Update README.md

README.md

@@ -6,4 +6,16 @@ tags:
 - code
 size_categories:
 - n<1K
----
+---
+### Baseline datasets for testing LLM runtime reasoning on WebAssembly
+
+All entries share common metadata; only the field required by the task (`stack_values`, `flag_states`, or `return_values`) is populated.
+
+1. dataset\_stack – 10 000 samples
+   Predict the runtime stack at each `;;INSPECT` marker.
+
+2. dataset\_flags – 10 000 samples
+   Identify which flags (`;;FLAG_1 … ;;FLAG_N`) are reachable.
+
+3. dataset\_result – 1 000 samples
+   Predict the program’s final return values.