Add model-index evaluation metadata for LLaDA2.0-flash to README

- Added a `model-index` block to the README, converting the existing benchmark table into structured evaluation metadata for LLaDA2.0-flash.
- The metrics are taken from the LLaDA2.0-flash column of the existing benchmark table.

README.md

@@ -6,6 +6,101 @@ tags:
 - diffusion
 - llm
 - text_generation
+model-index:
+  - name: LLaDA2.0-flash
+    results:
+      - task:
+          name: Text Generation
+          type: text-generation
+        dataset:
+          name: Benchmarks
+          type: benchmarks
+        metrics:
+          - name: Average
+            type: average
+            value: 79.32
+
+          # Knowledge
+          - name: MMLU
+            type: mmlu
+            value: 87.69
+          - name: MMLU-Pro
+            type: mmlu-pro
+            value: 73.36
+          - name: GPQA
+            type: gpqa
+            value: 61.98
+          - name: ARC-C
+            type: arc-c
+            value: 95.93
+          - name: CMMLU
+            type: cmmlu
+            value: 85.13
+          - name: C-EVAL
+            type: c-eval
+            value: 86.75
+          - name: GAOKAO-Bench
+            type: gaokao-bench
+            value: 93.90
+
+          # Reasoning
+          - name: SQuAD 2.0
+            type: squad-v2
+            value: 90.00
+          - name: DROP
+            type: drop
+            value: 87.90
+          - name: KOR-Bench
+            type: kor-bench
+            value: 64.24
+          - name: HellaSwag
+            type: hellaswag
+            value: 84.97
+
+          # Coding
+          - name: CRUXEval-O
+            type: cruxeval-o
+            value: 85.12
+          - name: MBPP
+            type: mbpp
+            value: 88.29
+          - name: MultiPL-E
+            type: multipl-e
+            value: 74.87
+          - name: HumanEval
+            type: humaneval
+            value: 94.51
+          - name: Bigcodebench-Full
+            type: bigcodebench-full
+            value: 41.58
+          - name: LiveCodeBench
+            type: livecodebench
+            value: 42.29
+          - name: Spider
+            type: spider
+            value: 82.49
+
+          # Math
+          - name: GSM8K
+            type: gsm8k
+            value: 96.06
+          - name: MATH
+            type: math
+            value: 95.44
+          - name: OlympiadBench
+            type: olympiadbench
+            value: 74.07
+          - name: AIME 2025
+            type: aime-2025
+            value: 60.00
+
+          # Agent & Alignment
+          - name: BFCL_Live
+            type: bfcl_live
+            value: 75.43
+          - name: IFEval-strict -prompt
+            type: ifeval-strict
+            value: 81.70
 ---
 # LLaDA2.0-flash