Retire the economy-fidelity scatter for the clearer two-panel dense trade-off chart; correct the dense-slice family list

README.md

@@ -186,8 +186,8 @@ the benchmark, namely both channels as text, the content as a rendered
 image with the instruction as text, the instruction as a rendered image
 with the content as text, and both channels combined into a single image.
 The second slice covers long prose documents, the extraction, reasoning,
-transcription and calculation families at tiers four and five rendered at
-the dense profile, where a document runs to roughly eight thousand
+transcription, calculation and instruction families at tiers four and five
+rendered at the dense profile, where a document runs to roughly eight thousand
 characters in English and where optical compression can save tokens. The
 third slice covers the structured family, machine-oriented text at the same
 long tiers and the same dense profile, so that the economy and the fidelity
@@ -560,18 +560,18 @@ between the task families far more than between the models.
 
 ![Each model's task profile, ranked](charts/en/7_model_task_cascade.png)
 
-The third places each task family on the trade-off between the tokens saved
-and the accuracy lost on long dense documents, for all three models. The
-top-right region, real savings with little accuracy loss, is occupied only
-by the tasks that do not depend on reading exact strings, namely reasoning
-and instruction over prose. Every task that must read exactly sits
-elsewhere. Prose retrieval and transcription fall to the bottom right,
-saving the same tokens but losing almost everything, and structured machine
-text occupies the left half, keeping most of its accuracy but costing more
-tokens than the text it replaced. No task that requires exact reading
-achieves both the saving and the fidelity.
+The third sets what imaging saves against what it costs on the long dense
+documents, family by family, in two aligned panels over shared rows. A
+family would need a rightward bar in both panels for imaging to be a clean
+win, and none has one. Reasoning and instruction over prose come closest,
+saving a quarter to a third of input tokens at a cost of eight and
+seventeen points respectively. The prose retrieval families save the same
+tokens whilst losing sixty-seven and ninety-two points. Structured machine
+text and the line-oriented invoices of the calculation family keep their
+accuracy losses moderate but cost more to image than to tokenise. No family
+that requires exact reading achieves both the saving and the fidelity.
 
-![The economy-fidelity trade-off](charts/en/4_economy_vs_fidelity.png)
+![The dense-document trade-off](charts/en/9_dense_tradeoff.png)
 
 Pooled-language versions of every chart, plus the full family-by-model
 grid, are in the charts directory alongside the per-language sets.
@@ -676,8 +676,8 @@ per-family chart shows the transcription collapse at the bottom, the
 retrieval curve shows the loss deepening with document length, the cascade
 shows GPT 5.5's instruction-channel failure, the economy bars show the
 larger relative prose saving discussed in the language comparison, and the
-trade-off plane shows the prose retrieval families lying on the floor of
-the accuracy axis.
+trade-off panels show the prose retrieval families losing everything that
+imaging saves, with extraction at minus one hundred points.
 
 ![Accuracy by modality condition, Chinese](charts/zh/1_modality_gradient.png)
 
@@ -689,7 +689,7 @@ the accuracy axis.
 
 ![When imaging saves input tokens, Chinese](charts/zh/6_economy_bars.png)
 
-![The Chinese economy-fidelity trade-off](charts/zh/4_economy_vs_fidelity.png)
+![The Chinese dense-document trade-off](charts/zh/9_dense_tradeoff.png)
 
 ## Synthesis