docs: improve forecasting model search metadata

docs/models.informer.html.md

@@ -1,8 +1,8 @@
 ---
 description: >-
-  Informer: Efficient Transformer with ProbSparse attention for long-sequence time series forecasting. Reduces O(L^2) complexity for scalable predictions.
+  Build long horizon forecasts with Informer in NeuralForecast. Learn ProbSparse attention, architecture, parameters, and a complete Python workflow.
 output-file: models.informer.html
-title: Informer
+title: Informer Time Series Forecasting in Python
 ---
 
 The Informer model tackles the vanilla Transformer computational
@@ -119,4 +119,4 @@ else:
 
 ::: neuralforecast.models.informer.ProbMask
     options:
-      members: []
+      members: []

docs/models.itransformer.html.md

@@ -1,8 +1,8 @@
 ---
 description: >-
-  iTransformer: Inverted Transformer architecture for multivariate time series forecasting with attention on time points and feed-forward on series dimensions.
+  Build multivariate forecasts with iTransformer in NeuralForecast. Learn its inverted attention architecture, parameters, and Python workflow.
 output-file: models.itransformer.html
-title: iTransformer
+title: iTransformer Time Series Forecasting in Python
 ---
 
 The iTransformer model simply takes the Transformer architecture but it
@@ -79,4 +79,4 @@ ax.set_ylabel('Monthly Passengers', fontsize=20)
 ax.set_xlabel('Year', fontsize=20)
 ax.legend(prop={'size': 15})
 ax.grid()
-```
+```

docs/models.kan.html.md

@@ -1,8 +1,8 @@
 ---
 description: >-
-  KAN: Kolmogorov-Arnold Networks for time series forecasting. MLP alternative using learnable activation functions for improved non-linear pattern modeling.
+  Build time series forecasts with Kolmogorov Arnold Networks in NeuralForecast. Compare KAN with MLP models and run a complete Python example.
 output-file: models.kan.html
-title: KAN
+title: KAN Time Series Forecasting in Python
 ---
 
 Kolmogorov-Arnold Networks (KANs) are an alternative to Multi-Layer
@@ -78,4 +78,4 @@ plt.grid()
 
 ::: neuralforecast.models.kan.KANLinear
     options:
-      members: []
+      members: []

docs/models.patchtst.html.md

@@ -1,8 +1,8 @@
 ---
 description: >-
-  PatchTST: Efficient Transformer model for multivariate forecasting using patched time series and channel-independence for scalable long-term predictions.
+  Build long horizon forecasts with PatchTST in NeuralForecast. Learn how temporal patches and channel independence work, then run a complete Python example.
 output-file: models.patchtst.html
-title: PatchTST
+title: PatchTST Time Series Forecasting in Python
 ---
 
 The PatchTST model is an efficient Transformer-based model for
@@ -144,4 +144,4 @@ else:
 
 ::: neuralforecast.models.patchtst.PatchTST_backbone
     options:
-      members: []
+      members: []

docs/models.tft.html.md

@@ -1,8 +1,8 @@
 ---
 description: >-
-  TFT: Temporal Fusion Transformer with interpretable multi-horizon forecasting. LSTM encoder, multi-head attention, variable selection for complex time series.
+  Build interpretable forecasts with Temporal Fusion Transformer in NeuralForecast using static, historic, and future variables in Python.
 output-file: models.tft.html
-title: TFT
+title: Temporal Fusion Transformer Forecasting in Python
 ---
 
 In summary Temporal Fusion Transformer (TFT) combines gating layers, an
@@ -469,4 +469,3 @@ mechanism has a great resemblence to a single attention layer, but it
 allows for $M$ multiple attention weights, and can be therefore be
 interpreted as the average ensemble of $M$ single attention layers.
 
-

nbs/docs/capabilities/cross_validation.ipynb

@@ -1,11 +1,19 @@
 {
  "cells": [
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "---\n",
+    "title: NeuralForecast Cross Validation API and Parameters\n",
+    "description: Understand NeuralForecast cross validation parameters, including horizons, windows, step size, validation size, test size, and refitting behavior.\n",
+    "---"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# Cross-validation | NeuralForecast\n",
-    "\n",
     ":::{.callout-warning collapse=\"true\"}\n",
     "## Prerequesites\n",
     "This Guide assumes basic familiarity with NeuralForecast. For a minimal example visit the [Quick Start](../getting-started/quickstart.html)\n",

nbs/docs/tutorials/cross_validation.ipynb

@@ -1,10 +1,19 @@
 {
  "cells": [
+  {
+   "cell_type": "raw",
+   "metadata": {},
+   "source": [
+    "---\n",
+    "title: Time Series Cross Validation Tutorial with NeuralForecast\n",
+    "description: Build a complete NeuralForecast cross validation workflow in Python, compare predictions with actual values, and evaluate models across rolling windows.\n",
+    "---"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# Cross-validation| NeuralForecast\n",
     "> Implement cross-validation to evaluate models on historical data"
    ]
   },
@@ -969,4 +978,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 4
-}
+}