Reframes model selection as data compression: the best hypothesis is the one that lets you describe the data in the fewest bits. Walks through MDL twice — once conceptually, once with full math — turning Occam's razor into a usable inference principle.
Condenses Stanford's CS 229 into one-page visual cheatsheets spanning supervised, unsupervised, and deep learning, plus probability and linear-algebra refreshers. Available in 10+ languages, with all topics merged into one Super VIP PDF.
AI Train2020
PyTorch library for operator learning: neural networks that map between whole function spaces, not fixed grids, so a model trained at one resolution runs at any other. Bundles FNO, Tensorized FNO and related architectures, mainly for solving PDEs.
AI Dataset2022
Benchmark dataset of ~8.5k grade-school math word problems with step-by-step solutions and calculator annotations for evaluating multi-step arithmetic reasoning in language models. Provided in two configs (main and socratic) and commonly used for chain-of-thought prompting, fine-tuning, and verifier training.
Breaks hard reasoning problems into simpler subproblems, then solves them using earlier answers. With 14 exemplars, code-davinci-002 reaches at least 99% accuracy on SCAN splits, including the length split.
Builds a single rigorous theory from one question: why some bit strings look random. Defines plain and prefix complexity, the incompressibility method, and Martin-Löf randomness, tying information content to whether a short program can reproduce a string.
AI Dataset2022
Contains 40,000 teacher-generated reasoning traces distilled from the Qwen3.8-27B model for supervised fine-tuning and analysis. Covers code, math, science and logic; each example pairs a <think> chain-of-thought with a final response and is distributed in JSONL/Parquet for SFT workflows.
Improves language-model problem solving by generating candidate answers, retaining verified successes, and fine-tuning on them. Repeats this loop with an expectation-maximization interpretation instead of collecting more human-written solutions.
Reaches 51.7% on the competition-level MATH benchmark with a 7B model and no tools or voting, rivaling Gemini-Ultra and GPT-4. Built on a 120B-token math corpus mined from Common Crawl, and introduces GRPO, a memory-efficient PPO variant for reasoning.
AI Dataset2024
A mixed instruction dataset for SFT and RLHF research that combines chat, math, code and instruction-following samples from multiple public datasets under an Apache-2.0-compatible license; intended for instruction tuning and evaluation.
Explains how base-model behavior and GRPO normalization distort claims about reasoning gains. Introduces Dr. GRPO to reduce length bias, with a 7B training recipe reaching 43.3% on AIME 2024.
AI Dataset2025
Open synthetic corpus for training small reasoning-focused language models — ~79.65M generated samples (≈75B tokens with Pleias tokenizer) amplified from ~58.7k Wikipedia/Wikibooks seeds; includes explicit synthetic reasoning traces, multilingual coverage, and parquet splits.