RSSAmplifier

Blog

Nihal Jain

Recent content on Nihal Jain

nihaljn.github.ioRSS feed ↗14 posts

Latest posts

Nature

I love nature and am regularly blown away by the beauty, diversity, adaptavity and scale of natural phenomena. This is a catalogue of some facts about nature I learned recently and find fascinating. The Saguaro are one of the largest cacti in the world. Found mostly in west US, they can grow up to 40 feet in height, and store up to 2 tonnes of water. They can live up to 200 years of age……

Reinforcement Learning

Algorithms for training LLMs There are two frameworks that are commonly used for (post-)training LLMs with RL: Reinforcement Learning from Human Feedback (RLHF) trains the LLM using RL with rewards derived from a reward model trained on human preferences; Reinforcement Learning with Verifiable Rewards (RLVR) trains the LLM using RL with rewards derived from rule-based or deterministic verifiers.…

Stumbling on Happiness

In his book on human psychology and neuroscience, with a blend of unexpected humor, Gilbert says our happiness is dependent on our ability to predict the future, i.e., to imagine . For e.g., in choosing among options of which city to move to, we imagine how happy we are likely to be in any of these places in the future , affecting our present happiness. Thus, imagination is very load-bearing, so…

Quantum Computing

Selected notes from COMS 4281 Introduction to Quantum Computing taught by Prof. Henry Yuen in Fall 2025. Qubits A quantum bit (or qubit) is the basic unit of information in quantum computing. Unlike classical bits that are exclusively 0 or 1, a qubit exists in a superposition of both states. However, upon measurement, the qubit collapses to a classical bit value. Mathematically, a qubit state is…

On Mitigating Code LLM Hallucinations with API Documentation

The Amazon Nova Family of Models

CrossCodeEval: A Diverse and Multilingual Benchmark for Cross-File Code Completion

ContraCLM: Contrastive Learning For Causal Language Model

MultiViz: Towards Visualizing and Understanding Multimodal Models

Self-supervised Multi-view Disentanglement for Expansion of Visual Collections

Multimodal prompting for image generation

Enabling the use of multiple modalities while prompting Stable Diffusion.

Generating Compositional Color Representations from Text

Inspiration Retrieval for Visual Exploration

A novel approach to use semantic segmentation based deep learning networks to classify multi-temporal SAR data