Microsoft's MarkItDown

The MarkItDown library is a utility tool for converting various files to Markdown (e.g., for indexing, text analysis, etc.)

Repo: github.com/microsoft/ma...

For anyone interested in fine-tuning or aligning LLMs, I’m running this free and open course called smol course. It’s not a big deal, it’s just smol.

🧵>>

Struggling with RAG over PDF files?

You might want to give Docling a try.

𝗪𝗵𝗮𝘁'𝘀 𝗗𝗼𝗰𝗹𝗶𝗻𝗴?
• Python package by IBM
• OS (MIT license)
• PDF, DOCX, PPTX → Markdown, JSON

𝗪𝗵𝘆 𝘂𝘀𝗲 𝗗𝗼𝗰𝗹𝗶𝗻𝗴?
• Doesn’t require fancy gear, lots of memory, or cloud services
• Works on regular computers or Google Colab Pro

Improve the LLM inference with a long context time by up to 11x while preserving 95-100% of accuracy.

Nvidia's Star Attention: Efficient LLM Inference over Long Sequences

My deep learning course at the University of Geneva is available on-line. 1000+ slides, ~20h of screen-casts. Full of examples in PyTorch.

fleuret.org/dlc/

And my "Little Book of Deep Learning" is available as a phone-formatted pdf (nearing 700k downloads!)

fleuret.org/lbdl/

Excited to announce "BALROG: a Benchmark for Agentic LLM and VLM Reasoning On Games" led b UCL DARK's @dpaglieri.bsky.social! Douwe Kiela plot below is maybe the scariest for AI progress — LLM benchmarks are saturating at an accelerating rate. BALROG to the rescue. This will keep us busy for years.

LLMs generate novel word sequences not contained in their pretraining data. However, compared to humans, models generate significantly fewer novel n-grams.

RLHF = 30% *more* copying than base!

Awesome work from the awesome Ximing Lu (gloriaximinglu.github.io) et al. 🤩

arxiv.org/pdf/2410.04265

Hot take: if you believe that talk therapy is useful, you have to believe that LLMs will eventually be the best and most available therapists

Just realized BlueSky allows sharing valuable stuff cause it doesn't punish links. 🤩

Let's start with "What are embeddings" by @vickiboykis.com

The book is a great summary of embeddings, from history to modern approaches.

The best part: it's free.

Link: vickiboykis.com/what_are_emb...

Why are some LLMs better at chess than others

Part 1: dynomight.net/chess/

Part 2: dynomight.net/more-chess/

What's the secret sauce of SmolLM2 to beat LLM titans like Llama3.2 and Qwen2.5?

Unsurprisingly: data, data, data!

The SmolTalk is open and available here: huggingface.co/datasets/Hug...

Free eBook: Machine Learning Systems by Vijay Janapa Reddi

Principles and Practices of Engineering Artificially Intelligent Systems

mlsysbook.ai

Oh wow!

A surprising result from Databricks when measuring embeddings and rerankers on internal evals.

1- Reranking few docs improves recall (expected).
2- Reranking many docs degrades quality (!).
3- Reranking too many documents is quite often worse than using embedding model alone (!!).

DeepSeek-R1-Lite-Preview Test Number 1

All the things you need to know to pretrain an LLM at home*!

Gave a workshop at Uni Bern: starts with scaling laws and goes to web scale data processing and finishes training with 4D parallelism and ZeRO.

*assuming your home includes an H100 cluster