Merge pull request #3080 from madeline-underwood/sme_ex

Fix formatting and enhance descriptions in profiling documentation

Author: pareenaverma

content/learning-paths/cross-platform/sme-executorch-profiling/01-overview.md

@@ -52,7 +52,7 @@ Before running the pipeline, it helps to understand how the execution stack is c
 
 ### PyTorch to ExecuTorch export
 
-Models are defined using standard PyTorch APIs and exported to a `.pte` (Portable ExecuTorch Executable) format. During export, you specify backend delegation—in this case, XNNPACK—which determines which operators the backend executes at runtime.
+Models are defined using standard PyTorch APIs and exported to a `.pte` (Portable ExecuTorch Executable) format. During export, you specify backend delegation (in this case, XNNPACK) which determines which operators the backend executes at runtime.
 
 ### ExecuTorch runtime and delegation
 

content/learning-paths/cross-platform/sme-executorch-profiling/03-model-onboarding-and-profiling.md

@@ -78,8 +78,8 @@ Keep exported artifacts under out_<model>/artifacts/ to separate model binaries
 ## Configure timing and trace profiling runs
 
 For complete analysis you must run two experiment types:
-  * Timing-only runs — no kernel trace logging; use these for accurate latency comparisons (SME2 on vs SME2 off).
-  * Trace-enabled runs — enable XNNPACK kernel trace logging; use these to collect kernel-selection evidence and traces for kernel-level analysis. Trace logging affects timing and should never be used to compare end-to-end latency.
+  * Timing-only runs - no kernel trace logging; use these for accurate latency comparisons (SME2 on vs SME2 off).
+  * Trace-enabled runs - enable XNNPACK kernel trace logging; use these to collect kernel-selection evidence and traces for kernel-level analysis. Trace logging affects timing and should never be used to compare end-to-end latency.
 
 Run timing-only for latency measurement, and run trace-enabled only when you need kernel-level details.
 
@@ -203,7 +203,7 @@ Timing-only run example:
 python model_profiling/scripts/mac_pipeline.py --config model_profiling/configs/examples/mac_mobilenet_fp16.json
 ```
 
-Trace-enabled (kernel analysis) run example — use a config that points to trace-enabled runners:
+Trace-enabled (kernel analysis) run example - use a config that points to trace-enabled runners:
 
 ```bash
 python model_profiling/scripts/mac_pipeline.py --config model_profiling/configs/examples/mac_mobilenet_fp16.json

content/learning-paths/cross-platform/sme-executorch-profiling/_index.md

@@ -3,6 +3,8 @@ title: Profile ExecuTorch models with SME2 on Arm
 
 minutes_to_complete: 90
 
+description: Learn how to profile and optimize ExecuTorch models using SME2 acceleration on Arm platforms, including operator-level analysis and performance bottleneck identification.
+
 who_is_this_for: This is an advanced topic for developers and performance engineers who deploy ExecuTorch models on Arm devices and want to understand and reduce inference latency.
 
 learning_objectives: