Standardize Chart 2/3 labels and Per-Category Recall table/caption parity

Remove 'Cheap' prefix in Chart 2, replace tc2000 with zero_cue_2000 in Chart 3 Panel 4 and tables, and sync SHA-256 provenance.

README.md

@@ -24,7 +24,7 @@ pinned: true
 
 | Model Repository | Role | Footprint | Hardware-Attributed Speed | 6-Dim Composite | 11-Class Category (Unmutated Keys) | Movie vs. TV (`oct15` `N=1,296`) | Audio Language (`oct15` `N=545`) | Swarm Liveliness (AUROC) |
 | :--- | :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
-| **[`NorviTech/armiger-ultra`](https://huggingface.co/NorviTech/armiger-ultra)** | **Unified Flagship Suite** (`Armiger` + `Ultra FP32` + `armiger_inference.py` + **All 6 Plugins Rolled In**) | `32.3 MB` model (`+26.0 MB` `BLAKE2b-64` catalog) | **`0.21 ms` RTX GPU / `0.29 ms` M4 Pro MPS** (batch-64) | **`97.60%`** | **`95.49%`** `oct15` / **`95.12%`** `beyond` / **`98.47%`** `mlang` (`95.42%` 5-split `N=10,262`) | **`97.30%`** (`96.54%` `beyond`) | **`97.80%`** | **`96.83%`** Corpus / **`92.42%`** Pub (`100%` trap recall) |
+| **[`NorviTech/armiger-ultra`](https://huggingface.co/NorviTech/armiger-ultra)** | **Unified Flagship Suite** (`Armiger` + `Ultra FP32` + `armiger_inference.py` + **All 6 Plugins Rolled In**) | `32.3 MB` model (`+26.0 MB` `BLAKE2b-64` catalog) | **`0.21 ms` RTX GPU / `0.29 ms` M4 Pro MPS** (batch-64) | **`96.86%`** Pub / **`97.60%`** Corpus | **`95.49%`** `oct15` / **`95.12%`** `beyond` / **`98.47%`** `mlang` (`95.42%` 5-split `N=10,262`) | **`97.30%`** (`96.54%` `beyond`) | **`97.80%`** | **`96.83%`** Corpus / **`92.42%`** Pub (`100%` trap recall) |
 | **[`NorviTech/armiger`](https://huggingface.co/NorviTech/armiger)** | **Universal Fast CPU Classifier** (`ONNX INT8` + `armiger_inference.py`) | **`8.3 MB`** | **`2.44 ms` single-item / `3.07 ms` batch-64** (M4 Pro CPU) | **`96.85%`** (`96.50%` Raw) | **`95.06%`** `oct15` / **`92.88%`** `beyond` / **`97.37%`** `mlang` (`93.95%` 5-split `N=10,262`) | **`96.60%`** | **`96.88%`** | **`86.64%`** Priors (`92.42%` w/ Adapter) |
 
 ## 2. Armiger Modular Plugins (6 Standalone Repos & Rolled Into `Armiger Ultra`)
@@ -56,7 +56,7 @@ All benchmark charts and headline metrics across the Armiger `v2.1.0` suite are
 1. **Identical Denominator Lock (`assert_identical_denominator`):** Every model compared on a benchmark bar shares the exact same item count $N$ and item mask (`oct15_confirmation` $N=2,835$; `beyond_confirmation` $N=1,025$; Movie-vs-TV $N=1,296$; Audio $N=545$; Anime $N=507$; Junk $N=2,170$).
 2. **Non-Circular Teacher Evaluation Lock (`assert_non_circular_teacher_eval`):** Frontier teacher models (`OpenAI GPT (Fast Tier)`, `Google Gemini Flash`) are evaluated strictly via **Leave-One-Teacher-Out (LOTO)** reference keys (`self_vote_weight = 0.0`), removing $+1.47\text{ pp}$ and $+1.81\text{ pp}$ of circular self-vote inflation.
 3. **Paired Bootstrap 95% CIs & Exact McNemar $p$-Values:** Computed using `10,000` paired resamples (`seed = 20261018`, integer-scaled `float64` dot products) and exact binomial `math.comb` McNemar tests (`p = 1.42e-11` vs. `Google Gemini Flash` LOTO; `p = 3.88e-24` vs. `Anthropic Claude Haiku 4.5`).
-4. **Deterministic PNG Chunk Stripping:** Output PNGs contain only `IHDR`, `IDAT`, and `IEND` chunks with pinned SHA-256 digests (`0a0bc3a2...`, `3cf46fd5...`, `3768e7ec...`, `1a77a643...`).
+4. **Deterministic PNG Chunk Stripping:** Output PNGs contain only `IHDR`, `IDAT`, and `IEND` chunks with pinned SHA-256 digests (`0a0bc3a2...`, `7344ba71...`, `51e1f229...`, `1a77a643...`).
 
 ---