Doug Shipman

Love the outdoors, adventures, helping others, data science, astronomy, ML, & AI.

[World of Statistics @stats_feed]

Number of Super Bowl wins, by country:

🇺🇸 USA: 59
🇨🇦 Canada: 0
🇦🇺 Australia: 0
🇨🇳 China: 0
🇮🇳 India: 0
🇯🇵 Japan: 0
🇬🇧 UK: 0
🇷🇺 Russia: 0
🇩🇪 Germany: 0
🇪🇸 Spain: 0
🇮🇹 Italy: 0

[elvis @omarsar0]

Large Language Diffusion Models (LLaDA)

Proposes a diffusion-based approach that can match or beat leading autoregressive LLMs in many tasks.

If true, this could open a new path for large-scale language modeling beyond autoregression.

More on the paper:

Questioning autoregressive dominance
While almost all large language models (LLMs) use the next-token prediction paradigm, the authors propose that key capabilities (scalability, in-context learning, instruction-following) actually derive from general generative principles rather than strictly from autoregressive modeling.

Masked diffusion + Transformers
LLaDA is built on a masked diffusion framework that learns by progressively masking tokens and training a Transformer to recover the original text. This yields a non-autoregressive generative model—potentially addressing left-to-right constraints in standard LLMs.

Strong scalability
Trained on 2.3T tokens (8B parameters), LLaDA performs competitively with top LLaMA-based LLMs across math (GSM8K, MATH), code (HumanEval), and general benchmarks (MMLU). It demonstrates that the diffusion paradigm scales similarly well to autoregressive baselines.

Breaks the “reversal curse”
LLaDA shows balanced forward/backward reasoning, outperforming GPT-4 and other AR models on reversal tasks (e.g. reversing a poem line). Because diffusion does not enforce left-to-right generation, it is robust at backward completions.

Multi-turn dialogue and instruction-following
After supervised fine-tuning, LLaDA can carry on multi-turn conversations. It exhibits strong instruction adherence and fluency similar to chat-based AR LLMs—further evidence that advanced LLM traits do not necessarily rely on autoregression.

https://t.co/8LNlzq0VoR

[Phys.org @physorg_com]

#A spinning #AtomicNucleus is confirmed as a quantum resource, challenging classical assumptions. This breakthrough enhances understanding of quantum phenomena in spin precession, with implications for quantum computing. @NUSingapore https://t.co/9FQo1LSAsd https://t.co/vQUplpJWVC

[Doug Shipman @explr77]

My creative type is Maker, and we're known for our focus and dedication. Take the #mycreativetype quiz to discover your type. https://t.co/T4opYGjKOv

[Aishwarya Nevrekar @nevrekaraishwa2]

Data Analysis Projects for beginners with source code
in Python😲

Part 1

A thread 🧵👇

[Quanta Magazine @QuantaMagazine]

Behold! President Biden has unveiled the first image from the James Webb Space Telescope, which launched in December. Here, we see the deep field SMACS 0723, in which a galaxy acts as a gravitational lens, magnifying thousands of extremely distant galaxies.

[Doug Shipman @explr77]

#heroworkout #wod @midnightmonsterfit Week 5 done. Another great workout! Looking forward to the next! https://t.co/nPmuCYl2Cl

[/MachineLearning @slashML]

How to compute the probability of trajectories term in Stochastic Gradient Meta Reinforcement Learning https://t.co/QcUgbRQmAe

[World of Engineering @engineers_feed]

Mathematical puzzle. Find the missing number.

[Astra @Astra]

We can confirm the successful deployment of the satellites on Spaceflight's Astra-1 mission today. We are grateful to our partner @spaceflightinc and incredibly proud of this mission success. #AdAstra

[Astra @Astra]

We had a nominal first and upper stage flight and are awaiting payload deployment confirmation. Stay tuned for updates.

[Peggy Whitson @AstroPeggy]

Sometimes it is difficult to visualize change over time. Take a look at this one….watch to the end.

[Saurav Jain (Open Source + Communities) @Sauain]

Seven Machine Learning Algorithms / Models and a short explanation about each 🤖

Thread 🧵👇

[Doug Shipman @explr77]

@midnightmonsterfit #herochallenge #wod Week 4 in the books. A little behind, but going to catch up. This was one of my favorites, so far. Killer workout! https://t.co/0sUBWU7v9a

[Spaceflight @SpaceflightInc]

An overview of who we are in 30 seconds? You got it.

[Astra @Astra]

Due to weather, now targeting tomorrow, Tues., March 15 at 9:22 AM PDT for the Astra-1 @Spaceflightinc mission

[Doug Shipman @explr77]

Your Brain on Running, and How to Get Higher - Outside Online https://t.co/G5THDboh4o

[Doug Shipman @explr77]

@sonukg4india I get the 9 answer, but I measured them out. 5x=36degrees = 7.2, 7x=57deg = 8.14, and 3x=18deg = 6. I know, I know.

[Doug Shipman @explr77]

@engineers_feed 0-4-2