Chris Mayfield

Blogger, Excited Gamer and has an unhealthy relationship to RULES

[Chris Mayfield @abandonrules]

Wonder what this is?

[Chris Mayfield @abandonrules]

Christmas Ornament Has Hidden Compartment, Clever Design via @hackaday https://t.co/lAwaRGMDfR

[Sundar Pichai @sundarpichai]

What a finish! Gemini 2.5 Pro just completed Pokémon Blue!  Special thanks to @TheCodeOfJoel for creating and running the livestream, and to everyone who cheered Gem on along the way.

[Chris Mayfield @abandonrules]

A Programming Language For Building NES Games via @hackaday https://t.co/nwEQBlIE4P

[Thread Reader App @threadreaderapp]

@jiayi_pirate Your thread is creating a buzz! #TopUnroll https://t.co/p0GZkF7Tpk 🙏🏼@vankous for 🥇unroll

[Jiayi Pan @jiayi_pirate]

We reproduced DeepSeek R1-Zero in the CountDown game, and it just works

Through RL, the 3B base LM develops self-verification and search abilities all on its own

You can experience the Ahah moment yourself for < $30
Code: https://t.co/UcGKN2SVGj

Here's what we learned 🧵

[Chris Mayfield @abandonrules]

This QR Code Leads To Two Websites, But How? via @hackaday https://t.co/QpRtXOcmq4

[Chris Mayfield @abandonrules]

3D-Printed RC Car Focuses On Performance Fundamentals via @hackaday https://t.co/ti7x0qggHk

[Chris Mayfield @abandonrules]

USB-C For Hackers: Reusing Cables via @hackaday https://t.co/xHgxmYxVh5

[Chris Mayfield @abandonrules]

@stickermule @rickyberwick So where is the story? I need details

[Sticker Mule @stickermule]

Our last CEO f'd things up. I'm fixing them.

Introducing white label hot sauce.

Go fucking buy it now.

https://t.co/dnBsPvbK8f

Ricky Berwick (@rickyberwick)
CEO, Sticker Mule

P.S. 20 people who repost this win $500. Follow us so we can DM you.

[Chris Mayfield @abandonrules]

@AmabelHolland This article brought me here. https://t.co/ER6a1jpNOz

[Jimmy Fallon @jimmyfallon]

Wedding RSVPs should have options for “Yes,” “No,” and “I feel obligated.”

[Chris Mayfield @abandonrules]

New Part Day: A Hackable Smart Ring via @hackaday https://t.co/y2tjFLKAyp

[elvis @omarsar0]

Nemotron-4 340B is a huge release by NVIDIA!

Why?

The Nemotron-4 340B instruct model lets you generate high-quality data and then the reward model (also released) can filter out data on several attributes.

It's often not enough just to have a good instruct model for generating high-quality synthetic data. You also need to filter that generated data on a set of requirements. It's common to use the same model to grade the outputs (LLM-as-Judge) but the reward model (Reward-Model-as-Judge), which is often not released, is ideal for this.

Also, check out in the paper how Nemotron-4-340B compares with GPT-4-1106-preview when evaluated by humans on several tasks.

Seems that GPT-4 is a lot better at rewrite and extraction but Nemotron-4-340B compared nicely on other tasks. Multi-turn chat has a high score!

Something else that caught my attention is that the majority of preference data used for alignment are synthetic.

The results show that Nemotron-4 340B is a strong model. Check out those MMLU, GSM8K, and Arena Hard numbers.

GPT-4 is still ahead but the gap keeps closing.

What's Meta's move now? Are we going to see a similar type of release? Could be a huge moment for the field and open LLMs.

Nice release from NVIDIA. They even released a preference dataset (link in the replies). Well done to the team!

[Chris Mayfield @abandonrules]

After I recently finished building my new RatRig V-Core 3, I thought it would be a great test for ... https://t.co/cppjOE1Bku

[Chris Mayfield @abandonrules]

https://t.co/XUVQxonaST

[Chris Mayfield @abandonrules]

A Scope Test Tool You Can Build With Just A Pico via @hackaday https://t.co/EJa9Upr55l

[Chris Mayfield @abandonrules]

Optimizing database queries in Django for improved performance by @s_emanuilov https://t.co/3L03Ogkuir