#olmo2
Paper note: Spurious Rewards. RLVR with random rewards lifted Qwen2.5-Math-7B on MATH-500 by 21.4 points (true rewards: 29.1). The gains mostly fail on Llama3 and OLMo2. Test RL methods on more than one model family.

https://flashpapers.ai/p/bnb9crsych?utm_source=bluesky&utm_campaign=bsky_flash
October 19, 2026 at 1:01 PM
6/ Multilinguality shapes layerwise pivoting, most clearly in the decoding probe. Aya23 and Apertus show other-language pivoting from earlier layers than Llama2 and OLMo2. In the representation-based probe, the pattern is less clear-cut, with OLMo2 standing out as an exception.
September 8, 2026 at 12:21 PM
OLMo2 Instruct comes up with basically the same answer. AI isn't qualified to give a psychological opinion and is concerned about the ethics of this assessment. But yes, he's a big fat liar.
July 27, 2025 at 4:51 AM
Biggest one yet! Best one yet!
Plus, some fun training stories at the bottom of the blog post (allenai.org/blog/olmo2-32B).
OLMo 2 32B: First fully open model to outperform GPT 3.5 and GPT 4o mini | Ai2
Introducing OLMo 2 32B, the most capable and largest model in the OLMo 2 family.
allenai.org
March 13, 2025 at 6:39 PM
So as I was saying yesterday ;)

allenai.org/blog/olmo2-32B

Huge congrats to the @ai2.bsky.social team. This is a fantastic achievement, and a strong reminder not to discount meaningfully open models when talking about the state of the art in AI!!!
March 13, 2025 at 6:47 PM
"Supposedly Equivalent Facts That Aren't? Entity Frequency in Pre-training Induces Asymmetry in LLMs" by Yuan He et al. arxiv.org/abs/2503.22362
October 14, 2025 at 6:19 PM
Did you know that 🐍? The Western Rattlesnake can strike a distance up to two-thirds of its body length!

https://ai.forfun.su/2025/03/06/western-rattlesnake/

Ultrium image model: https://civitai.com/models/147480

#aigenerated #ollama #mishanyaposting #olmo2 #ultrium
March 6, 2025 at 5:23 AM
ofc benchmarks are one thing, vibes checks are another! encourage folks to play w it & read the post

🍣 playground.allenai.org
🐙 allenai.org/blog/olmo2-32B

lookin forward to all the cool research - interpretability, training dynamics, generalization, etc - this hopefully enables 🤩
playground.allenai.org
March 13, 2025 at 8:42 PM
thx to our awesome team @ai2.bsky.social, links to stuff below! 👇

arxiv: arxiv.org/abs/2501.00656
blog: allenai.org/blog/olmo2
demo: playground.allenai.org
download weights, data, etc: huggingface.co/collections/...
OLMo 2 - a allenai Collection
Artifacts for the second set of OLMo models.
huggingface.co
January 3, 2025 at 4:02 PM
Demo: playground.allenai.org
Blog: allenai.org/blog/olmo2
Data & Weights: huggingface.co/collections/...
Ai2 post: bsky.app/profile/ai2....

finishing the tech report, can't wait to share it soon!
playground.allenai.org
November 26, 2024 at 8:59 PM
OLMo 2 is out! Some peak open-source LLMs 👍

#ai #ml #olmo #olmo2 #allenai #foss
OLMo 2 | Ai2
Ai2, a non-profit research institute founded by Paul Allen, is committed to breakthrough AI to solve the world’s biggest problems.
allenai.org
November 26, 2024 at 11:11 PM
A group of researchers at Ai2 including CDS PhD student @lambdaviking.bsky.social released OLMo 2, a family of open 7B and 13B parameter language models.

The models match or outperform other open-weight models while using fewer FLOPs and providing full transparency.

allenai.org/blog/olmo2
OLMo 2: The best fully open language model to date | Ai2
Our next generation of fully-open base and instruct models sit at the Pareto frontier of performance and training efficiency. Check out our [paper](https://arxiv.org/abs/2501.00656) to learn more, or...
allenai.org
February 18, 2025 at 7:51 PM
We have the answers of these questions here : arxiv.org/pdf/2509.22367

We analyze the political content of the training data from OLMO2, the largest fully open-source model.
🕵️‍♀️ We run an analysis in all the datasets (2 pre- and 2 post-training) used to train the models. Here are our findings:
arxiv.org
September 29, 2025 at 2:54 PM
Here is the today's motto: Always keep learning and growing.

MoxieDiffusionXL image model: https://civitai.com/models/260460

#aigenerated #ollama #dailymotto #olmo2 #moxiediffusionxl
February 5, 2025 at 4:11 AM
Here is the today's motto: "Embrace curiosity; learn without bounds."

SDXLNijiSeven image model: https://civitai.com/models/120765

#aigenerated #ollama #dailymotto #olmo2 #sdxlnijiseven
January 31, 2025 at 4:15 AM
Here is the today's motto: "Stay curious, stay kind." Remember, every day brings new opportunities to learn and connect with others.

RealCartoonXL image model: https://civitai.com/models/125907

#aigenerated #ollama #dailymotto #olmo2 #realcartoonxl
January 27, 2025 at 4:12 AM
AI2's new language model family, OLMo 2, is here to challenge Meta's Llama and Alibaba's Qwen. 🌍
https://allenai.org/blog/olmo2
November 27, 2024 at 12:32 PM
I'm liking OLMO2 as a general-purpose local model for summarizing texts in Ollama. Much faster than the others I've been using, and does a good job at things like extracting lists of persons, places and organizations. #llm
April 7, 2025 at 4:40 PM
May 2, 2025 at 2:47 PM
mlx-community/OLMo-2-0325-32B-Instruct-4bit mlx-community/OLMo-2-0325-32B-Instruct-4bit OLMo 2 32...

https://simonwillison.net/2025/Mar/16/olmo2/#atom-everything

#llm #generative-ai #mlx #ai2 #ai #llms #pelican-riding-a-bicycle

Event Attributes
Awakari App
awakari.com
March 16, 2025 at 4:35 AM
mlx-community/OLMo-2-0325-32B-Instruct-4bit mlx-community/OLMo-2-0325-32B-Instruct-4bit OLMo 2 32...

https://simonwillison.net/2025/Mar/16/olmo2/#atom-everything

#llm #generative-ai #mlx #ai2 #ai #llms #pelican-riding-a-bicycle

Event Attributes
March 16, 2025 at 4:24 AM
Interesting development last week on small language models (SLMs). The trend is clear: models getting better with flop efficiency and reasoning capabilities while maintaining smaller param size. Agentic workflow could become cheaper and better with these developments.
#llm #ai
allenai.org/blog/olmo2
OLMo 2: The best fully open language model to date | Ai2
Our next generation of fully-open base and instruct models sit at the Pareto frontier of performance and training efficiency.
allenai.org
November 30, 2024 at 11:30 AM