Author

ML Point

@ml-point

articles: 32
total claps: 1,707
top claps: 263
avg read: 3.9 min
paywalled: 32
active: 2025-09-20 → 2026-07-06

Wiki topics

LLM · Large Language Models · 7 AI · AI · General · 7 MM · Multimodal & Generative Media · 6 EDU · Education & Learning · 5 ML · Machine Learning · 5 AGT · AI Agents · 4 RAG · RAG & Retrieval · 4 📐 · Mathematics · 4

Top tags

#transformers ·6 #reasoning ·4 #retrieval-augmented-gen ·3 #ai ·3 #llm ·3 #generalization ·3 #noise ·3 #diffusion-models ·3 #agentic-ai ·2 #context ·2

HUM AI

ML ML Point · Towards AI

Supervised Reinforcement Learning: From Expert Trajectories to Step-wise Reasoning (Paper Review)

Non-members can read for free

Nov 01, 2025 · 6 min read · 263 claps paywalled #trajectory #reinforcement-learning #reasoning

Supervised Reinforcement Learning: From Expert Trajectories to Step-wise Reasoning (Paper Review)

ML ML Point · Towards AI

When Transformers Multiply Their Heads: What Increasing Multi-Head Attention Really Does

Transformers have become the backbone of many AI breakthroughs, in NLP, vision, speech, etc. A central component is multi-head…

Oct 14, 2025 · 8 min read · 220 claps paywalled #multi-head-attention #transformers #memory-usage

When Transformers Multiply Their Heads: What Increasing Multi-Head Attention Really Does

AI TCH

ML ML Point · Towards AI

DeepSpeed is Revolutionizing LLM Training

Non-members can read for free

Nov 27, 2025 · 3 min read · 144 claps paywalled #deepspeed #llm #microsoft

DeepSpeed is Revolutionizing LLM Training

AI HUM SCI

ML ML Point · Data And Beyond

Diffusion Models in Machine Learning: From Chaos to Creation

1. Introduction: The Art of Noise

Sep 20, 2025 · 7 min read · 116 claps paywalled #diffusion-models #probability #noise

Diffusion Models in Machine Learning: From Chaos to Creation

AI ECO

ML ML Point

Agent Memory: How AI Systems Move Beyond Stateless Intelligence

Agent memory refers to the ability of an AI system (an “agent”) to store, retrieve, and use information over time to improve future…

Jun 13, 2026 · 2 min read · 101 claps paywalled #agents #memory-management #context

Agent Memory: How AI Systems Move Beyond Stateless Intelligence

AI SCI

ML ML Point

LocateAnything by Nvidia AI

Recent vision-language models rely on a single concept: treating absolutely everything as a text generation problem. Images, user prompts…

Jun 04, 2026 · 4 min read · 100 claps paywalled #vision-language-model #bounding-box #localization

ML ML Point

Do Rude Prompts Make AI More Accurate?

Among the many debates across AI communities and research forums, one of the most surprising is the suggestion that rude prompts may…

May 30, 2026 · 4 min read · 100 claps paywalled #clarity #communication #rudeness

SOC AI

ML ML Point

What is Grokking?

Introduction

Jun 09, 2026 · 2 min read · 100 claps paywalled #grokking #interpretability #overfitting

HUM GEN AI

ML ML Point

The Learning Rate: The Number That Decides Whether a Model Learns or Fails

There is a moment in every training run when everything hangs in balance.

Feb 12, 2026 · 4 min read · 53 claps paywalled #learning-rate #progress #convergence

The Learning Rate: The Number That Decides Whether a Model Learns or Fails

ML ML Point

Types of RAG

For a long time, people assumed the intelligence of modern AI came mainly from larger models and more training data.

May 07, 2026 · 7 min read · 50 claps paywalled #retrieval-augmented-gen #evolution #search

ML ML Point

The Questions We Don’t Ask About Modern AI

Source Image

Apr 30, 2026 · 4 min read · 50 claps paywalled #baseline #failure-mode #approximation

The Questions We Don’t Ask About Modern AI

AI TCH

ML ML Point

Speculative Decoding (Technical Explanation)

Large language models generate text autoregressively, meaning:

May 08, 2026 · 4 min read · 50 claps paywalled #speculative-decoding #acceptance-rate #kv-cache

Speculative Decoding (Technical Explanation)

ML ML Point

Generative Refinement Networks (GRN): Toward More Efficient Visual Generation

The progress in image generation over the past few years has been undeniable. Models can now produce visuals that are often…

Apr 19, 2026 · 4 min read · 50 claps paywalled #generative-ai-solution #text-to-image-generation #text-to-video

Generative Refinement Networks (GRN): Toward More Efficient Visual Generation

ML ML Point

CVPR 2026: Research Direction from Accepted Work

Source Image

Jun 05, 2026 · 4 min read · 50 claps paywalled #vision-language-model #foundation-models #embodied-ai

CVPR 2026: Research Direction from Accepted Work

ML ML Point

The Top 9 Agentic LLM Workflows

For years, we have observed the dominant interaction pattern.

Jun 16, 2026 · 4 min read · 50 claps paywalled #llm #planning #execution

ML ML Point

The Physical Reality of Computation

1. AI and Computing Hardware

May 24, 2026 · 4 min read · 50 claps paywalled #computation #data-center #infrastructure

ML ML Point

Impact of Latency During Inference

Latency during inference is the time it takes a model to process data and deliver a result. It is a critical system constraint that…

Jun 03, 2026 · 3 min read · 50 claps paywalled #latency #gpu-computing #distributed-systems

AI LIF

ML ML Point · The Startup

Relationship of Finetuning Capacity and Model Size: Adaptation without Regret

Fine-tuning is where a pre-trained model transforms from a generalist to a specialist. But, if your fine-tuning capacity doesn’t match your…

Oct 02, 2025 · 3 min read · 50 claps paywalled #lora #model-finetuning #parameters-and-arguments

Relationship of Finetuning Capacity and Model Size: Adaptation without Regret

ML ML Point

Chain-of-Thought Is Not Reasoning

Most people believe that if a model explains its steps, it must be reasoning. Chain-of-Thought (CoT) strengthened this belief. It improved…

May 06, 2026 · 3 min read · 50 claps paywalled #chain-of-thought #reasoning #transformers

SCI

ML ML Point

The Problem Swin Transformer Solves

The biggest challenge with using standard Transformers for images is the math. Standard self-attention compares every single pixel to every…

Jun 19, 2026 · 3 min read · 8 claps paywalled #yolo #transformers #convolutional-network

ML ML Point · Towards AI

Inter-GPU Communication

Introduction

Oct 02, 2025 · 4 min read · 1 claps paywalled #distributeddataparallel #nccl #latency

AI HUM

ML ML Point

Munich 1991: The Foundation for the Most of Today’s AI Revolution

Artificial intelligence is today popular for modern breakthroughs like ChatGPT, large language models, and autonomous vehicles. However…

Jul 06, 2026 · 3 min read · 1 claps paywalled #transformers #neural-networks #deep-learning

Munich 1991: The Foundation for the Most of Today’s AI Revolution

ML ML Point

Why is the World’s Digital Economy Splitting Apart?

The World Isn’t Building One Digital Future

Jun 06, 2026 · 6 min read · 0 claps paywalled #digital-economy #ai-governance #fragmentation

ML ML Point

The Forgotten Twin of the Transformer

In June 2017, Google published two papers only four days apart.

Jun 04, 2026 · 4 min read · 0 claps paywalled #transformers #generalization #attention

AI SCI LIF

ML ML Point

Why AI Images Look Real But They Aren’t

The AI-generated look strange because reality is harder than appearance. Most discussions about these images focus on visual mistakes.

May 21, 2026 · 4 min read · 0 claps paywalled #linear-perspective #projective-geometry #diffusion-models