Place Unbind first and shuffle competing model columns

BENCHMARKS.md

@@ -50,4 +50,4 @@ UnbindAI reports 320/320 tests passed. Suite, pass criteria, and logs were not s
 
 [CSV comparison table](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/results-v1/comparison-table.csv) · [Source data and provenance](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/results-v1/source-data.json) · [Asset manifest](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/results-v1/manifest.json)
 
-Every graphic has a matching editable SVG under `assets/results-v1/`. Charts are sorted by supplied score; tables preserve the supplied model order. No combined score has been computed.
+Every graphic has a matching editable SVG under `assets/results-v1/`. Individual benchmark charts are sorted by supplied score. The full comparison puts Unbind first, followed by other models in a fixed shuffled order independent of scores; outlined cells mark the highest score for each benchmark. The coding table retains the supplied model order. No combined score has been computed.

assets/results-v1/README.txt

@@ -1,19 +1,14 @@
 UNBIND FLASH V1 — SUPPLIED RESULTS GRAPHICS
 
-10 graphics, each supplied as PNG and editable SVG:
-01–07: individual benchmark comparisons
-08: complete seven-benchmark comparison table
-09: coding comparison table
-10: reported 320/320 test result
-preview-sheet.png: visual index
-comparison-table.csv: exact supplied table
-source-data.json: exact values, asterisks and claim provenance
-manifest.json: image dimensions and asset inventory
+10 graphics, each as PNG and editable SVG.
+01–07: individual benchmark charts, sorted by supplied score.
+08: full comparison, benchmarks on the left and models across the top. Unbind is first and green; other columns use a saved random order independent of scores. White outlines identify every benchmark maximum, including ties.
+09: coding comparison, retaining the supplied model order.
+10: user-reported 320/320 test result.
 
-Figures supplied by UnbindAI in chat. No external verification performed.
-All asterisks are retained. Their definition was not supplied.
-Benchmark charts are sorted by supplied score; comparison tables retain supplied model order.
-No average, combined score, unsupported win, competitor refusal claim, or universal uncensored claim was added.
-The separate 320/320 card records the user-reported result; suite, criteria and raw logs were not supplied.
+comparison-table.csv and source-data.json preserve the original supplied data.
+full-comparison-transposed.csv matches the revised full chart.
+comparison-column-order.json records the column order.
+preview-sheet.png is a visual index; manifest.json lists exports and dimensions.
 
-Full comparison layout: benchmarks down the left; models across the top. Unbind has a green column; outlined cells mark each benchmark maximum, including tied maxima. full-comparison-transposed.csv matches this layout.
+Data supplied by UnbindAI; not independently verified. Asterisks preserved as supplied; their meaning was not specified. The 320/320 test result was supplied without suite name, pass criteria or raw logs.

assets/results-v1/comparison-column-order.json

@@ -0,0 +1,11 @@
+[
+  "Unbind Flash V1",
+  "Claude Sonnet 5",
+  "GPT-5.6 Terra",
+  "GPT-5.6 Sol",
+  "Claude Opus 5",
+  "Muse Spark 1.3",
+  "GPT-6 Sol",
+  "Gemini 3.1 Pro",
+  "Claude Fable 5"
+]

assets/results-v1/full-comparison-transposed.csv

@@ -1,8 +1,8 @@
-Benchmark,Claude Opus 5,GPT-5.6 Sol,GPT-6 Sol,Claude Sonnet 5,GPT-5.6 Terra,Gemini 3.1 Pro,Muse Spark 1.3,Claude Fable 5,Unbind Flash V1
-AA Index v5,45.0*,39.5*,43.0*,30.0*,34.0*,30.0*,36.4,38.0*,31.0*
-HLE-Diamond,49.0*,40.0*,45.2,30.0*,28.8,31.4,33.7,45.0*,26.9
-GPQA Diamond,93.9,93.5,94.3,80.3,93.3,94.4,93.5,85.9,93.1
-Terminal-Bench 4,49.0,39.9,43.9,14.1,35.4,4.0,33.3,42.4,12.6
-FrontierCode 2,35.0*,28.0*,32.0*,20.0*,25.0*,18.0*,23.1,30.0*,10.9
-LiveBench 2,78.0*,73.0*,77.0*,68.0*,67.8,69.0*,63.0,76.0*,64.6
-LiveCodeBench,89.0,82.6,87.0*,82.4,85.9,88.5,85.0*,89.8,87.2
+Benchmark,Unbind Flash V1,Claude Sonnet 5,GPT-5.6 Terra,GPT-5.6 Sol,Claude Opus 5,Muse Spark 1.3,GPT-6 Sol,Gemini 3.1 Pro,Claude Fable 5
+AA Index v5,31.0*,30.0*,34.0*,39.5*,45.0*,36.4,43.0*,30.0*,38.0*
+HLE-Diamond,26.9,30.0*,28.8,40.0*,49.0*,33.7,45.2,31.4,45.0*
+GPQA Diamond,93.1,80.3,93.3,93.5,93.9,93.5,94.3,94.4,85.9
+Terminal-Bench 4,12.6,14.1,35.4,39.9,49.0,33.3,43.9,4.0,42.4
+FrontierCode 2,10.9,20.0*,25.0*,28.0*,35.0*,23.1,32.0*,18.0*,30.0*
+LiveBench 2,64.6,68.0*,67.8,73.0*,78.0*,63.0,77.0*,69.0*,76.0*
+LiveCodeBench,87.2,82.4,85.9,82.6,89.0,85.0*,87.0*,88.5,89.8