hU3scBrF27DmPGjMFvv/2Ghw8fIiEhAdnZ2bh16xb+++8/YblWrVoWu35Zzj/lbbe6H1PqFF9p+e/Xr5/QHbQkyr7HDQA++GCy8KP0X3/9hfPnzyMzMxNff/0NkpPzHnrdsGFDuR5H8fHxclf24uPjK7S9ly9fYtSoUTh37jwSE5MQHx+PI0eOYNy41+e1L7/8stirnbduvX6ETVnaW1DFhgqiMtmxY0exT0c3NTXBDz/8gN69e79xG8OGDcO2bduRm5uLc+fOCZdrO3XqBEdHR4SFyXcNsrOzw6ZNmzB58mRkZ2fj+PETOH78hNwyyuoO1LlzJ2zbtg1A3nNH7O3zuu0tWDAfs2bNKnE9ZcTo4uKMzz77DGvWrAEALF++AsuXrwCQV4QUN8CIg4MDfvjhe0yZ8hHEYjGWLFmCJUuKPiSxbVvF7/XLycnG4cOH5e41K8jW1hY9exb/nKWCyppba2trDB48WO75dc2bN0fDhg3LFXfXrl1Qt25duVGQOnfuXOThqMpqnzKV9zOiTOU9plSVv27dumLbtm14//33IZVKsXTpUhgY6GPKlClYuHAh7t69h2vXruHhw0fFbr+k+wFatWqFS5cuwcXFVW66lZUVFi36UnhtZmYGb29vPHjwAACEQgoAkpOTSo29vPEtWrQI9+/n/TKaPyDM8OEjcOPGDURERGDXrl3w9fXF0KFDYGJigr59++Cff/JG9Xznnbz7PkUiEUaOHIEXL/4osq/58+fh3r27uHjxElJTUzFlykdy8wsOOlORduebOPF/2Lhxk9wv3aNHl+3+1HzDhr0lnP9Onz4l3Lt07tx54fg3NDRE+/btiqzbuHFj2NnZITIyEqmpqfj777+F9VWlrDlW5P2rSlUdb3k/S8HBIThx4iSWLl1a7PYK3t+siIyMDBw9egxHj5Y80nTDhg3luuwWVNbzT3naXdXv0ZuoU3wuLi44cOAAtm/fLjfdysoKixcvKmEt1fL29sayZUsxb958ZGVlFXk+pIWFBbZt21rmLqKKbE8qlRb7vTXfggULMHTokCLT09PTcfly3g9lbm5uaN68WZlizMcrbpVEJBLB0NAQdevWRdeuXbF8+TLcu3dPGAr6TVq3bo19+/4QBvGwsrLC+++/j717f4dIVPyIRkOGDMbFixfw7rvvwtXVFbq6urC0tET37t1x6NAhub7lFTFo0CAsXLgQ9erVK3O3PGXGuHDhAqxevQru7u7Q0dGBp6cntm/fVuqgHCNGjMD58+cwduxYODs7Q1dXFyYmJmjYsCFGjBiB3377FRMmlD4qXWneeust7N37O0aMGAEPDw8YGBhAV1cXrq6umDBhAs6c+Rfm5mZv3E55cltwdDqg/FfbgLyra/m//Ocr7h9tZbVPmRT5jChTeY4pVeZvyJDBcqNazZkzF7/8shOGhoY4duwo1q5di/bt28HMzAza2tpwcHBA27ZtMX/+PFy4UPxD7w0MDHDs2FEMHz4cpqYmMDY2xoAB/XHq1Ck4OTnJLfvzzz+he/fuRa5CvUl54jt58iQ2bdoMIK9o2rjxRwCAubkZNm3aJPxSPW3aNOFHiM2bN+PDDz+Ao6Mj9PT00LRpU+zd+3uJ94Dq6+vjyJEj+O6779CpU0eYm5tDR0cHHh4e+Oyzz+QeslqRdudzdXVBjx49hNdWVlYlPkS7JOPHjxd+/S44sNDvv7/u+tipU8cSe3j4+r6+ClNwhD5VKU+Oy/v+VbWqjLe8n/U//zyAL7/8Em3btoWNjQ00NTVhbm6Ojh07YNOmjaX+AFsW77wzCosWLUKfPr3h5uYGc3NzaGpqwtjYGE2bNsW8eXNx+vSpEn84Kuv5p7ztVvdjSl3iq1u3Lk6dOonevXvD2Ni41PN/Zfr4449x+PBh9OzZE+bm5tDW1oaTkyPee+89XLx4QRj0SVXbq1evHg4c2I8+fXrDwcEBOjo6sLe3x7Bhw/Dvv6cxa9bMYvdz/PjxAvdZvv/GEXgL05BIxFV3lzkRKZWLiysSEhKgo6ODgAD/EocFJ3qTn376GZ99lvfFuUuXLvjrryNVHFHNVzDnU6Z8iG+++abc25g5cxa2bdsGkUiEW7dulvhMNyJ1xvNP1WL+VadXr964fv06bGxscOfO7SIDybwJr7gR1QBisRi7du1CQkICgLx7rli0EVUfUVFRQtdukUiEiRMnKrSdOXNmw8TEBFKptMjzhIiIqOpcunQJ16/njZ46d+7cchdtAO9xI6r2Jk2ajH379gmvTUxMMGvWF1UYERGV1fPnz+Ht3VRu2vjx41G/fn2FtmdlZYWXL1+8eUEiIqpUHTt2LPN9zyXhFTeiGsLAwABt27bFgQP74er65meXEJH60NDQgJ2dHT755GOsXr2qqsMhIiI1xHvciIiIiIiI1ByvuBEREREREak5Fm5ERERERERqjoWbkslkMkilUqU93JqIiIiIiIiFm5LJZDLEREfUiMJNKpUiKjIMUqm0qkOpcZhb1WFuVYe5VR3mVnWYW9VhblWL+aXCWLgRERERERGpORZuREREREREao6FGxERERERkZpj4UZERERERKTmWLgRERERERGpORZuREREREREao6FGxERERERkZpj4UZERERERKTmWLgRVREDA6OqDoGI1IiOjm5Vh1Bj8XxLRDWBVlUHQFQbiUQimJiaVXUYRKQmRCIRLCzrVHUYNRLPt0RUU1R54ZaZmYkrVy7j3H9nERAQgN69+2Di/yYVu+yN69exb99eREREwMLCAn369sOAAQMVWu7WzZvYufMXpKenoXPnLhg//l2INDWF+Xt/34Oo6ChMn/6ZchtMRERERERUTlVeuB0//g9ioqMxZux4bFi/tsTlHj9+hHXr1mDChIno2s0XT548xto1q6GpqYm+ffuVa7nU1FR8990GfPLpVLi6umHRlwvh7OyMrt18AQBhYWH499/TWLNmnWobXwvJpBJIwo8hN8UfmiYNoOXQHxqish2GEmkuzsQEIig1Du7GVuhu7QEtkeabV1RRPOqwT0muDBcC0hESmwPXOjro7GkILU0NtY23KqkqVzUlP9VZZXwOVKm6x6/OmFsiqkmq/NvFW28NK9Nyfx44gMaNm6BX7z4AgGbNmqNXr944+OcB9O7VW7haVpblgoODYGhohNat2wAAOnTogEePHgmF27atWzBy5NswMzdXdnNrtdyEe0i/MByy9OfCNA3DejDsfACaFs1KXfdxciQm39yLsMwkYZqjvhm2thqFxqZ2lR6PopS5T7+ILMzYE4mIJIkwzd5MC+tH26GhvZ7axVvZNDRefzlTVa6qc35qisr4HKhSdY9fnTG3RFTTVIvBSSRiMfz8nqBp06Zy072bNkVycjJevHhRruWK9epL3tmzZyCRiNGjZy/lNqKWk0klRb7gAoAs/XnedKmkhDXzrrQVLtoAICwzCZNv7oVEmlup8ShKmfuU5MqKfCEBgIgkCWbsiYQkV6ZW8VYlVeWqpuSnOquMz4EqVff41RlzS0Q1UZVfcSuLuPh45ObmwqqO/I3bdazyXkfHRMPZxaXMy7m5uSMtLQ03blyHu5s7rl27iqFvDUNKSgp+37MH8xcshEhUtppWLBZDLBYLr2WyvH8MpFKpwu1VF/ltqEhb8vMoCT9W5AtuPln6c0jC/4G20yC5/eWveyYmsEjRli8sMwlnY56il22DMsVakXgUpcx95m/rQkB6kS8k+SKSJLgYmI5uDY0Uir8qcqRMUqkUWlrawhU3ZeequuenIpRxTlCGyvgcqFJ1j1+dMbeVS13OCTVVVee3rN+FqfJUi8ItJzsbAKCjrSM3XUcn73X2q/llXc7Y2BifTp2KX3ftRFpaOjp17owunbtg06aN6Ny5M+zt7bHxxx9w48Z1WFhYYPLkD9GgYcNiYzt06CAO7N8nvNbT08WqVasQFxtV0WarDUXboqOjK4ySlpviX+qyuSn+0Ebel9ykxHgAENYNSo0rdd2g1FihcEtKjEdOTnaxyxkYGAkji5UnnrTUFGRkpJW6fEmUuc+C+QyJzSl1WyGxOej26pAtLSeFVeQ9K+s+KkPBYdWVmauakp+KqsrzW2V8DlSpusevzgqeb8uT24qc4ylPTfrOo46qKr+2do5Vsl8qWbUo3HR0876E5YjlT8Q5OXmvdV/NL+tyANC6dRvhHjcAePL4cd5AJus24Nixo3j58gW++/5HXL9+DRs2rMP3P2yEtrZ2kdiGDn1LbsRKmUyG1JREWNWxrfa/VEilUsTFRimlLZomDco838zcUm6eu7FVqeu6G7++wlp4XWXEY2RsAiNjkzJtt7L26VpHp8R5heeXNSelxfOm+YruQxWkUqlQ/AOqy1V1zU9FKPOcoAyV8TlQpeoevzorT26VdY6vjdTtnFDTML9UWLUo3KwsLaGpqYm42Fi56bFxea9trG3KtVxhErEY27ZtwfsTJkJPTw+PHz1Cp85dYGJigu7de+CXHT8jIiIc9eo5F1lXW1tbrqCTSqVITUmESCSqMR8yZbRFy6E/NAzrFdu1TMOwHrQcXo8MWnhf3a094KhvVmx3SUd9M/ha1y9xXVXEoyhl7rOzpyHszbSK7Qpkb6aFTh6GZd5WZcRb2SQSMWQyGTQ0NFSWq+qcn4pSl/NbZXwOVKm6x6/OmNvKpS7nhJqK+aV81eIo0NLWRsOGjXD//n256Q/u34epqSnq1q1bruUKO3z4EBwcHdGyZSsAr+9TyyeTyaABDh9cERoiLRh2PgANw3ry01+NwFfa8OlaIk1sbTUKjvpmctPzR5VU5JEAFYlHUcrcp5amBtaPtoO9mfw6+SOmKWO466rIkSqoKlc1JT/VWWV8DlSpusevzphbIqqJqs03i7eGDcOypUtx6uQJdO3mCz+/Jzh16iRGjxkr9+Dssi6XLzIyAidOHMeqVWuEaY0aN8bFC+fRsWMnXL9+DcbGxrCzt6+UdtZkmhbNYDzIH5Lwfwo886pfmb7gNja1w3nfqTgb8xRBqbFwN64DX+v6FXqOW0XiUYd9NrTXw9EZzrgY+PoZRZ08lPuMoqrIkbLkX3EDVJer6pyfmqIyPgeqVN3jV2fMLRHVNBoSibhKx8S9ceM61qxeVWS6SCTC3j/2yy97/Tr++GMvIiMjYG5ugb59+2LAwEFFt1nG5QBg2dIl8PHxQb/+A4Rp2dnZ2LplM27dugULCwtMmjwZjRo1LlN7pFIpYqIjYG1jX+0va9ektlDtweNWdZhbIiqI5wTVYn6psCov3GqamvQhq0ltodqDx63qMLdEVBDPCarF/FJhPAqIqoBUKkVKchKffUNEAPLOCQnxsTwnqADPt0RUU7BwI6oifG4QERXE57SpDs+3RFQTsHAjIiIiIiJScyzciIiIiIiI1BwLNyIiIiIiIjXHwo2IiIiIiEjNsXAjIiIiIiJScyzciIiIiIiI1BwLNyIiIiIiIjXHwo2IiIiIiEjNsXAjIiIiIiJScyzciIiIiIiI1BwLNyIiIiIiIjXHwo2IiIiIiEjNsXAjIiIiIiJScyzciIiIiIiI1BwLNyIiIiIiIjXHwo2IiIiIiEjNsXAjIiIiIiJScwoVbqmpqcjNzVV2LERERERERFQMhQq3Gzeu46MpH2Lfvj+QEB+v7JiIiIiIiIioAC1FVvLxaYno6GicPnUKB/88gJYtW6FXr97w8vaGhoaGsmMkIiIiIiKq1RQq3MzMzDB69BiMHPk2bly/jlOnTmL58qWwtbVFz1690bVrNxgbGys7ViIiIiIiolpJocJNWFlLC+07dED7Dh0Q9vIlfvvtV/y6ayf2/v47OnbsiLeGDYONja2yYiUiIiIiIqqVKjyqpDQ3Fzdv3sDOnb/g7t07cHZ2wZChQxESEoxZMz9HaGiIMuIkIiIiIiKqtRS+4paYmIgzZ/7FmX9PIykpGW3atsGSpcvRoEEDAMDw4SOwaeOPuHD+PFxcXJUWMBERERERUW2jUOF28+YNrFu7BqampujRoyd69OwFMzMzuWU0NDTQsFEjBD19qow4iYiIiIiIai2FCjdDQ0NMnTYdrVu3gaamZonLdevmi27dfBUOjoiIiIiIiBQs3Bo1aqzsOIiIiIiIiKgEZS7c/P38cO7cf2VatkHDhujatZvCQREREREREdFrZS7cMjIyEBUVVaZl7e3tFQ6IiIiIiIiI5JW5cGvh44MWPj6qjIWIiIiIiIiKodBz3KKiohAYGFjueURERERERFR+ChVujx8/wtkz/5Z7HhEREREREZWfQoVbadLT06Gvr6/szRIREREREdVa5XocgL+/Py5cOI+I8HAkJydh69YtcvNzsrNx9+4dvPve+0oNkoiIiIiIqDYrV+GWnp6Gly9eIDk5CZmZWXj54oXcfAMDfQwcOAgdOnRUapBERERERES1WbkKNx+flvDxaYk7t28jLOwlBg0eoqKwiIiIiIiIKF+5Crd8LXx84OHpieTkZJiamgrTMzMzkZmZCQsLC6UFSEREREREVNspNDiJNDcXX61Yjvi4OPnpUimWL1uCpMREpQRHREREREREChZu/v7+0NXVgaubm9x0Q0ND+Pi0xIWLF5QSHBERERERESlYuMXFxcLY2KTYecYmJoiNialQUERERERERPSaQoWbja0tAgL8kZmZWWTevbt3YW1jU+HAiIiIiIiIKI9ChVv9+h4wMjLGVyuW4/79e4iJjkZgQAC++3YDgoKeomPHTsqOk4iIiIiIqNZSaFRJkUiEWbO+wNq1q7Fi+TJhurm5Ob74Yg7Mzc2VFiAREREREVFtp1DhBgC2dnZYtXotQkNDER8fB2MjY7i7u0NLW1uZ8REREREREdV6ChduAKChoQFXV1e4uroqKx4iIiIiIiIqROHCTSqV4sKF8wh6+hQpKSmQyWTCPC9vb/Tq1VspARIREREREdV2Chdua9esRmBgIMzMTJGRkQkLCwsEBQXB0NAAXt7eyoyRiIiIiIioVlOocHv+/Bn8/f2wfsN3uHnzBp4GBuLDKR8hLCwMixcthLt7fWXHSUREREREVGsp9DiA8LBwNG7cBCYmJhCJRBCLxQAAR0dH9OjZCzdvXFdqkEQ1kY6OblWHQERqxNjYrKpDICI1w+8KVJBCV9wyszJhYGAAADAxMUFCQoIwz8jICFFRUcqJrpDk5GRcuXIZ8XFxMDE1hY+PDxwcHIssl5WVhUsXLyAiIgIWFhbo2LETzAo9oiArMxNnzp5Beloa2rZrj7p168rNDwwMhL+/HwYNGqyStlDtJhKJYGFZp6rDICI1IRKJYGhkVNVhEJEa4XcFKkyhK24FuTi7ICjoKW7dvInnz5/hzL//wt7OThmxyXn6NBCffPwR7ty+BWMTEzx/9gwzP/8c5879J7dcWmoq5s6ZjTNnzsDM3ByPHz/GZ59Nx8uXL4RlZDIZli1bgocPHiBHLMaC+XPl5ovFYmz88Xs4FlMUEhERERERVTaFrrjZ2NhAJpUCACwsLTFgwECsWvUNAMDZ2QXdfLsrL8JX/ti7Fw4ODpg3fyE0NDQAANra2vh110507dpNWG7//n3IysrENytXQVdXF4MGDcaSxYvw0/btWLxkKQAgOCgIYWFh2Lb9Z+jo6CApMRH/nj6N9ydMBAAcPnwITk510cLHR+ntUBVZrgTiR5eRG/0cmjb1oN2kAzQ0K/S0B5WQSHNxJiYQQalxcDe2QndrD2iJNFW+bkmqIm/V5b3KJ5NKIAk/htwUf2iaNICWQ39oiNQ33somyZXhQkA6QmJz4FpHB509DaGlqVGhbdbmnKsin+quNrYZqL3tVpXqkM/qEKO6YK6oOAp9E2jSxAtNmngJr98e9Q569uyFlNRUODk5QVOzYl+mixMbGwvPBg2Eog0AnF1ccPbsGWRlZkJPXx8AcOXKZXTq3AW6uq/7BHfr1g0//PA9EhISYGFhgcSkRJiZmUFHRwcAYGNrK1xxi4yMwMkTx7Fq1Rqlt0FVJC8Dkb5tNqQJr7uoiixsYThpJbScPKowMnmPkyMx+eZehGUmCdMc9c2wtdUoNDYt/SptRdYtSVXkrbq8V/lyE+4h/cJwyNKfC9M0DOvBsPMBaFo0q7rA1IRfRBZm7IlERJJEmGZvpoX1o+3Q0F5PoW3W5pyrIp/qrja2Gai97VaV6pDP6hCjumCuqCQKdZUs+My2fBaWlnB2doampiZysrMrHFhhDRs1QoC/nzAQCgA8eHAfLi6uQtGWlJiI5ORk1HVyklvX6dX9ay9e5H0RsrG2QVxcPNLS0gDkjZJpY20DANi+bSveGjYcFpaWSm+DKshyJUUKAQCQJkQhfdtsyHIlJaxZuSTS3CKFFwCEZSZh8s29kEhzVbJuSaoib9Xlvconk0qKFBAAIEt/njddql7xVjZJrqzIP6wAEJEkwYw9kZDkFj1Pvkltzrkq8qnuamObgdrbblWpDvmsDjGqC+aKSqPQFbdLly4iOzsbPXr0LDLv7Jl/8ezZM0yY+L8KB1fQu+++hx0//4SZn38GzwYN8PLFCxgYGGDmrFnCMqlpqQBQ5AZvQ8O812mpeYVa3Xr10Lx5cyz6ciFsbW0RGBiI996bgAvnzyE9PQN9+vTF06eBuH3rFiwsLODr2x1a2trFxiUWi+WKyfyiVvqqK6mqiER5Nbf40eUihUA+aUIUxI+vQMe7s0Ix5S9fkbbkx3kmJrBI4ZUvLDMJZ2OeopdtA7n9VWTdN8Wjyrypwz4rIj9eSfixIgVEPln6c0jC/4G20yAAVRtvYco4bkuTn58LAelF/mHNF5EkwcXAdHRraFSmWKpLzlWRW1XkU93VxjYDVdduVZ8Tqoo6HEdvyq06xFhdqGOu8mMi9aFQ4WZuZo4VK5bD388PkyZ/AF1dXWRnZ2Pbtq24fOkSpnz0kbLjRHBQEB4+fAgPDw84OjhAS0sLN29cx/1799CjZy8AgAbyulEWuSL46rWG6HU3y88+n4n79+8hPS0dH3zwIUSamti9+zfMnj0X/n5+WLXqG/TrPwCXL1/G/fv3MeuL2cXGdejQQRzYv094raeni1WrViEuVjUjawJ5Q8PmjzKUG138l7x80qhnwKtiICkxHjk55b8aqmhbDAyMYGJqBgAISo0rddmg1Fih+EpLTQEAhdfNyEgrdjljYzOhqC9P3jIzMpCamlTq8iWp7Peqogq+Z7kp/qUum5viD23kFRGl5b2qqOIzWPD9DInNKXXZkNgcdGuY93dp76fcMVKOnFfVMQIoL7cFP5PlyWdFPpNVrTa2GQBMzSyh/6p3THnanZ2djeSkeKXEoMp/lyubKs5FFVFcbtUtRnWmrrmyteMgfepGsXvcvLywZOkyrF+3FvPmzsboMWOxZ/dvyMjIwJIlS+Hh6anUICViMdavX4v27Ttg4v8mCdNdXFywffs2NG7SBHZ29jA1MwMApKamyq2f8uq1qYmpME0kEqF58xbC682bN6Fdu/ZwdXPDunVr0Kt3H4wc+TZS+vTFpP9NQFxcHKysrIrENnToWxgwYKDwWiaTITUlEVZ1bCvllwpNm3qlzhfZOgt/m5mXr/unVCpFXGyUUtriblw0d/LzXw93a2RsUqF1C69fnPLkTd/AAPqvHn9REap8r1RB06RBmeeXNe+VQZnHbWlc6+iUeX5Z38/y5LwqjhFV5rY8+VTWZ7Kq1cY2A+Vrt66uLqxt7Cu0v8o6J1QVVZyLyqqsua3KGKsb5opKo/AwZR4eHli1eg0WLpiHld98DRcXV6xctQYmJsr/8packoLU1FS4ubvLTa/vXh9SqRSREZGws7OHsbEx6tSxRkhwMLp37yEsFxISDJFIBGdn52K37+/nh/v37mHd+g0AgKTEJHh7eQPIe06drq4ukhITiy3ctLW1oV2gG6VUKkVqSiJEIlGl/AOh3aQDRBa2xXbBE1nYQrtx+9evFYxHGW3pbu0BR32zYrs8Ouqbwde6vtz+lLVuSSojb+qwz4rQcugPDcN6xXbd0zCsBy2HfsJrdYi3MFV/Bjt7GsLeTKvYLi32Zlro5GEoF0tZVJecqyK3qsinuquNbQaqrt2V9e9yZVOH4+hNuVWHGKsL5opKo/A7LhGL8ccfexEVFQVv76Z48eI5zp75t9iBSyrKwsICJiYmuHPnttz0W7duQiQSwanu68FIunXrhitXriApMRFA3j1op0+dQstWrWBkbFy0HRIJtm3bgvffnyB047C2scHz53lfnGKio5GdnYM61tZKb5cyaGhqwXDSSogsbOWm549UqC7DzGuJNLG11Sg46pvJTc8fGbK0Yf0rsm5JqiJv1eW9yqch0oJh5wPQMJS/Upg/wmFtGZ6+JFqaGlg/2g72ZvJ5yB/5S5Fhm2tzzlWRT3VXG9sM1N52q0p1yGd1iFFdMFdUGg2JRFzuSis+Ph5rVq9EfHw8pk2fgcaNm+DSxYvYsmUzGjdujE8++bTYIqki7t69g++/+xZWVnXg4uKC6OhoPH0aiDFjx6Ffv/7Ccjk5OVi96hu8fPkSzZo1x9OnTyGV5mLhl4thYWFRZLuHDh1EYEAAZs+ZK0x7FhqKJUsWoWnTZggODkbzFi0w4dUz3t5EKpUiJjoC1jb2lfpLiCxXAvHjK5BGPYPI1hnajdtXuBBQRVsk0lycjXmKoNRYuBvXga91/XI9x03RdUuiiryp4z4rIu+ZYv8UeKZYP7UuICr7MyjJleFi4Otn7XTyUNZz3NQv55WRW1XkU93VxjYDldfuqvp3ubJVxXFU3tzW1mNdEcwVFUehwu3MmX9x4fx5TJ/xGczNzYXpYWFhWLd2NRo0aIjJH3yo1EABIDMzE35+T5CYmAhjY2PUr+8ht/98MpkMfk+eICIyAhbmFvDy9pbrzljQ2bNn0KxpsyLD/8fFxeHJ40ewsLBEEy+vYtctTk36B6ImtYVqDx63qsPcUnXE41Z1mFuiyqVQ4RYXGwsLCwuIinnQdlZWFh4/fgQfn5ZKCbC6qUknsZrUFqo9eNyqDnNL1RGPW9Vhbokql0J9b6zq5I3il5WVhTu3byMqKhJ29vbwauKF1LTUWlu0EZWVVCpFUmI8zMwt+Y8dEUEqlSIzIwP6BgY8JxARAH5XoKIUPgpu3bqJTz/5GJs2/Yjjx//B/Xv3IAPw1YoVyMrKUmKIRDVTbXtODRGVrjo/p42IVIPfFagghQq3lORkfPftBgwcOBA7duzEqHdGAwCMjY3RuEkTXLp0UalBEhERERER1WYKdZV89PgR6tf3wKDBQ4rMc3JywvNnzyoYFhEREREREeVT6IqbNFcKHZ3in+yek5NT4giOREREREREVH4KFW7OLi54+PABQkND5KbLZDLcunkT7vXrKyU4IiIiIiIiUrCrpKOjI1r4+GD+vLlo36EjcnKykZiQiNWrViI7OwutW7dRdpxERERERES1lsKjSn766TT06z8A9+7ewbWrVxESEgIdHR0sWLgIWloK1YNERERERERUDIUrLG1tbYwdOw5jx45DdnY2tLW1+YwJIiIiIiIiFVDKpTFdXV1lbIaIiIiIiIiKwUtkREREREREao6FGxERERERkZpj4UZERERERKTmWLgRERERERGpOYUKtydPHuO//86Wex4RERERERGVn0KFW2RkJAL8/UueFxBQoaCIiIiIiIjoNaV3lYyMiICJsbGyN0tERERERFRrles5blevXsGe3buRmZkBsViCx48fy83PyclGUlISFi1aotQgiYiIiIiIarNyFW42Nrbo1LkzQkNCkJAQD5+WreTmG+jro0HDRnB3d1dqkERERERERLVZuQo3V1dXuLq64lloKBISEtDCx0dVcREREREREdEr5Src8jm7uMDZxUXZsRAREREREVExFCrc8t28eQNXLl9GbGwszMzM4OXtje7de0BLq0KbJSIiIiIiogIUHlVy448/YM3qVUhMTISdnR0kEgl+3bUTCxfMQ052tjJjJCIiIiIiqtUUujQWHByEK1cuY9myFfDw9BSmJyQk4MuF83H2v7Po06ev0oIkIiIiIiKqzRS64hYeHg5v76ZyRRsAWFhYoHuPnogID1dKcERERERERKRg4VanTh1kZmYWOy8zMxNWdepUKCgiIiIiIiJ6TaHCzd29PlJSkvH3339BIpEI02/fvoVz/51F2zZtlRYgERERERFRbafQPW537txGamoqft21E/v+2Atzc3OkpqYiPT0d+voGWLnya2HZNm3aYuTbo5QWMBERERERUW2jUOFmaWmFLl26lmnZes7OiuyCiIiIiIiIXlGocHN3d4e7u7uyYyEiIiIiIqJiVOhJ2VlZWbhz+zaioiJhZ28PryZeSE1LhZ2dvbLiIyIiIiIiqvUULtxu3bqJLZs3IysrE3p6evDxaYkmTbzw1YoVWL1mLfT09JQZJxERERERUa2l0KiSKcnJ+O7bDRg4cCB27NiJUe+MBgAYGxujcZMmuHTpolKDJCIiIiIiqs0UuuL26PEj1K/vgUGDhxSZ5+TkhOfPnlUwLCIiIiIiIsqn0BU3aa4UOjo6xc7LycmBtrZ2hYIiIiIiIiKi1xQq3JxdXPDw4QOEhobITZfJZLh18ybc69dXSnBERERERESkYFdJR0dHtPDxwfx5c9G+Q0fk5GQjMSERq1etRHZ2Flq3bqPsOImIiIiIiGotha64AcCnn05Dv/4DcO/uHVy7ehUhISHQ0dHBgoWLoKVVoacMEBERERERUQEKVVgymQza2toYO3Ycxo4dh+zsbGhra0MkyqsDs7Ozoaurq9RAiYiIiIiIaiuFrrhdvXoFzwqMHKmrqysUbSEhIfhj7+9KCY6IiIiIiIgULNxEIhG++XoFEuLj5aaHh4fhqxXLYGJqqpTgiIiIiIiISMHCrW3bdvD2bopvvvkaWZmZAIDY2BgsX7YULVr4YHAxz3cjIiIiIiIixSg8OMnkDz6EoaEh1q9fh4T4eCxbugRu7u748MMp0NDQUGaMREREREREtZrChZuWlhZmzvoCMTHRmD59KqytbTB92gyINDWVGR8REREREVGtV+ZRJSMjIxDg719kert27XH06N9o2aoVLl26CACws7eHp2cD5UVJRERERERUi5W5cPPz88PWLZtLnP/Ljp+Fv319u7NwIyIiIiIiUpIyF26+vt3h69tdlbEQERERERFRMRS+x42IiIiIiIgqBws3IiIiIiIiNcfCjYiIiIiISM1Vy8ItIyMDCfHxpS4jFosREx0tPCC8ODnZ2UhOTi52XnZ2NqIiIysUJxERERERkTKUeXASdRAdHYXt27bB398fJiYm0NLSxISJ/0PTps3kljty5DAO/nkARkZGSElJQadOnTHxf5OgWeAZc//8cwy/79kDbW0tuLm5Y9asL6CjqyvM//mn7dA3MMB7771fWc0jIiIiIiIqllIKt5ycHEgkEhgYGChjc8VKSUnBoi8XolnzFti2/Sfo6ekhLi4O169fkyvcLl++hL2/78H8+QvRxMsLUZGRWLBgHoyMjDB6zFgAQEJ8PHb/9itWfPUNHBwcMH/eXJw6fQoDBgwEAPj5PcGDBw+wfv0GlbWHiIiIiIiorBTuKpmVmYn9+/dhyoeTMW7saOza+QsiIyOwedNGZcYnOHBgP7S1tTFp0mTo6ekBAKysrNC//wC55Y7+/TdatmyFJl5eAABbOzv06dsPJ0+egFgsBgA8e/YM5ubmcHZ2hra2Nnx8fBAcFAQAkEgk2LZ1C96fMAF6+voqaQsREREREVF5KFS4yWQyfP31V7h08SIGDByE3r37AADs7OwRGRmJwIAApQYJANeuXkGr1m2gqamJuLg4ZGRkFFkmOzsbISEhaNiokdz0Rg0bITMzE8+ehQIAdHR0kJ2dLczPysqC7qtukn/9dQS2tnZo3bqN0ttARERERESkCIW6Svr5PUFMTAzWrl0HA0NDnDnzL54GBgIAGjVqhNu3b8HD01NpQaalpiIpKQm5EglmTJ+KtLR0ZGZmwMnJCR98+BGcnZ0BAAkJ8ZDJpLCwsJBb39zCHAAQHxeP+vUB9/r1AQCHDx+Cm5sbLl26iA8+mILo6Cj8c+wovvlmFQAgKTERRkZG0NLWLjE2sVgsXMkD8opaAJBKpUprf1XJb0NNaIu6YW5Vh7lVHeZWdZhb1WFuVYe5Va2qzq9IVC3HMKzRFCrcYmJi4OHhAQNDwyLzjE1MEBEeXuHACsrOyQEAHD/+Dz6fOQtt2rRFVlYW1q1dg1Urv8a69d9CT08P4py8AkpLS75Z2lp5hVeOOG87enp6mDt3Pv74Yy9u3byJ4SNGomWrVlixfBmGDBkKXT09zJ83F9HRURCLxZj8wYfo0KFjsbEdOnQQB/bvE17r6eli1apViIuNUmoOqlJNaou6YW5Vh7lVHeZWdZhb1WFuVYe5Va2qyq+tnWOV7JdKplDhZmZmhmfPQiGRSIoUSWEvX8LaxkYpweXTf3VPW8OGDdGmTVsAecXXqFHvYM6cL+Dv54dmzZsL96RlZWXJrZ//2kD/9eAprm5umDtvvvD68uVLSEpKQr9+/bFnz24YGhpi67af8PDhA6xftxY+Pi2Fe+sKGjr0LWFQEyDviltqSiKs6thW+18qpFIp4mKjakRb1A1zqzrMreowt6rD3KoOc6s6zK1qMb9UmEKFW/49Y2vXrMY774yG7NUl3CdPHuPq1Sv4ZuUqpQZpYGgIMzMzmJmZy003M897nZScBACwsrSEjo4OoqLkf5nIf21nb1fs9tPT07Fr507MmvUFRJqaCA4OQtt27SESidC0aTNIJBJERETA1dW1yLra2trQLtCVUiqVIjUlESKRqMZ8yGpSW9QNc6s6zK3qMLeqw9yqDnOrOsytajG/lE+hwk1HVxfTZ3yGNatXYebMz6ChkXdAnT9/DpMmTYaNja2y40TLlq3w+PEjSKVS4eANCQkGADg5OQEARJqaaNasOW7fuonhw0cI6966dRM2NrZwcCj+ku+e3b+hdevWwr1vOjo6yH51lU4iFiM3Nxe6ujpKbxMREREREVFZKPwct0aNGuP7Hzbi5s0biIuNhaGREXxa+Ci9m2S+t94ahuvXr2Pzpo3o2asXEhMS8csvP6N1mzZwc3MXlhs58m3Mnz8Xv+z4GV27dsMTvyc4f/4cpk2fUex2AwMDcfv2Laxbt0GY5uPTEn///Tc8PDxx69ZN2Nraws62+Kt1REREREREqqYhkYhlVR1EWUVFReHgnwfw7NkzGBsbo3nz5ujbrz80NTXllgsNDcGhgwcRGRkBc3ML9OnTFy18fIrd5vp1a9GufXu0bdtOmCbNzcXhw4dw69YtWFhYYMzYsbCzsy9TjFKpFDHREbC2sa/2l7VrUlvUDXOrOsyt6jC3qsPcqg5zqzrMrWoxv1RYtSrcqoOa9CGrSW1RN8yt6jC3qsPcqg5zqzrMreowt6rF/FJhCneVvHHjOvb9sRcRERGQSCRy83x9u+PDKR9VODgiIiIiIiJSsHCLjo7ChvXr0KVLV/TrP6DIIwFseT8YERERERGR0ihUuD19+hSNGzfGBx9OUXY8REREREREVIhCHWb19fWhq1v0YdRERERERESkfAoVbo0bNUZoaChevHih7HiIiIiIiIioEIW6Surp62Po0KGYO+cLuLvXh6GRodx8b++m6NOnr1ICJCIiIiIiqu0UKtySk5Px66+/wsTEBAYGBtAUyT9HjUOWEhERERERKY9Chdvjx49gYWGBVavXQFtbW9kxERERERERUQEKXRrT0tJCPed6LNqIiIiIiIgqgUKFm4uLC4KePkVOdray4yEiIiIiIqJCFOoqKZFIYGJiioUL56NT5y4wMpQfnMTO3h6eng2UEiAREREREVFtp1Dh5ufnh5CQYADA8193FZnv69udhRsREREREZGSKFS4+fp2h69vd2XHQkRERERERMXguP1ERERERERqTqErbvlCQ0MQHBSElJQUyApMd3F2QQsfnwqGRkREREREREAFCre9v+/BoUOHYGBgAKk0FwCQmZkJXV099O3Xj4UbERERERGRkihUuEVHR+HYsWNYtXo1goKC8DQwEB9O+QjXrl3Fls2b0aNHT2XHSUREREREVGspdI/b82fP0aRJE9Sr5wwNALm5eVfc2rZth06dO+Py5UvKjJGIiIiIiKhWU6hwS0tLhampKQDAyMgYKSkpwjxbW1vEx8crJzoiIiIiIiJSrHArOBCJo6Mj/P39ERMTA7FYjJs3bsDc3FxJ4REREREREZFC97gZGhjC7FVxZu/ggKZNm+LTTz6GlpYW9PR08emnU5UaJBERERERUW2mUOHWtl07tG3XTng9ffoM3Lt3DympKWjWtJlQ1BEREREREVHFVeg5bvlEmpoc/p+IiIiIiEhFFLrHjYiIiIiIiCqPwlfcAgL8cfDPA4iIiIBYLJab17FjJ4wdN77CwREREREREZGChVtSUhK+WrEcnp4N0K2bL7S0teXm16tXTynBERERERERkYKFW0CAPxwdHTFv/gJlx0NERERERESFKHSPm6ZIE3WsrZUdCxERERERERVDocLN1c0NwUHByM7OVnY8REREREREVIhCXSWzs7NgZWWJLxfOR6fOXWBkaCg3387eHp6eDZQSIBERERERUW2nUOHm5+cHPz8/AMDzX3cVme/r252FGxERERERkZIoVLj5+naHr293ZcdCRERERERExeADuImIiIiIiNQcCzciIiIiIiI1x8KNiIiIiIhIzbFwIyIiIiIiUnMs3IiIiIiIiNQcCzciIiIiIiI1x8KNiIiIiIhIzSn0HDcACAjwx8E/DyAiIgJisVhuXseOnTB23PgKB0dEREREREQKFm5JSUn4asVyeHo2QLduvtDS1pabX69ePaUER0RERERERAoWbgEB/nB0dMS8+QuUHQ8REREREREVotA9bpoiTdSxtlZ2LERERERERFQMhQo3Vzc3BAcFIzs7W9nxEBERERERUSFl7ioZGRmBAH9/4bWVlSW+XDgfnTp3gZGhodyydvb28PRsoLwoiYiIiIiIarEyF25+fn7YumVzkenPf91VZJqvb3cWbkREREREREpS5sLN17c7fH27qzIWIiIiIiIiKoZC97ilpaUhISGh3POIiIiIiIio/BQq3K5fv4Z9f+wt9zwiIiIiIiIqP4UKt9KIc3Kgo6Oj7M0SERERERHVWuV6AHd4eBgeP36MAH9/xMTE4NSpk3Lzc7KzcebMv+jdu49SgyQiIiIiIqrNylW4PX36FL/u2gWpNBdSqRTPnj2Tm29goI+GDRuha9duyoyxiLi4OGzZvBEikQhz5y0oOj82FseOHUVERAQsLCzQs1dvuLq6yi0TGxuDQ4cOIT0tDZ06d0bLlq3k5t+4fh0BAf4YN/5dlbaFiIiIiIjoTcpVuHXt2g1du3bDhfPnEBQcjAkTJqoqrhLJZDJs2vgDQkNDi30AeExMDObOmY3GTRqjV+/eePL4MebPm4v5CxagSRMvAIBELMaSxYvg49MSzZo3x/fffYvZc+aiUaPGAID09HT89NN2zPpidqW2jYiIiIiIqDjlKtzyde7SFZ27dFVyKGVz4sRxJCQkoEePnjh27GiR+fv+2AsTExNMnzYDIk1N+Pi0RGxcLHb+sgOr16wDADwNeoqMjAy8+977EIlECA4KwqVLl4TCbffu39CmTRu4u7tXatuIiIiIiIiKU+bCLezlSzx69BCOTk4wMzXDo0cPS1zW0clJuLqlTBEREfh9z27MX7AQ9+/fLzJfJpPh5s2b6NuvH0SamsL09u07YN3aNYiOjoKNjS3S09NhaGgEkShvbBYTU1OEh4UBAAIDA3Hn9i2sW/+t0uMnIiIiIiJSRJkLt5CQYOzd+zs6d+kKdzc37N37e4nLdu7SVemFmzQ3Fz/+8B18fbvD07NBsYVbfHw8MjMzYG9vLzfd3i7vdVhYGGxsbOHo6IS4uDjExMSgTp068HvyGE28vJGbm4ttWzfj3ffeh4GBQZniEovFEIvFwmuZTJYXr1SqaFPVRn4bakJb1A1zqzrMreowt6rD3KoOc6s6zK1qVXV+8y9wkPooc+FWuHtkZXeVPHT4EJKTkzHqndElLpORkQEAMNCXL7r0DfTl5tva2qJ///6YO2c2TE1NoKmpiT59+uLo0b9hYWGBdu3a4+LFC7hx/TosLCwwbPgImJiYFB/XoYM4sH+f8FpPTxerVq1CXGxUhdqrTmpSW9QNc6s6zK3qMLeqw9yqDnOrOsytalVVfm3tHKtkv1Qyhe5xS0pMhJa2NoyMjJQdT7GePXuGPw/sx9x586Gnp1ficpqaeb8M5Epz5abn5kpfzX/dfXLsuPHo2asX0tMzULduXSQmJuCvI0fw9dff4MqVy/h11068+977uHvnDlZ+8xVWfPVNsfscOvQtDBgwUHgtk8mQmpIIqzq21f6XCqlUirjYqBrRFnXD3KoOc6s6zK3qMLeqw9yqDnOrWswvFaZQ4Xb7zm1s3bIFrq6u8Pb2hpe3Nzw9G0BbW1vZ8QEAHj9+BE1NTRw+dBCHDx0EAERHR0MiycWypYvRoUNH+HbvAQsLSwBAUlKS3Pr5ry3MLeSm29jYCn//tH07Bg4aBGsbG/zyyw707tMXHTp0hI9PS7w7fhyiIiNha2dXJDZtbW25dkulUqSmJEIkEtWYD1lNaou6YW5Vh7lVHeZWdZhb1WFuVYe5VS3ml/IpVLh17NgJ5ubmePDgAW7duolDhw5CV1cXDRo0hHfTpmjVslWxRY6i2rRpCycnJ7lpF86fR3z8ZQweMhTWdawBAPr6+nByckJgQIDcQ8ADAwOgra0DZxeXYrd/7epVxMbGYuasLwDkPQ7A2NgYAKCnpwcdHR2kpacrrT1ERERERETloVDhpqurixYtfNCihQ8AIDExEffv38OJ48fx666dCA8Lw4dTPlJakFZWVrCyspKb5u/vD5FIA97eTeWm9+7dBzt37sTgwUNQt149pKSk4OSJ4+jStWux3SwzMzOxc+cOTJ/xGbS08tLhVLcunjx+jF69eiM4OAhSqRR2SixEiYiIiIiIykOhwi1fUlISHj58gAf37+PBgwdIS0uFl5eXUNBVhZ69eiMsPBzz58+Fs7MLwsLC4OFRH+PHv1vs8r//vgfNm7eAp2cDYdrgwYOxZPEizJr5OeLi4jB6zBgYGhpWVhOIiIiIiIjkKFS4PXnyGDt+/gkvXrxE3bp14e3tjY8+/hgNGzSEjq6usmMsVpcuXdCoUaMi0zU0NDBhwkQMHfoWoqIiYW5uAVtb22K2kKd169ZwdXWTm1anjjXWrf8WL168gLm5OSwtLZUePxERERERUVkpVLhFRkbi+fPnaNiwITp16gxv76awtrFRdmylsrGxlRtcpDBzc3OYm5u/cTslPW9OR0cH7u7uCsdHRERERESkLAoVbl26dIWdnR0ePHiA//47i+3bt8Ha2hre3k3h3bQpmjRuAgN2LSQiIiIiIlIKhQo3LS0tNGrUGI0aNcaoUe8gPT0dDx7cx+FDh3Dq1En4+nZX6uAkREREREREtVmFBieJjIzAwwcP8ODBAzx+/Ajp6emwsbFBfQ8PZcVHRERERERU6ylUuD24fx9btmxCbGwsDA2N0KRJE4weMxZNq+BeNyIiIiIioppOocJNJBKhe/ce8PZuClc3Nz7NnYiIiIiISIUUKtyaeHmhiVfxozESERERERGRcvFSGRERERERkZpj4UZERERERKTmWLgRERERERGpORZuREREREREao6FGxERERERkZpT+AHcUZGRkEqlsHdwgEwmw5EjhxEcFIRu3XzRwsdHmTESERERERHVagpdcROLxVi7djVEmpoAgMuXL+HgnweE6ZGREUoNkoiIiIiIqDZTqHB7GhgIY2MT2NraAgCuXrmCUe+Mxpy589Ctmy+uXbum1CCJiIiIiIhqM4UKt/iEeJiZmwMAZDIZAgL80bx5cwCAo5MTEhMTlRchERERERFRLafQPW6WlpZ4GhiInJwcBAYEQFNTC3Z29gCA2JgYWFlZKTVIIiIiIiKi2kyhwq1Bg4bQ0dHBR1M+QFZWFgYPGQoAkObm4u7dO/jiizlKDZKIiIiIiKg2U6hwE4lEWLJ0GW5cvwYDQ0O0adMWABAbF4v+AwbC1s5OqUESERERERHVZgoVbjKZDEZGRvDt3kNuuo2NLWxsbJGdnQ1dXV2lBEhERERERFTbKTQ4ydWrV/Ds2bNi54WEhOCPvb9XJCYiIiIiIiIqQKHCTSQS4ZuvVyAhPl5uenh4GL5asQwmpqZKCY6IiIiIiIgULNzatm0Hb++m+Oabr5GVmQkAiI2NwfJlS9GihQ8GDx6izBiJiIiIiIhqNYUKNwCY/MGHMDQ0xPr165AQH49lS5fAzd0dH344BRoaGsqMkYiIiIiIqFZTuHDT0tLCzFlfICYmGtOnT4W1tQ2mT5sBkaamMuMjIiIiIiKq9co8qmRkZAQC/P2LTG/Xrj2OHv0bLVu1wqVLFwEAdvb28PRsoLwoiYiIiIiIarEyF25+fn7YumVzifN/2fGz8Levb3cWbkREREREREpS5sLN17c7fH27qzIWIiIiIiIiKobC97gRERERERFR5SjzFbfCoiIjIZVKYe/gAJlMhiNHDiM4KAjduvmihY+PMmMkIiIiIiKq1RS64iYWi7F27WphBMnLly/h4J8HhOmRkRFKDZKIiIiIiKg2U6hwexoYCGNjE9ja2gIArl65glHvjMacufPQrZsvrl27ptQgiYiIiIiIajOFCrf4hHiYmZsDAGQyGQIC/NG8eXMAgKOTExITE5UXIRERERERUS2n0D1ulpaWeBoYiJycHAQGBEBTUwt2dvYAgNiYGFhZWSk1SCIiIiIiotpMocKtQYOG0NHRwUdTPkBWVhYGDxkKAJDm5uLu3Tv44os5Sg2SiIiIiIioNlOocBOJRFiydBluXL8GA0NDtGnTFgAQGxeL/gMGwtbOTqlBEhERERER1WYKPw7AyMgIvt17yE2zsbGFjY1thYMiIiIiIiKi1xQu3AAgNDQEwUFBSElJgazAdBdnFz7LjYiIiIiISEkULtz2/r4Hhw4dgoGBAaTSXABAZmYmdHX10LdfPxZuRERERERESqJQ4RYdHYVjx45h1erVCAoKwtPAQHw45SNcu3YVWzZvRo8ePZUdJxERERERUa2l0HPcnj97jiZNmqBePWdoAMjNzbvi1rZtO3Tq3BmXL19SZoxERERERES1mkKFW1paKkxNTQEARkbGSElJEebZ2toiPj5eOdERERERERGRYoVbwYFIHB0d4e/vj5iYGIjFYty8cQPm5uZKCo+IiIiIiIgUusfN0MAQZq+KM3sHBzRt2hSffvIxtLS0oKeni08/narUIImIiIiIiGozhQq3tu3aoW27dsLr6dNn4N69e0hJTUGzps2Eoo6IiIiIiIgqTqHCLS0tDTk5ObCwsAAAiDQ1heH/09LSkJCQIMwjIiIiIiKiilHoHrfr169h3x97yz2PiIiIiIiIyk+hwq004pwc6OjoKHuzREREREREtVa5ukqGh4fh8ePHCHg1iuSpUyfl5udkZ+PMmX/Ru3cfpQZJRERERERUm5WrcHv69Cl+3bULUmkupFIpnj17JjffwEAfDRs2Qteu3ZQZIxERERERUa1WrsKta9du6Nq1Gy6cP4eg4GBMmDBRVXFRLTB//nw4OjpiypQpVR0KEREREZFaU2hUyc5duqJzl65KDuXNcnJy8PLlC2RlZcHB3qHUxw5ERUUhMjICFhYWqFfPudhlXrx4gfT0NLi5uRe5Ly8pMREvXryAd9OmymxCrbNhw7eIj4/HsmVLi8yLiIiEnp5+mbazevVqpKdnYPHiRcoOkYiIiIhI7SlUuFWF3377FWf+PQ1LS0sYGRkjKOgpOnXqjIkT/wctbW1hOalUis2bN+H6tavw8PDE8+fPYGdnj9mz58DA0FBY7ueff8L1a9dgbm6O7OwsLF22AsbGxsL8TZs2olHjxrWmcMvKysLw4SPw9tsjMW7cOGF6To4Yw4YNw8CBgzBhwvvl3u60aVMxc+YsrF27Dp9//plCse3atQtPnvhh27atCq1PRERERFTdVZvC7fSpU/j440/Quk0bAMCz0FAsWDAPFpaWGDFipLDcieP/4OqVy1i5ag3s7e2RkZ6O2bO/wC+/7MBHH38CAIiMjMC/p0/hhx82wsLSEkuXLMaJE8eF7Vy9egXx8XEYMGBg5Te0iujp6eGzzz7DuHHj0LZtW9SvXx8AsH7Dt0hISMS4cWMV2q6GhgbWrl2DoKAghWPr0KEjxowZA01NTYW3QURERERUnSn9cQCqMm/+fKFoAwBnFxc0aNAQD+7fk1vu9OlTaNeuPezt7QEABoaG6N2nDy5duojMzEwAQHhYOKysrGBhaQkAaNCwIcJevgQAZGRkYNfOXzB58odqVShIcmU4+yQN288n4OyTNEhyZUrfh69vN7z33ruYOPF/yMnJwaVLl7B37z5s3boFWlpa2L79JwwePAQ9evTEggULkZqaKqw7f/58rF+/AYsXL0GvXr2xdetWfPzxx9i3bx8AwN3dHQDQuXMXPHz4UFgvKysLCxYsRI8ePfHOO+/gzp27wrwNG75Fy5atMGrUKAwePARbtmyBTKb8dhMRERERqbtqU7h5ejYoMi0xMREmJqbC64z0dISHhwtFQj6P+h6QSCR49iwUAGBkZITU1FTk5uYCAJKTkoRuknt/34PmLXzg4empqqaUm19EFgasf4bpeyLx3el4TN8TiQHrn8EvIkvp+1q0aBGk0lzMnj0HH344BXNmz0L9+vWxcuVKrFu3DpMnT8Ly5ctw7949jBgxUiikIiIisWLFCpiZmWHt2jUYNmwYXr4MQ1JSktz2g4KCkJX1Ou7NmzfDyMgQK1euRNOmTdGvXz88f/4cADB27Bjs2bMbv/32Kz799BNs3boNW7eyuyQRERER1T4KdZWUSqUQiaq25rt48QJevnyBwUOGCNMSkxIBAGZm8oOWmJrlFXeJiXnz63t4wMTEBNu2boGziwsuXbqI+fMXIigoCNevX8e69RuQlZmJ4JBgmJtbCFfviiMWiyEWi4XX+YWMVCqtUPs0NDSgoaEBSa4MM/ZEIiJJIjc/IkmCGXsicXSGM7Q0NSCTyZRyNUpLSwtbt25Ft26+6Nq1K95+ewSkUil++OFHfPvtBvTt2xcA8NNP29GoUWNcv34drVu3hkwmQ69evTB16qdy25NKZUVyIZVKIZVKIZPJ0LFjB3zxxRcAgObNm+HKlavYseMXfPnlQlhYWMDCwgIAUL9+fYjFYqxduw6TJk2qcDurWn5OKnqcUFHMreowt6rD3KoOc6s6zK1qVXV+q/q7PhWlUOF27tx/uHrlMnr26o2WPi0hquQuhSHBwdi6ZQtat2mDzp27CNNzc/MO7MIHWv7r/CtsmpqaWLR4KY7+/RdCQkLw+cxZcHd3x9y5szFu/Hikp6dj0ZcLYWVlhcjICAwYMBBDhr5VbCyHDh3Egf37hNd6erpYtWoV4mKjFG6flpY2rOrYAAAuBKQXKdryRSRJcDEwHd0aGkFDQwPxcTGQSMTFLlseFuYmsLaug1YtWwAAAvwfIzU1FQ72toiJjhCWs7aug3v37sC5niOyszPh5OgoNz8nJxtpqcly02QyKRIT4hATHYHs7Ey4ujjLzXd1cYa//xPEREfg+fMX2L79ZwQEBiI1NRXZWdnIys6SW766q8hxQqVjblWHuVUd5lZ1mFvVYW5Vq6rya2vnWCX7pZIpVLh51PfA9WtXsXbNGpibm6FHj57w7d5DuDqiSuHhYfjqq+Vwd3fH1KnT5eYZvho1MiMjXW56RkaG3HwAsLCwwPh33xNeHz36N4yNTdCxYyds27oFjRs3xiefTsXz588wd85s9OjZC0ZGRkXiGTr0LblBTGQyGVJTEmFVx1Ypv1SExOa8cX63hq/aZFmnwvvLp6mpBUMjEwCAq5sHtLS0kCvVgLVN3tVHiUSC5OQUuLnVh7WNPXR19WFkbCrMBwADA0Po6RsI03JycpCdnQNzCythnbT0TLl10tIz4ODgBGsbe/TpOxDdu3fHypUrYWpqilu3bmH+/AVyy1dXUqkUcbFRSjtO6DXmVnWYW9VhblWHuVUd5la1mF8qTKHCzdHJCXPnLUBMTAz+PX0KJ0+ewJ9/HkDLlq3Qq3dvNGniBQ0NDWXHipiYGCxfthR2dnaYPXtOkWevWVhYwNDQEOHh4XLT8187OjoVu934+HgcOngQK1Z8BQAIC3uJjp06AwDq1XOGlpYWoqKiitw7BwDa2trQLvQ4gtSURIhEIqV8yFzr6JR5vrI/1Pnvoa6uLnr37oV169bht99+hb6+PtatWwdDQ0O0bdsWIpFI6NpZMAYPj/o4e/Y/fPjhh9DS0sL33/+A3NxcITcaGho4evQoHj9+DC8vLzx8+BDHjh3D7t27IZVKERYWhv79+6Ndu3ZIS0vDsmXLoKFRsy7dK+s4oaKYW9VhblWHuVUd5lZ1mFvVYn4pX4WOAmtra4weMxabNm3Bx598ioSEBCxbugTTp03F8eP/ICen9KtF5ZGYmIhlS5fAzMwMc+ctgJ5+0Qc3a2hooFXrNrh27SokktfdCy9dughXVzdYW1sXu+2fftqGfv37w9bODgBgaGQkDKqRmZmJ7OwcGBsXvdpWGTp7GsLerPj62t5MC508DIudp2xr166FRCKBm5s7XF3d8Pvve7Fr1065Z98V9sknnyAsLAzu7vXh4eGJ2NgY6Bd63/JGsnwfDRs2QvfuPTBt2lT4+naDlpYW5s6di1GjRsHLyxve3k1hXsoD14mIiIiIajINiURc4REtYqKjcfr0Kfz331mIxWI0aeKFR48ewsTEFIsWL4GVlVWFti8WizFn9izExMRiwsSJcl0edXV10bRpM+F1fHw85s2dDRcXV3Tq3Bl+T57gv//+w5eLFhU7MuWtmzexe/dvWL16jfAg72vXrmLb1q0YPWYM7t+7l1c0Ll9RplilUilioiNgbWOvtF9H/CKyigxQYm+mhfWj7dDQXk8p+ygsNPQZzMxMkZ2VLteW1NRUZGZmFimCIyMjoaOjA8tXj1jIJ5PJEBkZCUtLS+jq6iIoKAiOjo7Q09OTWyc6OhpmZmbQ1dWVWz8tLQ2xsXGwt7eDTCZDeHg43NzcVNLmyqSK44TyMLeqw9yqDnOrOsyt6jC3qsX8UmEKF27S3FzcuXsHp0+dxL179+HgYI9evfugS5eu0NfXR3p6OtavWws3Nze8M3pMhYLMzMzE999/W+w8UxNTfPDhFLlpSYmJOHHyBCIjImBuYYEe3XvA0an4bpI///wT2rfvgAYN5Iu6a9eu4tatW7CwsMDgwUPkisXSqOpDJsmV4WJgOkJic+BaRwedPAyhpan87qgF8YShOsyt6jC3qsPcqg5zqzrMreowt6rF/FJhChVu/n5++PbbDUhMTEDLlq3Qp09fNPHyKrLcv6dPISg4GB8WKqxqspr0IatJbVE3zK3qMLeqw9yqDnOrOsyt6jC3qsX8UmEKDU6SkpKCzp07o1fvPkW6xhXUvn0HtGzZSuHgiIiIiIiISMHCrXWbNmjdps0blzMwNIRBGbsYEhERERERUfEUKtzy5ebm4vnz54iNjYGZmRnq1XOGnp5qBssgIiIiIiKqrRQu3Pz9/bFl80a5Z6aZmJhg/Ph30blLV2XERkRERERERFCwcEtLTcXKb76Ct3dTfPLpVNSpY43U1BRcvHABP/74I2xt7eDh6ansWImIiIiIiGolhQq3h48ewsrKCtOmzxBGuTExMcGod0YjKSkJ165fY+FGRERERESkJAqNLaqpqQk7++KHJnVwdISWpmaFA6PqYfr0GYiMjKzqMIiIiIiIajSFCjdXF1cEBgQgKSlJbnpubi5uXL8O9/r1lREbVQM7duxAYmJiVYdBRERERFSjKdRVUiwRw8bGFl/M+hxdu3ZDHWtrpKSk4Mrly0hPT0NGejrO/XcWAGBnbw9PzwZKDZpU58svF2HAgP5o3bp1VYdCRERERESvKFS4+fn5ISDAHwDw119HiszfvHmT8Levb3cWbtXI77//Dk9PDxZuRERERERqRKHCzde3O3x9uys7FiqFLFcC8aPLyI1+Dk2betBu0gEamhV6DF+Z5ObmYvfu3QgICISnpweaNWuGP/88iMWLFwnLZGRkYNu27Xjx4gV8fHwwZMhguW0cOXIEt2/fgampKQYNGoj67EpLRERERFQuCt3jRpVL8jIQKUtGIH3bHGT9tQnp2+YgZckISF4Gqnzfn06dgR9++BGWlpa4du06Ro58G3/88YfcMu+++x4CAwNgYmKCOXPmYO7cucK8iRP/hy+/XARTU1OEhYWhS5euOHfuvMrjJiIiIiKqSRS+ZJOTnY3Dhw/h8uXLiI2NgZmZGby8vPHOO6NhZm6uzBhrNVmuBOnbZkOaECU3XZoQhfRts2GyaL/KrrzduHEDV65cxYMH92FtbQ0gbxTJkydPyi03fvw4zJ49GwDQp09vdOnSFR999BGio6Nx5MgR3LlzG3Xr1gUAmJubY+HChbh48YJKYiYiIiIiqokUvuK2atVKHDt2DF5eXhj1zmi079ARjx49xNy5c5CRnq7MGGs18aPLRYq2fNKEKIgfX1HZvh8+fAQPj/qwsrISpnXp0qXIct26dRP+9vLygoWFBR4/foz79x/A09NDKNoAoHt3Xzx58gQSiURlcRMRERER1TQKXaoJCgpCcHAwVq9eA2sbG2H6yBEjsWDBfJy/cB59+/ZTWpC1WW7081LnS6OeAd6dVbJvHR1tiMViuWlicU6R5XJycoq81tbWgY6ONnJyxIXmiaGlpQVNPuuPiIiIiKjMFLriFhEehiZNmsgVbQCgo6uLjh07Ijw8XCnBEaBpU6/U+SJbZ5Xtu1WrVnj69Cn8/fNGEJXJZDhw4M8iy+3ff0D4+9SpUxCLxWja1ButW7dGcHAwbt++Lcz//fc9aNOmDTQ0NFQWNxERERFRTaPQFTc9fX0kJMQXOy8uPh76enoVCope027SASIL22K7S4osbKHduL3K9t2gQQO899676Nu3H3r06IHg4GCIRCKIRPL1/qNHj9CvXz/UqWON06dPY+nSJbCysoKVlRXmzp2DoUOHonv3HggLC0N4eDgOHNivspiJiIiIiGoihQq3Rg0b4Yfvv8Ovu3Zi6FvDYGRkBLFYjPPnz+H0qVP4ctGiN2+EykRDUwuGk1YWGaBEZGELw0krlT4wydKlS9CyZUvh9czPZ2D06DEICgqCh4cHHj16hM2bNwvz161bh6FDhyIwMAAvXrzA/Pnz4OHhIcyfNWsWhg4dinv37sHExATt2rWDsbGxUmMmIiIiIqrpNCQSsUyRFW/cuI7NmzYiLS0NBgYGyMzMgkgkwogRI/DWsOHKjrPakEqliImOgLWNfZErUxUhy5VA/PgKpFHPILJ1hnbj9ip/jptUKsXZM6fRzbcHNDU1IZFIMHjwYDRs2Ahr1qxW6b5rOlUdJ8TcqhJzqzrMreowt6rD3KoW80uFKfzNv3XrNmjYsBHu37uHmFePA2jUqDFsbW2VGR+9oqGpBR3vziobiKQkjx49xoKFi+Hl5YV79+5BU1MTn3/+WaXGQERERERU25WrcIuMjMDLly9Rr1492NjYwtjYGB07dQIApKWl4cmTx0hLTYV7/foqCZYq36hRI9Gv/wDcu3cPY8aMQbt2baGrq1vVYRERERER1SrlKtw2bdwIfX09zJw1u8g8IyMjBAUF4fSpU9i0eQv0OEBJjeHu7i533xoREREREVWuMneYjYyMwLNnofjk02nQ1tYudpl33hkNa2tr3LhxXWkBEhERERER1XZlLtzCw8PRxMur1BEBNTQ00LZdO0RERCglOCIiIiIiIipH4ZaRkQE9Pf03Lqenp4f0tLQKBUVERERERESvlblws7K0QmhIyBuXCw0JgVWdOhUKioiIiIiIiF4rc+FW38MDKSkp+O+/syUuExoagsuXL8OnhY9SgiMiIiIiIqJyjCqpra2NMWPHYvOmTXgaGIiOnTrB3i7vgYBx8fG4c/sW/vrrCLp16wZHJydVxkxERERERFSrlOtxAL6+3SHNzcXu3b/h339Py83T1NRC3379MGbMWKUGSJXr2LFjaNy4MZydneWmHzp0GG3atIa9vX3VBEZEREREVIuVq3ADgB49e6FDx054+PABoiIjIZVKYWllBa8mXjAzN1dFjFSJpk+fgcWLFxUp3D744AP8/PNPLNyIiIiIiKpAuQs3ANDX10fr1m2UHQuVQiLNxZmYQASlxsHd2ArdrT2gJdKssnj++ecfZGRkQltbC66urvDy8hLm5eTk4K+//sagQQOho6MDAIiMjMS9e/fQt29fYZnr168jKysLLVq0gKWlZZm2TURERERUGylUuFHlepwcick39yIsM0mY5qhvhq2tRqGxqV2VxHTmzFkkJCQgJycHd+7cQYsWzfHbb79BQ0MDqampmDhxIkJCgoWC7P79+/jiiy/Qt29fREZGomfPnrCwsISdnS38/QPwzTdfC0VdadsmIiIiIqqNWLipOYk0t0jRBgBhmUmYfHMvzvtOVfqVt1u3bkNXVw8ymRQpyYkwMTWHVCqVW2bt2jXC35mZmejSpSuOHTuGAQMGvHH7//77L+rUsRZGKE1PT8ft27eVsm0iIiIiopqIhZuaOxMTWKRoyxeWmYSzMU/Ry7aBUvd5//59JCUlQSaTITs7E7q6+sjNzS2y3L179xAWFoasrGzY2dnh3r17ZSqufHx8sGTJUmzatAm+vr7w8PBA586dlbJtIiIiIqKaiIWbmgtKjXvD/FilF24TJ07AmDFjIJVKERMdAWsbe9ja/iPMT05OxqBBg5GcnIwGDTyhr2+Aly9fwtXVFQCK7dJY8Ipdo0aNcOHCeRw9ehSrVq3G3bt38eWXX2LIkMFv3DYRERERUW3Ewk3NuRtbvWF+nUqK5LV//z2DlJQU3LlzGyJR3jPcBw4cBJlMBgAwNjaGpqYmoqOjhXvcHjx4IKwfHR0Ne3t7TJ48GZMnT8ahQ4cxd+5cDBky+I3bJiIiIiKqjVi4qbnu1h5w1Dcrtruko74ZfK3rV3pMbm6uCA8Px8qVK+Hk5IQzZ87i/v17cHNzA5D3sPbu3btj2rTpGD9+HIKDQ7Bv3z5oauYVYqdOncLu3XswYEB/GBub4Ndff0WLFs3LtG0iIiIiotqIhZua0xJpYmurUSWOKqnsgUn69++PevXqFZk+dOgQ2Ns7AACaNWuGffv24ciRIwgLC0ffvn3Qs2cPSCQSYfmfftqOLVu24Pr1G/DyaoKdO3/Bvn37AADjxo2Dp6cnjhw5goCAQIwaNQqjR79T5m0TEREREdU2GhKJmH3QlKjgfWH5Xf2UQSLNxdmYpwhKjYW7cR34WtdX+XPcVNUWYm5ViblVHeZWdZhb1WFuVYe5VS3mlwrjFbdqQkukiV62DZQ+EAkREREREak/lu9ERERERERqjoUbERERERGRmmPhRkREREREpOZYuBEREREREak5Fm5ERERERERqjoUbERERERGRmmPhRpUiOjqaD9Gu5hISEpCVlVXVYRARERHVSizcSCExMTHlKsQ8PDwRGBiowohI1QYNGoQjR45UdRhEREREtRILN5ITFxeHzMzMItNjYmKQnZ0tvO7evTv8/PwqMzRSUGpqKpKTk6s6DCIiIiKqABZuJKddu/Y4ePBgkelNmnjh9OnTwuuHDx/Cy8urMkMjBX3zzUrMmTOnqsMgIiIiogpg4UYKKamrZHp6OgAgJycHcXFxxa6blpZWpn0UvC8uMTEJubm5cvuRyWTFrpeZmQmxWFzidovbf1ZWFhISEuSmpaamIiUlBYB8e2QyWZFl89td1v2VpqRtSaVSZGRkFJleMM6srCy59yUzM/PVf1mIjo5GdHR0kXVycnKEvwEgOzu71PwVjik1NbVsDSMiIiIihdXIwi0gwB9ff7Ucn37yERZ9uQBXrlwuskxgQAAWL/oSs2Z+hr/+KnrfzvHj/2DTxh8rI9wykUklEL88gqzHKyF+eQQyadUO9NGkiRcePHgovD537jwaN24Cd/f6aNKkCRYvXoxOnTrLrXPo0CE0bNgIHh6e8PLywu3bt0vdh4eHJ77++hs0bNgIrVq1wrNnz/HXX3+hWbPm8PDwRN269bBo0WJIpVIAwNOnT9G5cxc4O7vAzc0dH374oVxRsW3bdri5ucPDwxOurm7YsmWLMO/IkSMYOvQtuf0vWbIUX365CABw4cIFtG/fAXPmzIGbmzveemsYxGIxli9fDmdnF9Sv74H27dvj4cO8nIjFYqxYsQL16jmjQYOGaNiwEf76668S21rattLT0zF16jQ4OjrBzc0dbdq0xdWrV4V1V6xYgf/9bxKGDn0Lnp4NYGdnj/nzFwAADh8+jN9//x3Hjx9Hx46d0LFjJ2RnZwvrDBo0GJ6eDbBmzVoEBwejX79+cHZ2Qd269TBixEhERUWVGPP8+Qvg4OCIxo0bo2XLVrh06VKp7ycRERERKa7GFW4hISFYumQJnJ1dMH/+QnTp0hXff/cdLl28KCyTmZmJlSu/Rtu2bTFp0gf4+6+/cO3a6y/CcXFx+PPAfgx9a1hVNKGI3IR7SP2rATIuDEf2vQXIuDAcqX81QG7CPZXsLyUlRbg6ExsbJ1ylKUlCQiLGjx+HGTNmICIiHMePH8fx4yeKLHf37j1cu3YV4eFh6N27DxYsWPDGWK5du4orV64gKOgpkpISMWXKR9iwYQPCw8Nw48Z1nDhxAr/8shMAsHr1ajRq1BDh4WEIDg6Suw/vzJmzWLBgAXbt2omIiHDs2bMbixYtxsmTJ8ucl+joaOjrG+Dp00CcO/cf1qxZgz//PIgTJ44jIiIc27Ztw40bNwEAGzZswD///IMLF84jLOwltm/fho8++hjBwcHFbru0ba1YsQK3bt3E7du3EB4ehnfeeQejRr2D+Ph4Yf3z589j5szP8fz5M5w79x+2bduGe/fu4Z133sGECRPw1ltD8fRpIJ4+DYSuru6rnJzBtGnTEBoagkWLvsS4cePg4uKC58+f4enTvIFkPvjgw2LjvXv3Hnbs2IGrV6/gxYsX2LdvH27dKr0QJyIiIiLF1bjCbd++vXBxccE7o8fA1s4Ovt17wNfXF7//vlvoWhcU9BSampro07cfPDw90alTJ9y5c0fYxs8/bUe//gNga2tbVc0QyKQSpF8YDln6c/np6c/zpqvgyttXX32Njh07oXPnLhgydBg6d+4iNzBJYWfPnoW5uQX+97+J0NDQgJOTEz76aEqR5T77bAZMTU2hoaGBIUMGw8/PH0Bet738QrFw98rPP58Jc3MzAMCuXb+iT58+aNiwAaKjoyESiTBq1Cj8/fffAIA6deogKSkJCQkJ0NbWxogRI9C6dWsAwMGDf2Lo0CHo0KEDAKBt27YYNmwY9u8/UOa86OjoYP78edDU1AQA/Pbbbsye/QUaNGgAAGjcuDEmTpwAANi5cycmTZoMPT09REdHw93dHa1bt8aJE0UL2jdt688/D+Kzzz6DnZ0dRCIRpk2bCkNDQ7l7Dnv27Cm0rXHjxvDw8IC/v3+p7fH19UX37r4AAD8/Pzx+/ATLl6+Ajo4OjIyMsHjxIpw7d67YLq/m5mYQiUR48eIFpFIpXF1dMH36tDLnkoiIiIjKp0YVbrm5uXj44CFatGghN93HpyViY2MRFvYSQF63NG1tHWG+jq4uJK/u6blx4zoiIyMxaNDgygu8FJLwY0WKtnyy9OeQhP+j9H1+883XePo0EAEB/rh86TwCAvyFqzTFiYuLhZ2dndy0wq8BwMzMTPhbV1cXOTk5AICNGzcK3fiGDx8ht46trY3wd0REBE6cOCEs27FjJ2zcuBFZWXmjYC5cuBAdOnTApEmT0blzFyxYsFDoKhkdHQMHB0e5bTs42L/xamJBVlZW0NLSApB3n1tUVBTq1q1bZDmZTIbIyCgsWbJELtaHDx/K3UtWcPnSthUbGwsHBwdhmoaGBmxtbREdHSNMK5hbIC+/2dk5pbbHxuZ1bqOjY2BoaCgUyQBgb+/wal7RHDk7O2P//n3YvXs3unbthpEj35brvklEREREyqVV1QEoU3xcHMTiHNgUulJm/eoLamRkJJyc6sLFxRVJSYl4+jQQ9erWw+1bt+DbvTuyMjPxy46f8enUacIX9DcRi8VyAznkX9XLv+9KUSJRXk2dm1L6VZPcFH9oY5BS9plPJpNBKpUK28v/f/7018vlLePo6IiQkBDk5OQIefP3DwAgv3xx25RKpZg1axZmzZolt1xx67i4OMPCwgKbNm0sNl5tbW18/PHH+Pjjj5GVlYVx48bj66+/xvLly+Hk5AR/f3+5bfv5+aFuXSdIpVLo6+sjIyNdbn54eDisra0LxCDfHmdnZzx48ABt27aVi0VDQwPOzs6YPXs2hg+X725bXG5FIlGp23J0dISfnx/atGkDIG/wkNDQUDg5OUIqlb465mSF3n+Z8P6IRCJIpYXfO/l1nJwckZ6ejufPX8DJKa/AffLkCTQ0NODg4PBqPxC2I5PJ0KZNGyGmf/45jmHDhiMo6Cn09PRQlQofY6Q8zK3qMLeqw9yqDnOrWlWd3/zvoqQ+alThlpWVBQDQ1ZG/OqT36mpR/nxzc3O89/4ELFu6BADg7d0UPXr0xO7dv8HLyxueHp7Yu/d33LxxHRYWFnj33ffh6ORU7D4PHTqIA/v3vd6Xni5WrVqFuNiSB3V4Ey0tbVjVySs2NU0alLpswfkJ8bGQSMo2GmBJpNJcpCQnISY6QpgWFxsFmUyG5KQEYbpMJkNCfCxioiPg7dUImpp5XfjGjxuL0NBn+OGHH6CjoyO3nfzlASAxIQ4ymVRufnEKrjN82BAMfWsknJzs4dutG7Kzs3Hu/HmIc8SYOvUTfD5zNtq2aYWWLVsiMzMTERHhcHWph5joCAwe1A/DR4zCurVr0KlTR1y6dBnHjx/H3t93IyY6Ag72tggNDcXWLZvg49MCFy9dxqlTpzBs2FDEREcgKSkBubny8b733jh8/dXX0BRpwNvbC/fu38fzZ88xdeon+N/E97FgwQJkZqShUeOGiIqMwoEDBzF48EC0b99OLrdv2taot0fg66+/gbGRIewd7LBt60/Q19NDi+beiImOQEZGOjIzM+RiE4vFSE3Jex9NTYxw6tRJ3Lp5Dfr6BqhTx6rIOibGBmjfvi0mT56EuXNmITs7Bwu/XIwhgwchKzMNWZlpkEjESElJREx0BP49cxbnz13A0KGDYW1tjZs3r0FDQwPxcdHQ1tZW6NhTtop8Bql0zK3qMLeqw9yqDnOrWlWVX1s7xzcvRJWqRhVu2jp5XxgLD1MvflXM6BToHtmzZy/4+naHOCcHevr6CA0NwZXLl7Bu3QacOn0K165ewfTpn+HGzRtYvXol1q3/Vri3qaChQ9/CgAEDhdcymQypKYmwqmNboV8q8q+2aDn0h4ZhvWK7S2oY1oOWQz9heQvLOgrvL5+1tQ2sbWxhbWMPqVSKuNgoWNWxhY2NDaxt7GBtYw8Ar17bC6//+usvzJ+/AB99PBUeHh6YOnUqdu3aJcy3traGje3r5W1sY1/ty76UWOTXsbaxx7//nsaqVatx5Mhn0Nc3QK9ePTFz5hcwNDTEypUrsWrVauzctRtaWlro3t0X8+bNg76+Pqxt7HHw4EGsW7cOv/62B/Xq1cP+/fvRpUsXYds//rgRP/zwA3b8sgu+vr6YPHkStLS0YW1jDxsbO6HN+T7++FPY2trjl192YuOmLWjdujWWL18GS0tLfPDhFDjVrYft27dj3fpv4ejoiLFjx2LwkLegoaEhl1uRSFTqtmbPmQsLSyts3rIViYlJaNasGf45fhx16zq9ei9skZGRIRdbfm6tbewx8X+T8ejxE/xv0hRkZmbiwYP7xa6zZ89erFixArPnLICWlhb69u2H2bO/EK6g2djYwubVMTBq1GhIpcCGb39AVFQUPDw88OefB+DgWK9iB6ASFM4tKQ9zqzrMreowt6rD3KoW80uFaUgk4uIfhlUN5WRnY9y4sRg3frxcMfXo0UMsXbIYX329Eu7u7kXWk0qlmD9vLvr07YsuXbriqxXL4eXtjYEDB0EikWDc2NFYvXptiVfdCm8rJjoC1jb2SvuQ5SbcKzJAiYZhPRh2PgBNi2ZK2UdxytOW/EIz3/Lly3H79h0cOlT0Yd6kmuOE8jC3qsPcqg5zqzrMreowt6rF/FJhNeqKm46uLtzd3fHkyWO5wu3J48cwMDCAc73irwacPHkCenp66NKlK4D8wUvyrt5paWlBU1NTuGpXFTQtmsF4kD8k4f8gN8UfmiYNoOXQDxoi9Xn7Vq1ahfr1PdC0aVPcunUTmzdvwYYN66s6LCIiIiKiGqHGle+DBw/B7Vu3cetW3jOwXjx/juPHj6Nf/wHQKubem4SEBPx5YD8mTZosTKtfvz5uXL+O3Nxc3Lp5E9raOsIIe1VFQ6QFbadB0Gv8BbSdBqlV0QYA48aNw/HjxzF8+HDs2PEL1q5dg+HDh1d1WERERERENYJ6fftXgtZt2uD9CROwaeOPAICcHDF69OiB4cOKLyJ+2fEzevXuA/sCw60PGjQYa9euxvvvjYemphY+nDKl1OHwCbC3t8e2bVurOgwiIiIiohqpxhVuANCnT1/06tUbaampMDA0LHVo//feex+mhZ6BZWRsjEWLlyIzMxO6urrsV0xERERERFWqRhZuQN6zJ0xMTd+4nIWlZYnz9PX1lRkSERERERGRQngpiYiIiIiISM2xcCMiIiIiIlJzLNyIiIiIiIjUHAs3IiIiIiIiNcfCjYiIiIiISM2xcCMiIiIiIlJzLNyIiIiIiIjUXI19jltVkclkAACpVFrFkVRcfhtqQlvUDXOrOsyt6jC3qsPcqg5zqzrMrWqpQ341NDSgoaFRZfsneRoSiVhW1UHUJBKJBHGxUVUdBhERERFRhVjb2EMkYgc9dcHCTcmkUimkUmmN+IUiMzMTUz6chE2bt0FfX7+qw6lRmFvVYW5Vh7lVHeZWdZhb1WFuVUsd8lsTvs/WJOwqqWQikajG/DKhoaGBrKxsaGho1Jg2qQvmVnWYW9VhblWHuVUd5lZ1mFvVYn6pMB4FREREREREao6FGxERERERkZpj4UYl0tbWxvARI6GtrV3VodQ4zK3qMLeqw9yqDnOrOsyt6jC3qsX8UmEcnISIiIiIiEjN8YobERERERGRmmPhRkREREREpOZYuBEREREREak5PsetlkpNTcXNG9eRkpICp7p10bx5izI/IyQrKwvXr19DdHQ0OnToAAcHRxVHW71kpKfjxs0bSEpMhL29A3xatoSmpmaZ1g0MCEBwSDBEIhHc3d3h5uau4mirl6ysLNy8eQPxcXGwsbVFy5atynTT9onj/yAlNRUAYKCvD3sHBzRr2gyiMr4vtUFOTg5u3byJ6Jho1KlTB61atYaurm65tpGWmorjJ45DX08PAwYOUlGk1Y9EIsHt27cQEREBCwsLtGrVGgYGBm9c7/79ewgICCgyvWvXbrC2tlZFqNWONDcXd+7eQdjLlzAxNUXrVq1hZGxc5vVTkpNx+85tpKamwqO+Bxo0bKjCaKsXmUyG+/fu4dnzZzAyMkKrVq1hamr6xvVOnTqJpKSkYud1794DlpaWSo60enr06CGCg4Ohr6eHli1bwaKMeck/5iMiIqCvr48GDRrAyamuiqMldcErbrXQ8+fPMH3aVFy+fBlpaWn4+aft+GrFckjE4jeuu/f3PZj66ce4cvkSDuzfh/Dw8EqIuPqIiozEjBnT8d/ZM0hLT8fu3b9i0ZcLkZWVVep68fHxmDXzc6xc+TUiwsPxLDQUSxYvwrcb1kOam1tJ0au3hPh4zJr5GY7/8w/S09NxYP8+zJs7B2lpaeXaTmxsLLZt3YpZs2Yi9VUxV9ulpKRgzuxZOHToIDLS0/H3X3/hi1kzkZSYWK7tbN++DQf/PICjR/9WUaTVT1ZmJhZ9uQC/79mNjPR0/Hv6ND7/bDqio6PeuO79+/dx5t/TlRBl9SQWi7F8+TLs+PlnpKWn49LFi5gxYxpevnxRpvWvXb2KTz/9BLdu3kRaaioOHNiP7du3qTjq6iE3NxerV63E5s0bkZqSgps3bmD6tE8R9PSpQtu7fu0qDh38s8w/YtZkMpkMP/7wPdavW4ekxEQ8fPgQ06ZNxYMH99+4bkZ6Or74YiZ2/PwzkhIT4efnh9lfzMKBA/srIXJSB7ziVgtt2rgR7u7umDtvPgCgX/8BmDb1U5w4cfyNv5I3aNgQg4cMRUJCPO7evVsZ4VYr27ZtgbW1NRYtXgqRSIQhg4dg2rRPceTwIbw96p0S10tKSoSRkSG+XLQYxq9+Le7eoyfmzZ2Npk2boms338pqgtraseNn6OnpYemy5dDS0sKwYcMxY8Z07Nv3ByZMmFjqun369pN7PWLESEyb9in279/3xnVrgz27f4NYLMY336yCjq4uRorF+GLW5/jtt1/xyadTy7SNK5cvIyDAHx06dMSjRw9VHHH18eefBxATE4P1G76DkZERpLm5WLhwPn7+aTvmzlvwxvXNzS0wcuTblRBp9XPs6N8IDg7Gt99+BzNzc8hkMny1Yhm2bN6M5Su+KnXd58+f4dtvN+Djjz9Bx06dhOmhoSGqDrtaOHPmX9y/fw/r138LaxsbAMC6dWuwceMPWLtuAzQ0NEpct1ev3nKvJRIJTp08gWbNmsPMzEyVYVcL169dw4UL57Fy5Wo4u7gAALZt3YJNG3/E99//CK1SepEcPfo3oqNj8MOPG4Wrn66urvh110506+bLq5m1AK+41TLh4WEICQlGz169hGkWFhZo2bIlLly48Mb1mzVrDn19fVWGWG0lJibi0aNH6Nmzp9Dt1MjYGO3atX9jbuvUscacufOFog0A3N3dYWJigpAQfpHIzMzErVu34Nu9B7S08n5v0tPXR6fOnXHp4gVIpdJybc/I2Bh169ZFeNhLVYRbrUgkEly5chldunaDzquukdra2ujazRfXrl1FTnb2G7eRmJiI7du3YtLkD3h+KOTixQto374DjIyMAAAiTU1079ET9+7dR3JychVHV71dvHgBrVq3hpm5OQBAQ0MDPXr2QmBgAKIiI0td9/ChQ6hbt65c0QYALi6uKou3Orl44QKaNm0mFG0A0Ktnb4SFhSG0nP8m3blzGykpKejmyx8gAeDCxfPw8PAQijYA6NW7N+Lj4/Ho8aNS142KioKNjbVcl9UGng0gk8kQFVX6MU81Awu3WiY0JBQA4FzPWW56PWdnvHz5AhKJpAqiqhlCQ0Mgk8lQz9lZbno9Z2fExsaU2qXPxMSkyP1E8fHxSE1N470sAF68eI7cXAnq1asnN925njPS0tIQFxdXru1lZGTg5cuXvIcQed17s7Kyip4T6tVDTk4OIiIi3riNzZv+396dh0V53Qsc/w6DbGEREMMuoAIKRIOCLIoaDZhFk2brbZY2sdl7mzSLJhXBrU16mxiztUmeJO29TdJUzGobo+CCAkZQEYi44AiMooDLRGZABGeY+wcyOgzDYhiF8Ps8j0+YlzNnfu85b17mvGf7C9dfH0ts7CQbRTk4NTQ0oNFoLO4JIaNCMBrbUKure8yjsVHHf/69lm+++Q8lJXv6/JDip6q1tZWammOEdLonjLpwHVdXV3f7/rKyMmJiYjhy5Ahr135NdvYGjqjVNop28KmurrK433Zcx1XVVX3Ka2tuLh4eHnJ/uKCqssp0nXYICgxCqVRSXdV92Y4eM4b6+hNmD30OHDyAo6OjzHMbImSo5BCj1WkBTE9/O7i6umIwGDjb1IR7LyYfC0sd86VcXc0nxneUtU6ntSh3a4xGIx9++D5OTo4kJU/t30AHIZ22+7LVarU9NnDramvZlreN5rNn2bOnmNjYSdx51922CXgQsX5PaC9rrVbb7ftzcrKprKzktVVv2CbAQUyntX6/hZ7LFtp7P0+cOIFer+eLzz/H29ub5xcsHPIPdBobGzEa27opW+u9ma2treh0Wg4cOEBBQT7x8VPQNTby97/9jbvuvps777zLprEPdC0tLbS0tFgs8tKX67aDVquluLiYm26+2TRaYqjr6ruAnVKJs7Nzj2WbljaH+vp6li7J4PrrY2loaEClUvHMs8/h7u5uy7DFACH/Fw01RmP7f7sZny4uj/FC2fZHyX766T/ZvWs3zzzzLF5eXv2Q4+Bm5ELZ9kPhGgGlUsnRo0c5eeIEgUFBPz7TwezCLeFyCre+vo6P/vF/PP7Ek2bDfEW7i9ft5V24N8y8gQce+KXp/b+49z5eWPg87737DhmZS/otzkHpR/wt67hXV1VVsWrV66bhgJEREXz44YfExcUTHDx0ey/6829Zfn4eBoOeG2SetonRyGX/MdM2NFB7/Bj29sNAocDe3p6WlnNUV1VJj+YQIQ23Icbdvb03rVGnw8nJyXS8sbERpVKJyzXXXK3QBr2Op126Rp3Zsr4dQyTd3Hr3NGzt2q/56ssvePTRx0hITOz/QAchU9l2WgWyo2zde9Fo8PXzMy3y0GYwsHz5Mt588w3+/Mqr/Rzt4NJRto0WZdv+2s3detnmbtnCsGHDqKmpIStrNQAqlYrm5mayslYzYcIEIiIibRT5wNdxv7V+3XZ/T+j8UMHNzY2U6TP48osv0Ov1Q7oHw9XNDYXCrovr9sL9tpveB0dHRxwdnQgLCzObw5WUlMwHH7zPwYMHhnTDzcnJCUdHR6vXbV8e0mzN3cLYsWPlAdkl3N3dLa7bNoOB5ubmbq9bgPfee4eGBi1/+p8/m1bo3L9/H0syMwgeNYrJk+NsFrcYGGSO2xATGtY+Gba609wKdXU1QUHBQ/qLwI8VGhqGQqFA3Wluhbq6Gh+fkb0aJrkxJ5tPPv6IBx+az+wbU3tMP1QEBQWjVNqj7jQHpWN/oRE+Pn3Kz06pJDomBrVabXq6PFT5+vnh5ORkeU9Qqxk2zIEA/wCr742KiiZtzk02jnDw8vDwwMvLy+KeUK2uRqGws5j71itD/Hrt4ODgQGBgANWd7gkd8wZDQ0K7eNdFYWFd/P5CL8jl9pD+lISEhFrcbzuu47BeLuByRK2mqqqKmdLbZiY0LNRifuvRmqMYDAZCQ7u/bisrq4iMjDTbViEychxKpb0sZDZESMNtiAkICCQsbDQ52dmmYxqNhl27dpGSkmKWdsOG9WzbtvVKhzhoeXp6Eh0dTU5OjmkBgUadju++225Rtps3b2LTpo1mx/Lz8nj//ff5xb33cfPNt1yxuAcDFxcXJk+ezOZNG00L6JxrbiZv2zamTksx2zw+Py+P9d+uM73W6XTU1povsKHX6ynZs4fg4OAh/yXN3t6epKRktuZuMa0gef78eXK3bCYxMdG00iRAYeEO1q792vQ6OiaGe+75udm/MWPG4OzszD33/HxI97Z1mDYthe3bC2hqagLan6xv2pjDxIkTzFaGK9mzh8/WZJleG41GDh2qMMtLp9ORl7eN6OgoechGe9nuLCoy7TdoNBrZmJNNeHgEvn5+pnTl5XvJylpttlfp9BkzUalUnKivNx3bXpCPUqlknGzCzbSUFEpLS8zKJztnA4GBgYSGXWy4VVRUkJW1mrNnz1rkkZu7BUdHmafdWcq06VRUVJgtRJK9YQNeXl5ER0WbjqnV1WRlrTbbTzMwMIADBw5guGR/1wMH9mMw6AkMDLwyJyCuKrnzD0FPPPkky5ctY8XyZYSGhvLdd9sJDw9nTqcn59kb1uPl5U1KynTTseLi3ahUKtOk++0FBVRXV+Pr62uWbqh65JHHWLIkk2VLMwkPj6CoqBA/P39uu/1nZum2bN5MW5uBWbNmA+036L/85S18fEbQ2tpqGnYGEBgYSFJS8hU9j4HowYfmsyRzMZkZi4mKiqK4uBhXV1eLPa7y8/Ooq6s17d1mMBhY9dpreHl7EeAfQEtrK6UlJRiNbTz73IKrcSoDzr333c/SJRmkpy9i4sSJlJWV0dbWxv33P2CWrrCwkO/Lypg377arFOngc8edd1FeXk76oheJi4vnwIEDaDQannr6d2bpSkr2sH79t9x19z1AeyPko3/8A6VSyaiQEFpbWigqKsTb25tHH3v8KpzJwHPzLbdSWlrK4sWLSEhMoqryMGq1mswly8zSlZeX89maLObNnWfaI2vGjJmUlZaSnr6IpKQkdI2NFO7YwQMP/JKAAPkCPGvWbPYUF5OZuZjk5KnU1NRQUXGQ9PQMs4ddhw5V8NmaLGbNmo2Li4vpuMFgIC8vjylTEsyOC5iSkMC0lOmsWLGclJQUTp06RUlJCQsWLDTbw02tVvPZmizi4y5uefGrB+fzhxXLePGFBVx33QQaGxvZvn07cXHxJCTI1IqhQKHXn5dxF0OQTqejqKgQnVZLYFAQsbGTzHotoL3HzdnZucuGW2fScLvobFMThUWFnDlzBn9/fyZPjjMb1gDtPW5Go9HUcOtY8bAr0nC76FxzM0VFhZzWaLh25LXExcczrNNmpfl5eTQ26sw23W5ra6O8fC/q6mrshw3Dz8+PmOgY7DrVy1DW2trKzp1FnDhxAp8RPsTFx1tsUVFYuIP6+vpuG27Fxbs5fuwYt86dZ+uQBw29Xs+uXTupra3F09OT+PgpFl9mS/bsQaU6ZGq4dag4eBDVYRV2dnYEBgQSFR095HuJL9VmMFBcXMzRmqN4uLsTHz/FYjXE8vK9lJeXc8fP7rDY3Hjv99+jOqzC1dWV8eOj8Pf3v5LhD2hGo5GSkj2oq9uHpMfFTzHrJYb2HreSkj3ceutcs2tao9GwcWMOCQmJQ3q+YHc6rj1nZ2cmT46z2Dxbra6msLCQ1BtTTQ03gHPnzlFcvJtTJ0/i6OREaEgo4RERVzp8cZVIw00IIYQQQgghBjiZ4yaEEEIIIYQQA5w03IQQQgghhBBigJOGmxBCCCGEEEIMcNJwE0IIIYQQQogBThpuQgghhBBCCDHAScNNCCGEEEIIIQY4abgJIYQQQgghxAAnDTchhBBCCCGEGOCk4SaEEP3g6NEjrHz1FY4dq+k23bvvvkPx7t1XKKp2KpWKla++Qmtr6xX9XFvbv38fK1995WqH0W9+qvUkhBCif0jDTQgh+oFWq6WwcAc6ra7bdLt37aK2rtYmMVRUVLDy1VfQ6/VmxzWa0xQW7qDNYOg23WBz8uRJCgt3XO0w+k3neuoNa3X5U6ljIYQQF0nDTQghfiI0py988W9rMzs+ZsxYnn32eRwcHLpNJwYfa3UpdSyEED899lc7ACGE+KnSNjTw7bfrqKurw9PLi7S0tC7T6c+fJ78gn/379mEwGBg3fjwzZsxEqVQC7UMC133zDY888iibNm/i+LFjeAwfTmpqGiNHjgTah9n9+99rAXjzjddRKBRc6+vL/fc/gEajoaAgn9hJkzhSVdVlunlz5/HBB+9z+8/uICwszCy+v334ASEhIdwwa3aX8XfE98Avf0V+3jaO19Yy3MOD1LQ5eHp6smXzJioOHcLF2ZnpM2YwevQY03tra4/zz08+AUCpVOLt7c3kuDjGjRtvkf+vH36E9d+uo76+nuSpU7uMpbx8LxvWr+fmW24lMjISo9HIzp1FlJWV0drSYjoPJycnq/XW8Xnzf/0wmzZtpK62Fk9PT9Lm3MSIESMuq+46xz55cpzVz+9t/tbqPCEh0eq1cCViFkIIYRvS4yaEEDbQ1NTEiy8upKyslKioKFycnVmcns7Zs01m6VpaWliyJJNPPv6YgIAAxoaH8/VXX7Jq1UpTmo4hgYsXL0Kv1xMVHc3RI2peWLgAbUMDAF5eXkRERgKQkJhIcvJUJk6YCJgPwbOWzt3DA41Gw7pv/mMWX+Xhw6xf/y3+/gFWz7UjvuXLlgAQFRXF3r17SV/0e1YsX0ZlVRXR0dGca2lhcXq62TxAN1c3kpOnkpw8lclxcdjb2/Onl18iO3uDRf7pi36PQqEgLi6+y3i2bs3lTy+/xHUTJhB54Ry/+OJz3nv3XXx8fIiKjubU6dOsWL60u6ozfd6LLyygtaWFcePHo1KpeGHhAk6dOnVZdddT7F3pKX9rddndtWDrmIUQQtiO9LgJIYQNfP31VxgMBpYsWYaDoyMAvr5+vPnm6xbpKisreXXlSgICAgGIjY3lqd/+lu+/LyMm5jpT2gcfnE/spEkAJCUm8fDD8ynaWcTs2Tfi5eVF+NhwAOLjp5iGRXbWXbq0OXN4952/8qsHH8LNzQ2A7OwNBAUFEzluXI/nPP/XDxMb2x5faEgoCxc+T1xcHI8+9jgA06fPoHzvXvLz8vj5f/0CAFc3NxISE83y8R4xgs/WZJGaat5Dee9995GcfLGnTaU6ZPp59b8+Zd26dTy/YCETLjRSAHbtLCI1NZXbbrvddKyjsduT2267nZtvuRWAmTNm8txzz7AmazVPPPkboG911zn23uhN/tbq0tpxW8cshBDCdqTHTQghbGD/vnImTZpsarRBe++HQmF+2y3evZvo6CjTl2gAH5+RREREWKw+GR0TY/rZwdERb+8RnL6kB+jHSkxIxMXFhS1bNgNwtqmJgoJ8Uq0M8ewsOvpifCOvvRaAqOho0zGFQsHIkSM5rdGYvU+trubTf37C22+/xcpXX2F7QQFnzpyhqcm8d/L6idd3+blvvvE6ubm5rFjxB7NGG8Do0WPIzc1ly5bNpt4ydw+PXp1PYlKy6Wc7pZK4+Cns27fPdKwvdWct9u70JX9b5Hk5MQshhLAd6XETQggb0Gp1XHPNNWbH7O3tLeZWNTbqaGpqtFjWvr6+zuL9nXvR7OwUGI3GfovZftgwZs2azcacbObOnUfu1lwUCgXTpqX06v2XxmenUAAwbNgwy5gvWTCjsHAHr696jRtumMX4ceNwdnGhpqaGffvKOXfunFkZuHQqjw5Hjqjx8vLCu9P8M4AHH5pPcHAw2wsK+N+//x13dzfmzp1HatqcHs+nc/m7urqi1WpNr/tSd9Zi705f8rdFnpcTsxBCCNuRhpsQQtjAyJE+1NXVmR3TNjTQ3HzW7JiPz0gUCoXFkLTk5KkMHz68bx+q+PHpbkxN46uvvqS0tISc7GymTkvBxcWlb3H0wcacbKZOncYjjz5mOrZ586Y+5bF02QpefumPLF2SyeKMTDwu6VGzt7cnNW0OqWlzaDMYyN2ay7vv/JXQsDDGXhhOaE1dXR3BwcFmrzsWg4F+rrsu9Cp/a3Vp5bitYxZCCGE7MlRSCCFsIGX6DHbt2kVlZSUARqOR1av/ZZEuNS0NtbqawMBAEhITTf9GjBiBnV3fbtEdDZae5nB1l65jVccPP3ifY8dqLOaZ9Tel0p4ffvjB9LqhoYG1X3/VpzxcXV3JyMjE3d2NzIx0Tp08afrdxo05tLS0AO3DHSMj2hftaDnX0mO+n3+2xrSn2vHjxynIz2P69Omm3/dn3XWlN/lbq0trx20dsxBCCNuRHjchhLCBqVOnUVpaSsbidCIiIjhz5gc8PT0teq8SE5P4QaNh8eJFBAQE4ubmRm1tLW5ubjz++BN9+swxY8YSFBTE8uVLCQ4eha+fn2kJ+L6kmzPnJpYvW0p4eAQhISF9Pve+uPOuu3j5pT/y7DNP4+3tTWVlJaGhYRw/frxP+Tg5O/Pi79N5fdVrZGQsJiMjE/+AADSnT/PUb/8bX19fHBwcOHSoghkzZjJ+/Pge8zQYDPzud0/h4zOSQ4cqGB8VxZw5N5l+359115Xe5G+tLq0dt3XMQgghbEeh15/vvwkSQggxRGkbGti3bx9R0dGmFRkBjh49Qn1dPcM9PRk9ejTFxbsJ8A/A18/P7P3nmps5XHmYlpYW/P38zX5/6uRJVCqVxeqLZaWleHp5EhR0cTifXq9HpVKhbWjAxcWF6JgYNBoNFQcPEh8fj92Fvbq6StfhzA8/8Nhjj/Cb3/w3KdNn9HjuXcXXZjBQVFREeEQEXl5epuP79+/DwcHBbC+3s01NqA6r0Ov1jBoVgoODA+V79xIbG4uDo6PV87f2ubt278LV1ZXx46MAaG5upqqqkvPnzxMQEGixF1tn27Zt5e233iRrzefU1h6n9sI+bqGhYV2mv5y660pX9dRT/mC9Lrur4/6KWQghxJUjDTchhBBmPvn4I/LytvHW23+1WFxkKLi04SaEEEIMFDJUUgghBAB5edvYtnUre/d+z9NPPzMkG21CCCHEQCU9bkIIIQA4cuQIdXW1BAcFWwzHG0pkqKAQQoiBSBpuQgghhBBCCDHAydq/QgghhBBCCDHAScNNCCGEEEIIIQY4abgJIYQQQgghxAAnDTchhBBCCCGEGOCk4SaEEEIIIYQQA5w03IQQQgghhBBigJOGmxBCCCGEEEIMcNJwE0IIIYQQQogBThpuQgghhBBCCDHA/T/J3YE+lw6F8gAAAABJRU5ErkJggg==\n"}, "metadata": {}}, {"output_type": "display_data", "data": {"text/plain": ["<Figure size 800x300 with 1 Axes>"], "image/png": "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\n"}, "metadata": {}}, {"output_type": "display_data", "data": {"text/plain": ["<Figure size 800x300 with 1 Axes>"], "image/png": "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\n"}, "metadata": {}}]}, {"cell_type": "markdown", "metadata": {"id": "7GjnVafDYy40"}, "source": ["## 8. Write results into the README"]}, {"cell_type": "code", "source": ["def pct(v):\n", "    return \"–\" if v is None else f\"{v:.0%}\"\n", "\n", "S = summary\n", "afr = [\"yoruba\", \"igbo\", \"hausa\"]\n", "ex = {g: (S.get(g, {}).get(\"copy_exact\")) for g in GROUPS}\n", "examples = []\n", "for g in afr:\n", "    bad = copy_df[(copy_df.lang == g) & (copy_df.status != \"exact\") & (copy_df.n_marks > 0)]\n", "    if len(bad):\n", "        r = bad.iloc[0]\n", "        shown = r[\"rendition\"] if r[\"rendition\"] else f\"({r['status'].replace('_', ' ')})\"\n", "        examples.append(f\"| {GLABEL[g]} | {r['template']} | {r['name']} | {shown} |\")\n", "\n", "# --- derived numbers for the findings ---\n", "upper = {g: by_template.get(g, {}).get(\"c5_upper\") for g in GROUPS}\n", "other = {g: (sum(v for t, v in by_template.get(g, {}).items() if t != \"c5_upper\") /\n", "             max(1, len([t for t in by_template.get(g, {}) if t != \"c5_upper\"]))) for g in GROUPS}\n", "pair_df = pd.DataFrame(res[\"pair\"])\n", "collapse = []\n", "for sname, grp in pair_df.groupby(\"set\"):\n", "    for ans, sub in grp.groupby(grp.output.str.strip().str.lower().str.rstrip(\".\")):\n", "        if len(sub) > 1:\n", "            collapse.append((len(sub), sname, list(sub.word), sub.output.iloc[0].strip()))\n", "collapse.sort(reverse=True)\n", "n_sets = pair_df.set.nunique(); n_collapsed = len({c[1] for c in collapse})\n", "ex_c = collapse[0] if collapse else None\n", "up_ex = copy_df[(copy_df.template == \"c5_upper\") & (copy_df.lang != \"control\") & (copy_df.status != \"exact\")].head(3)\n", "up_txt = \", \".join(f\"*{r['name']}* → *{r['rendition'] or r['status']}*\" for _, r in up_ex.iterrows())\n", "\n", "findings = [\n", "    f\"**Copying works while the name sits in the prompt, until the letter case changes.** In the repeat, welcome-message, JSON and \"\n", "    f\"translation contexts the model kept names exactly {pct(other['yoruba'])} (Yorùbá), {pct(other['igbo'])} (Igbo) and \"\n", "    f\"{pct(other['hausa'])} (Hausa) of the time. Asked for capital letters, it kept only {pct(upper['yoruba'])}, {pct(upper['igbo'])} and \"\n", "    f\"{pct(upper['hausa'])} intact, against {pct(upper['control'])} for controls ({up_txt}). All copy-task loss comes from this one context.\",\n", "    f\"**It cannot produce the marks itself.** Restoring a stripped name: Yorùbá {pct(S['yoruba'].get('restore_exact'))} exact \"\n", "    f\"(mark F1 {S['yoruba'].get('restore_mark_f1')}), Igbo {pct(S['igbo'].get('restore_exact'))}, Hausa {pct(S['hausa'].get('restore_exact'))}, \"\n", "    f\"against {pct(S['control'].get('restore_exact'))} for controls such as *François* and *Nguyễn Thị Hương*.\",\n", "    f\"**It cannot read tone marks, even with the rule in the prompt.** Exact Yorùbá tone sequence: {pct(S['yoruba'].get('tone_read_exact'))}; \"\n", "    f\"per-vowel accuracy {pct(S['yoruba'].get('tone_read_position_acc'))}, close to the 33% expected from guessing among three tones.\",\n", "    f\"**It does not know what the names mean.** Meaning correct for the marked name: Yorùbá {pct(S['yoruba'].get('meaning_acc_marked'))}, \"\n", "    f\"Igbo {pct(S['igbo'].get('meaning_acc_marked'))}, Hausa {pct(S['hausa'].get('meaning_acc_marked'))} (only 3 Hausa names, all day-of-birth names).\",\n", "    f\"**Marks do not change its answer.** Minimal-pair accuracy: Yorùbá {pct(S['yoruba'].get('minimal_pair_acc'))}, Igbo {pct(S['igbo'].get('minimal_pair_acc'))}, \"\n", "    f\"Hausa {pct(S['hausa'].get('minimal_pair_acc'))}. In {n_collapsed} of {n_sets} sets it gave one answer to words that differ only by marks\"\n", "    + (f\" (e.g. {', '.join('*' + w + '*' for w in ex_c[2])} were all \\\"{ex_c[3]}\\\").\" if ex_c else \".\")\n", "    + f\" Asked to list what a bare form could mean, it recalled {pct(S['yoruba'].get('ambiguity_sense_recall'))} / {pct(S['igbo'].get('ambiguity_sense_recall'))} / \"\n", "    f\"{pct(S['hausa'].get('ambiguity_sense_recall'))} of the senses and spelled {pct(S['yoruba'].get('ambiguity_spelled_correctly'))} / \"\n", "    f\"{pct(S['igbo'].get('ambiguity_spelled_correctly'))} / {pct(S['hausa'].get('ambiguity_spelled_correctly'))} of the marked words correctly (Yorùbá / Igbo / Hausa).\",\n", "    f\"**The stripped spelling is its default.** It scores the stripped name as more likely than the correct one in {pct(lp_summary.get('yoruba'))} (Yorùbá), \"\n", "    f\"{pct(lp_summary.get('igbo'))} (Igbo) and {pct(lp_summary.get('hausa'))} (Hausa) of cases, against {pct(lp_summary.get('control'))} for controls. \"\n", "    f\"Correct Yorùbá names also cost {tok_summary.get('yoruba')}× the tokens of their stripped forms (Hausa {tok_summary.get('hausa')}×, controls {tok_summary.get('control')}×).\",\n", "    f\"**Diacritic density did not predict copy failures** (Spearman ρ = {rho_d:.2f}, p = {p_d:.2f}); the failures cluster in one context rather than in heavily marked names.\",\n", "]\n", "\n", "table = [\"| Metric | Yorùbá | Igbo | Hausa | Controls |\", \"|---|---|---|---|---|\"]\n", "for k, label in [(\"copy_exact\", \"Name kept exactly (copy tasks)\"), (\"copy_mark_recall\", \"Identity marks kept (copy)\"),\n", "                 (\"copy_fully_stripped\", \"All marks stripped (copy)\"), (\"copy_catastrophic\", \"Letter dropped entirely (copy)\"),\n", "                 (\"restore_exact\", \"Exact diacritic restoration\"), (\"restore_mark_f1\", \"Restoration mark F1\"),\n", "                 (\"restore_overcorrection\", \"Over-correction on unmarked names\"),\n", "                 (\"meaning_acc_marked\", \"Meaning correct (marked)\"), (\"meaning_acc_stripped\", \"Meaning correct (stripped)\"),\n", "                 (\"minimal_pair_acc\", \"Minimal-pair accuracy\"), (\"tone_read_exact\", \"Tone sequence read exactly\")]:\n", "    row = [S.get(g, {}).get(k) for g in GROUPS]\n", "    fmt = (lambda v: \"–\" if v is None else (f\"{v:.2f}\" if \"f1\" in k else f\"{v:.0%}\"))\n", "    table.append(f\"| {label} | \" + \" | \".join(fmt(v) for v in row) + \" |\")\n", "\n", "block = \"\\n\".join([\n", "    f\"### Results (model: `{MODEL_ID}`, greedy, non-thinking, {len(gens)} generations)\",\n", "    \"\", *[f\"{i + 1}. {f}\" for i, f in enumerate(findings)], \"\",\n", "    *table, \"\",\n", "    \"Examples from the run (first non-exact case per language):\", \"\",\n", "    \"| Language | Context | Correct name | Model wrote |\", \"|---|---|---|---|\", *examples, \"\",\n", "    \"![Copy outcomes](figures/fig1_copy_outcomes.png)\",\n", "    \"![Headline metrics](figures/fig2_headline.png)\",\n", "    \"![By context](figures/fig3_copy_by_context.png)\",\n", "    \"![Tokens](figures/fig5_token_ratio.png)\",\n", "    \"![Likelihood](figures/fig6_prefers_stripped.png)\",\n", "    \"\", \"*Numbers above are generated by the notebook; read `results/generations.jsonl` to audit every output.*\",\n", "])\n", "readme = open(\"README.md\", encoding=\"utf-8\").read()\n", "readme = re.sub(r\"<!-- RESULTS:START -->.*<!-- RESULTS:END -->\",\n", "                lambda m: \"<!-- RESULTS:START -->\\n\" + block + \"\\n<!-- RESULTS:END -->\", readme, flags=re.S)\n", "open(\"README.md\", \"w\", encoding=\"utf-8\").write(readme)\n", "json.dump(dict(model=MODEL_ID, summary=S, copy_by_template=by_template, lost_marks_copy=lost_copy,\n", "               tokenizer_ratio=tok_summary, prefers_stripped=lp_summary,\n", "               spearman_density=[rho_d, p_d], spearman_length=[rho_l, p_l]),\n", "          open(\"results/summary.json\", \"w\"), indent=2, ensure_ascii=False)\n", "from IPython.display import Markdown\n", "Markdown(block)"], "metadata": {"colab": {"base_uri": "https://localhost:8080/"}, "id": "N1G8l1vnuOco", "executionInfo": {"status": "ok", "timestamp": 1792105863210, "user_tz": -60, "elapsed": 6, "user": {"displayName": "Flora Oladipupo", "userId": "04462134423746892885"}}, "outputId": "46cd7b11-4c58-4640-ee7b-7368599831ca"}, "execution_count": 16, "outputs": [{"output_type": "execute_result", "data": {"text/plain": ["<IPython.core.display.Markdown object>"], "text/markdown": "### Results (model: `Qwen/Qwen3.5-4B`, greedy, non-thinking, 363 generations)\n\n1. **Copying works while the name sits in the prompt, until the letter case changes.** In the repeat, welcome-message, JSON and translation contexts the model kept names exactly 100% (Yorùbá), 100% (Igbo) and 100% (Hausa) of the time. Asked for capital letters, it kept only 30%, 0% and 33% intact, against 78% for controls (*Adéṣọlá* → *ADEṢOLÁ*, *Ọláyínká* → *ỌLAYINKA*, *Fọláṣadé* → *FOLÁṢADÉ*). All copy-task loss comes from this one context.\n2. **It cannot produce the marks itself.** Restoring a stripped name: Yorùbá 0% exact (mark F1 0.453), Igbo 12%, Hausa 0%, against 100% for controls such as *François* and *Nguyễn Thị Hương*.\n3. **It cannot read tone marks, even with the rule in the prompt.** Exact Yorùbá tone sequence: 5%; per-vowel accuracy 35%, close to the 33% expected from guessing among three tones.\n4. **It does not know what the names mean.** Meaning correct for the marked name: Yorùbá 5%, Igbo 0%, Hausa 50% (only 3 Hausa names, all day-of-birth names).\n5. **Marks do not change its answer.** Minimal-pair accuracy: Yorùbá 25%, Igbo 0%, Hausa 25%. In 4 of 5 sets it gave one answer to words that differ only by marks (e.g. *igbá*, *ìgbà*, *igba*, *ìgbá* were all \"Time\"). Asked to list what a bare form could mean, it recalled 12% / 0% / 25% of the senses and spelled 17% / 100% / 0% of the marked words correctly (Yorùbá / Igbo / Hausa).\n6. **The stripped spelling is its default.** It scores the stripped name as more likely than the correct one in 100% (Yorùbá), 100% (Igbo) and 100% (Hausa) of cases, against 50% for controls. Correct Yorùbá names also cost 2.13× the tokens of their stripped forms (Hausa 1.89×, controls 1.23×).\n7. **Diacritic density did not predict copy failures** (Spearman ρ = 0.09, p = 0.63); the failures cluster in one context rather than in heavily marked names.\n\n| Metric | Yorùbá | Igbo | Hausa | Controls |\n|---|---|---|---|---|\n| Name kept exactly (copy tasks) | 86% | 80% | 87% | 96% |\n| Identity marks kept (copy) | 92% | 82% | 87% | 91% |\n| All marks stripped (copy) | 4% | 18% | 13% | 4% |\n| Letter dropped entirely (copy) | 0% | 0% | 0% | 0% |\n| Exact diacritic restoration | 0% | 12% | 0% | 100% |\n| Restoration mark F1 | 0.45 | 0.20 | 0.00 | 1.00 |\n| Over-correction on unmarked names | – | 0% | 0% | – |\n| Meaning correct (marked) | 5% | 0% | 50% | – |\n| Meaning correct (stripped) | 0% | 0% | 67% | – |\n| Minimal-pair accuracy | 25% | 0% | 25% | – |\n| Tone sequence read exactly | 5% | – | – | – |\n\nExamples from the run (first non-exact case per language):\n\n| Language | Context | Correct name | Model wrote |\n|---|---|---|---|\n| Yorùbá | c5_upper | Adéṣọlá | ADEṢOLÁ |\n| Igbo | c5_upper | Chiọma | CHIOMA |\n| Hausa | c5_upper | Ɗanjuma | DANJUMA |\n\n![Copy outcomes](figures/fig1_copy_outcomes.png)\n![Headline metrics](figures/fig2_headline.png)\n![By context](figures/fig3_copy_by_context.png)\n![Tokens](figures/fig5_token_ratio.png)\n![Likelihood](figures/fig6_prefers_stripped.png)\n\n*Numbers above are generated by the notebook; read `results/generations.jsonl` to audit every output.*"}, "metadata": {}, "execution_count": 16}]}, {"cell_type": "markdown", "metadata": {"id": "QL8P9ReyYy41"}, "source": ["## 9. Publish to a public Hugging Face Storage Bucket"]}, {"cell_type": "code", "metadata": {"id": "15-OpgKbYy41"}, "execution_count": null, "outputs": [], "source": ["# Save a copy of this notebook (with outputs) into the folder so it ships with the results.\n", "try:\n", "    from google.colab import _message\n", "    nb = _message.blocking_request(\"get_ipynb\", timeout_sec=60)[\"ipynb\"]\n", "    json.dump(nb, open(\"oruko_eval.ipynb\", \"w\", encoding=\"utf-8\"), ensure_ascii=False)\n", "    print(\"Notebook saved into the bucket folder.\")\n", "except Exception as e:\n", "    print(\"Could not auto-save the notebook; upload oruko_eval.ipynb to\", WORKDIR, \"with the Files panel.\", e)\n", "\n", "from huggingface_hub import create_bucket, sync_bucket, copy_files\n", "bucket_id = f\"{HF_USER}/{BUCKET_NAME}\"\n", "create_bucket(bucket_id, private=False, exist_ok=True)\n", "sync_bucket(\".\", f\"hf://buckets/{bucket_id}\",\n", "            exclude=[\"__pycache__/*\", \"*.pyc\", \".ipynb_checkpoints/*\", \".config/*\", \"sample_data/*\"])\n", "if COPY_MODEL_TO_BUCKET:\n", "    try:\n", "        copy_files(f\"hf://{MODEL_ID}/\", f\"hf://buckets/{bucket_id}/model/\")\n", "        print(\"Model weights copied server-side into model/\")\n", "    except Exception as e:\n", "        print(\"Model copy skipped:\", e)\n", "print(\"\\nYour public bucket:  https://huggingface.co/buckets/\" + bucket_id)"]}], "metadata": {"accelerator": "GPU", "colab": {"provenance": [], "gpuType": "L4"}, "kernelspec": {"name": "python3", "display_name": "Python 3"}, "language_info": {"name": "python"}, "widgets": {"application/vnd.jupyter.widget-state+json": {"version_major": 2, "version_minor": 0, "state": {"df82e6ba303345d3abc0e6153b167b49": {"model_module": "@jupyter-widgets/controls", "model_name": "HBoxModel", "model_module_version": "1.5.0", "state": {"_dom_classes": [], "_model_module": "@jupyter-widgets/controls", "_model_module_version": "1.5.0", "_model_name": "HBoxModel", "_view_count": null, "_view_module": "@jupyter-widgets/controls", "_view_module_version": "1.5.0", "_view_name": "HBoxView", "box_style": "", "children": ["IPY_MODEL_541bbb7d39074dc08a17f783b0e46a69", "IPY_MODEL_1a8781a695f64c21894c986b5494fd58", "IPY_MODEL_a9f2a9894cae44c0b3028d5d135ad144"], "layout": "IPY_MODEL_1f2c6d7259f442da814bbe4ecdece838"}}, "541bbb7d39074dc08a17f783b0e46a69": {"model_module": "@jupyter-widgets/controls", "model_name": "HTMLModel", "model_module_version": "1.5.0", "state": {"_dom_classes": [], "_model_module": "@jupyter-widgets/controls", "_model_module_version": "1.5.0", "_model_name": "HTMLModel", "_view_count": null, "_view_module": "@jupyter-widgets/controls", "_view_module_version": "1.5.0", "_view_name": "HTMLView", "description": "", "description_tooltip": null, "layout": "IPY_MODEL_044e17399c674a128d34f72e2225e8f8", "placeholder": "​", "style": "IPY_MODEL_e9897add835f480c8c9c60d3565775a4", "value": "Loading weights: 100%"}}, "1a8781a695f64c21894c986b5494fd58": {"model_module": "@jupyter-widgets/controls", "model_name": "FloatProgressModel", "model_module_version": "1.5.0", "state": {"_dom_classes": [], "_model_module": "@jupyter-widgets/controls", "_model_module_version": "1.5.0", "_model_name": "FloatProgressModel", "_view_count": null, "_view_module": "@jupyter-widgets/controls", "_view_module_version": "1.5.0", "_view_name": "ProgressView", "bar_style": "success", "description": "", "description_tooltip": null, "layout": "IPY_MODEL_55901d4388d94f7ca3faf43b1126051e", "max": 723, "min": 0, "orientation": "horizontal", "style": "IPY_MODEL_e78a379c0ae14d4699ca65025ff3b89e", "value": 723}}, "a9f2a9894cae44c0b3028d5d135ad144": {"model_module": "@jupyter-widgets/controls", "model_name": "HTMLModel", "model_module_version": "1.5.0", "state": {"_dom_classes": [], "_model_module": "@jupyter-widgets/controls", "_model_module_version": "1.5.0", "_model_name": "HTMLModel", "_view_count": null, "_view_module": "@jupyter-widgets/controls", "_view_module_version": "1.5.0", "_view_name": "HTMLView", "description": "", "description_tooltip": null, "layout": "IPY_MODEL_1df95a82d8f34472a21f15ee11b225d7", "placeholder": "​", "style": "IPY_MODEL_09eb42db9ed64a818e40ec652e0e6040", "value": " 723/723 [00:02&lt;00:00, 785.01it/s]"}}, "1f2c6d7259f442da814bbe4ecdece838": {"model_module": "@jupyter-widgets/base", "model_name": "LayoutModel", "model_module_version": "1.2.0", "state": {"_model_module": "@jupyter-widgets/base", "_model_module_version": "1.2.0", "_model_name": "LayoutModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "1.2.0", "_view_name": "LayoutView", "align_content": null, "align_items": null, "align_self": null, "border": null, "bottom": null, "display": null, "flex": null, "flex_flow": null, "grid_area": null, "grid_auto_columns": null, "grid_auto_flow": null, "grid_auto_rows": null, "grid_column": null, "grid_gap": null, "grid_row": null, "grid_template_areas": null, "grid_template_columns": null, "grid_template_rows": null, "height": null, "justify_content": null, "justify_items": null, "left": null, "margin": null, "max_height": null, "max_width": null, "min_height": null, "min_width": null, "object_fit": null, "object_position": null, "order": null, "overflow": null, "overflow_x": null, "overflow_y": null, "padding": null, "right": null, "top": null, "visibility": null, "width": null}}, "044e17399c674a128d34f72e2225e8f8": {"model_module": "@jupyter-widgets/base", "model_name": "LayoutModel", "model_module_version": "1.2.0", "state": {"_model_module": "@jupyter-widgets/base", "_model_module_version": "1.2.0", "_model_name": "LayoutModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "1.2.0", "_view_name": "LayoutView", "align_content": null, "align_items": null, "align_self": null, "border": null, "bottom": null, "display": null, "flex": null, "flex_flow": null, "grid_area": null, "grid_auto_columns": null, "grid_auto_flow": null, "grid_auto_rows": null, "grid_column": null, "grid_gap": null, "grid_row": null, "grid_template_areas": null, "grid_template_columns": null, "grid_template_rows": null, "height": null, "justify_content": null, "justify_items": null, "left": null, "margin": null, "max_height": null, "max_width": null, "min_height": null, "min_width": null, "object_fit": null, "object_position": null, "order": null, "overflow": null, "overflow_x": null, "overflow_y": null, "padding": null, "right": null, "top": null, "visibility": null, "width": null}}, "e9897add835f480c8c9c60d3565775a4": {"model_module": "@jupyter-widgets/controls", "model_name": "DescriptionStyleModel", "model_module_version": "1.5.0", "state": {"_model_module": "@jupyter-widgets/controls", "_model_module_version": "1.5.0", "_model_name": "DescriptionStyleModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "1.2.0", "_view_name": "StyleView", "description_width": ""}}, "55901d4388d94f7ca3faf43b1126051e": {"model_module": "@jupyter-widgets/base", "model_name": "LayoutModel", "model_module_version": "1.2.0", "state": {"_model_module": "@jupyter-widgets/base", "_model_module_version": "1.2.0", "_model_name": "LayoutModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "1.2.0", "_view_name": "LayoutView", "align_content": null, "align_items": null, "align_self": null, "border": null, "bottom": null, "display": null, "flex": null, "flex_flow": null, "grid_area": null, "grid_auto_columns": null, "grid_auto_flow": null, "grid_auto_rows": null, "grid_column": null, "grid_gap": null, "grid_row": null, "grid_template_areas": null, "grid_template_columns": null, "grid_template_rows": null, "height": null, "justify_content": null, "justify_items": null, "left": null, "margin": null, "max_height": null, "max_width": null, "min_height": null, "min_width": null, "object_fit": null, "object_position": null, "order": null, "overflow": null, "overflow_x": null, "overflow_y": null, "padding": null, "right": null, "top": null, "visibility": null, "width": null}}, "e78a379c0ae14d4699ca65025ff3b89e": {"model_module": "@jupyter-widgets/controls", "model_name": "ProgressStyleModel", "model_module_version": "1.5.0", "state": {"_model_module": "@jupyter-widgets/controls", "_model_module_version": "1.5.0", "_model_name": "ProgressStyleModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "1.2.0", "_view_name": "StyleView", "bar_color": null, "description_width": ""}}, "1df95a82d8f34472a21f15ee11b225d7": {"model_module": "@jupyter-widgets/base", "model_name": "LayoutModel", "model_module_version": "1.2.0", "state": {"_model_module": "@jupyter-widgets/base", "_model_module_version": "1.2.0", "_model_name": "LayoutModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "1.2.0", "_view_name": "LayoutView", "align_content": null, "align_items": null, "align_self": null, "border": null, "bottom": null, "display": null, "flex": null, "flex_flow": null, "grid_area": null, "grid_auto_columns": null, "grid_auto_flow": null, "grid_auto_rows": null, "grid_column": null, "grid_gap": null, "grid_row": null, "grid_template_areas": null, "grid_template_columns": null, "grid_template_rows": null, "height": null, "justify_content": null, "justify_items": null, "left": null, "margin": null, "max_height": null, "max_width": null, "min_height": null, "min_width": null, "object_fit": null, "object_position": null, "order": null, "overflow": null, "overflow_x": null, "overflow_y": null, "padding": null, "right": null, "top": null, "visibility": null, "width": null}}, "09eb42db9ed64a818e40ec652e0e6040": {"model_module": "@jupyter-widgets/controls", "model_name": "DescriptionStyleModel", "model_module_version": "1.5.0", "state": {"_model_module": "@jupyter-widgets/controls", "_model_module_version": "1.5.0", "_model_name": "DescriptionStyleModel", "_view_count": null, "_view_module": "@jupyter-widgets/base", "_view_module_version": "1.2.0", "_view_name": "StyleView", "description_width": ""}}}}}}, "nbformat": 4, "nbformat_minor": 0}