Frontier AI Models Evaluation Benchmarks
A guide to frontier AI model benchmarks in 2026, covering MMLU, GPQA Diamond, HLE, SWE-bench, ARC-AGI-2, MMMU, Arena Elo, etc. What each benchmark measures, which models lead, why scores saturate.
Harshit Kumar
Machine Learning Engineer specializing in deep learning, computer vision, and NLP. Technical blog and portfolio by Harshit Kumar.
Latest posts
Introduction to Model Context Protocol (MCP)
MCP is an open-source protocol that standardizes how LLMs connect to external tools and data sources, replacing fragile custom integrations with a common interface.
Evaluation Metrics for Large Language Models
Walkthrough of evaluation metrics for large language models: perplexity, cross-entropy, BLEU, ROUGE, METEOR, CIDEr, BERTScore, RAG metrics, safety metrics, and LLM-as-a-judge, with equations and visualizations.
Prompt Engineering Techniques: How to Write Effective Prompts
A deep-dive into prompt engineering techniques from few-shot prompting and chain-of-thought, ReAct, and prompt injections with examples.
Distributed Training: How to train Large Language Models (LLM)
Comprehensive guide to distributed training for LLMs covering data parallelism, model parallelism, tensor parallelism, ZeRO optimizer, FSDP, 3D parallelism, DeepSpeed with interactive visualization, code examples.
Vision Language Models (VLM)
Overview of Vision Language Models (VLMs) and their training paradigms: contrastive learning (CLIP), masking (FLAVA), generative approaches (CoCa, Chameleon), and pretrained backbone methods (Frozen, LLaVA, BLIP-2).
Matrix Multiplication in CUDA
Implementing matrix multiplication in CUDA from a naive CPU baseline to GPU-accelerated versions using tiled shared memory for deep learning workloads.
Retrieval Augmented Generation (RAG) Chatbot for 10Q Financial Reports
Building a RAG-based chatbot for 10Q financial reports to reduce LLM hallucinations by grounding answers in retrieved document context.
Mixed Precision and Quantization: Accelerating Deep Learning Training and Inference
Comprehensive guide to mixed precision training (FP16/FP32) and INT8 quantization, covering GPU architecture, Tensor Cores, loss scaling, AMP, PTQ, QAT, and layer fusion with practical code examples.
PyTorch Basic Tutorial
A practical introduction to PyTorch covering tensors, autograd, neural network modules, and key libraries like torchvision and torchaudio.