Closed-Loop Supervised Fine-Tuning of Tokenized Traffic Models. CVPR Oral 2025.

Next-token prediction in JavaScript — build fast language and diffusion models.

Generative model for sequential recommendation based on Convolution Neural Networks (CNN))

[ICLR 2025] Where Am I and What Will I See : An Auto-Regressive Model for Spatial Localization and View Prediction

Curated collection of research on the limitations of next-token prediction and methods that go beyond it.

LLM pipeline: data→tokenizer→attention→GPT train/eval→instruction FT→sampling. Reproducible, clean configs, RTX-4060 defaults, ready for AMP/LoRA/DDP.

Implementation of a simplified GPT model trained on the Friends Series' dialogue dataset to generate dialogue-like text using a transformer architecture with multi-head attention and feed-forward networks. Includes training scripts, model definition, and text generation functionality.

Demonstrates, at very small scale, how a language model is trained (unigram language model).

A Next-Token Prediction Engine using LSTMs/RNNs. Demonstrates the foundational architecture of GenAI, focusing on sequence memory and perplexity optimization.natural-language-processing

Interactive visualization of next-word (token) prediction in GPT-style language models.

ngram language model next token prediction

Demonstrates, at very small scale, how a language model is trained (unigram language model).

Demonstrates, at very small scale, how a language model is trained (bigram language model).

Demonstrates, at very small scale, how a language model is trained (two-context model).

Demonstrates, at very small scale, how a language model is trained (bigram language model).

Demonstrates, at very small scale, how a language model is trained (two-context model).

🚀 Train a custom unigram model using simple and efficient methods, enabling easy adoption for natural language processing tasks.

Demonstrates, at very small scale, how a language model is trained (three-context model with explicit transform stage).

🤖 Visualize next-token prediction in GPT models with a minimal chat interface—focus on inference, not learning.

🐾 Train AI with 300 diverse contexts about animals to enhance understanding and interaction in natural language processing applications.