Add GPT-5.6 Luna benchmark result

Replace Gemma E4B benchmark with matched E2B

Add prompt-only baselines and paired fine-tune improvements

Describe evaluation content and distribution

Document external eval decoding contract

Add Fluid-1 Mini and FluidIntelligence eval results

Merge evaluation tables and remove WER column

Publish corrected model evaluation scoreboard

Publish corrected model evaluation scoreboard

Publish corrected model evaluation scoreboard

Publish corrected model evaluation scoreboard

Correct eval metric scoring cohort

Publish exact 0.8B text-only eval split