Artificial Intelligence on Nalar

Evaluating RAG Systems with a Small Golden Dataset

Tue, 01 Sep 2026 00:00:00 +0700

Retrieval-augmented generation (RAG) is easy to demo and surprisingly hard to evaluate. A fluent answer can hide weak retrieval, while a good retriever can be blamed for an answer model that ignores its evidence.

A useful evaluation process separates those failure modes. You do not need thousands of examples to begin. A carefully maintained golden dataset of 30 to 100 representative questions can catch many regressions before users do.

Define what the system is supposed to do

Start with the product contract rather than a model metric. For a documentation assistant, useful requirements might be:

Validate LLM Output with Structured Contracts

Tue, 01 Sep 2026 00:00:00 +0700

Large language models are useful when software needs to turn ambiguous text into a structured decision, extraction, or plan. The dangerous shortcut is to treat a model response as if it were already trusted application data.

Even when a provider can constrain output to JSON or a schema, the result can still be semantically wrong: a date can be impossible, an identifier can refer to a nonexistent record, or a supposedly positive amount can be negative. Reliable integrations therefore need a contract boundary between model output and the rest of the system.

Version Embeddings for Safe Semantic Search Migrations

Tue, 01 Sep 2026 00:00:00 +0700

Semantic search systems often look simple from the outside: encode a document, store its vector, encode a query, and compare the vectors. The operational difficulty appears later, when the embedding model changes.

Two models can produce vectors with the same dimension and still define completely different coordinate spaces. Mixing vectors from model A with query vectors from model B can silently destroy ranking quality without producing an obvious error.

The safe approach is to treat an embedding model as a versioned data dependency, not a drop-in function.

Running LLMs on Your Local Computer

Sun, 01 Sep 2024 00:00:00 +0700

Have you ever wanted to run a large language model (LLM) directly on your own computer without depending on a cloud service? Tools such as Ollama make it possible to manage and run language models locally from a laptop or desktop.

This article walks through the basic setup and shows how to run a model on Linux and macOS.

Installing Ollama on Linux

For Linux, the installation process is straightforward:

Object Detection with YOLOv8: Using a Pre-Trained Model on Images

Sat, 25 May 2024 00:00:00 +0700

In this article, we will experiment with YOLOv8 for object detection in images. YOLO (You Only Look Once) is one of the best-known object detection algorithms. YOLOv8 offers fast performance and accurate results. Fortunately, the ultralytics library makes it straightforward to use.

What You Need

Before you begin, make sure you have:

Python
The ultralytics library

If it is not installed yet, install it with pip:

$el.innerText = 'Copy', 2000)" class="absolute top-2 right-2 bg-neutral-700 text-white text-xs px-2 py-1 rounded opacity-0 group-hover:opacity-100 transition-opacity" > Copy

pip install ultralytics

Simple Code Example

Here is a short example that detects objects with YOLOv8:

Practical Guide to SMS Spam Detection with BERT and PyTorch

Sat, 25 May 2024 00:00:00 +0700

In this article, we will explore SMS spam detection with BERT. BERT (Bidirectional Encoder Representations from Transformers) is a widely used NLP model architecture that can be adapted to many text tasks, including classification. PyTorch and Hugging Face Transformers provide a convenient way to work with BERT models.

What You Need

Before you begin, make sure you have:

Python
PyTorch
Hugging Face Transformers
An SMS dataset, such as the SMS Spam Collection, if you plan to fine-tune a classifier

Install the required libraries with:

Real-Time Object Detection with YOLOv8 and OpenCV

Sat, 25 May 2024 00:00:00 +0700

YOLO (You Only Look Once) is one of the most popular object detection approaches. In this article, we will use YOLOv8 with OpenCV to perform real-time object detection through a webcam.

Prerequisites

Make sure you have:

Python 3.8+
The ultralytics library for YOLOv8
The opencv-python library for webcam access

Install the dependencies with pip:

$el.innerText = 'Copy', 2000)" class="absolute top-2 right-2 bg-neutral-700 text-white text-xs px-2 py-1 rounded opacity-0 group-hover:opacity-100 transition-opacity" > Copy

pip install ultralytics opencv-python

Example Code

The following Python example performs object detection from a webcam stream:

Using YOLOv8 for Object Detection with Labels and Confidence Scores

Sat, 25 May 2024 00:00:00 +0700

In this article, we will use YOLOv8 to detect objects in an image and print each object’s label and confidence score. YOLO (You Only Look Once) has long been a popular choice for object detection, and YOLOv8 provides a convenient Python API through the ultralytics package.

Here is how to use it.

Prerequisites

Make sure Python and the ultralytics library are installed on your system. If needed, install the package with pip: