Is the latent reasoning entangled?
Does the latent embedding entirely hold the puzzle logic?
Read article
Blogs
We explore the technology, science, and research behind Artificial Intelligence. Our goal is to break down complex ideas into clear, thoughtful insights that make AI easier to understand and discuss.
2026
4 articlesA Token Is Worth a Thousand Objects (VLMs for Object Detection and Segmentation)
Can vision-language models do more than describe an image? Explore how VLMs can identify objects and enable precise visual segmentation.
Read article
From Random Actions to Balance: Building a Reinforcement Learning Agent for the CartPole Problem
In this project, I built a reinforcement learning agent to solve the classic CartPole problem.
Read article
Machines Can Transcribe Speech, But Can They Understand It?
Does emotion play a role in understanding how something was said?
Read article
2025
3 articlesThrough the Lenses - A Circuit Odyssey
Why outputs from techniques like LogitLens need a careful relook.
Read article
Distiling Truth from Falsehoods
Extracting truthful representations even when the model is trained to respond untruthfully.
Read article
Zero to Mamba
Mamba is a model that was designed to do sequence modelling tasks effectively and is claimed to outperform transformer based models of similar parameter size.
Read article
2024
14 articlesAlphaZero: A Revolutionary Approach to Game-Playing AI
Games have long been a testing ground for artificial intelligence, offering structured environments with clear rules and measurable outcomes.
Read article
Genie 2: A large-scale foundation world model
This blog delves into Genie 2’s transformative features, showcasing how it redefines interactivity, realism, and creativity in gaming.
Read article
Paper Review : Search Augmented Instruction Learning
LLMs face a significant challenge: they often operate on static datasets that quickly become outdated.
Read article
Exploring KANs : A Unique Approach to Function Approximation in AI
We have come to accept Multi-Layer Perceptrons (MLPs) as the foundational building blocks of neural networks.
Read article
Paper Review : SoccerNet
For football enthusiasts, having access to a comprehensive breakdown of a match, even when not witnessing it live, is invaluable.
Read article
MINECRAFT AND AI: Pushing the Boundaries of GenAI, Block by Block
MineRL provides an OpenAI Gym interface to Minecraft, allowing researchers to develop and test AI models within the game.
Read article
Paper Review : GamenGen
The objective of this paper review is to analyse the working of the model without diving excessively into the heavy maths that drives it.
Read article
Microsoft’s GraphRAG: From Local to Global Query Focused Summarisation
Retrieval Augmented Generation(RAG) is the most famous term in AI these days.
Read article
Paper Review : Conformers
Conformer: Advancing the future of Automatic Speech Recognition
Read article
Emergent Abilities of Large Language Models
The Natural Language Processing (NLP) field witnessed a groundbreaking moment with the introduction of transformers.
Read article
NLPs and LLMs for the Indian Languages
Why is there a huge need to develop Natural Language Processing techniques for Indian languages today?
Read article
Gemini: A Journey into LLMs and Words
LLMs and AI like Gemini paint a vivid picture of a future brimming with possibilities.
Read article
Mixtral 8x7B: A Breakthrough in Sparse Mixture of Experts Models
Mixtral 8x7B is an open-source answer to these problems.
Read article
2023
2 articlesChatGPT: What does it do? How does it work? And Should you be worried?
So, all in all, chatGPT is wonderful but not without its shortcomings.
Read article
Learning to Paint with Deep RL
An interesting Paper Review along with our own implementation of the paper
Read article