<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Allen Institute for AI — FoundationIndex</title>
    <link>https://foundationindex.ustunozgur.com/s/allenai</link>
    <description>The Seattle-based nonprofit research institute founded by Paul G. Allen in 2014 that builds fully open language, multimodal, and reasoning models—publishing not only model weights (OLMo, Molmo, Tülu) but also complete pretraining corpora (Dolma), training code, intermediate checkpoints, and reinforcement learning recipes.</description>
    <language>en</language>
    <atom:link href="https://foundationindex.ustunozgur.com/s/allenai/feed.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Allen Institute for AI · Release · 2026-10-01</title>
      <link>https://foundationindex.ustunozgur.com/s/allenai#allenai-olmo-core-3-moe-2026</link>
      <guid>https://foundationindex.ustunozgur.com/s/allenai#allenai-olmo-core-3-moe-2026</guid>
      <pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Ai2 released Olmo-core 3, an open training framework for large mixtures of experts. Expert parallelism spreads expert weights across devices so the model can grow beyond the memory of a single device.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://allenai.org/blog/olmocore3&quot;&gt;Introducing Olmo-core 3: Open, scalable training infrastructure for large MoEs&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Allen Institute for AI · Release · 2025-11-20</title>
      <link>https://foundationindex.ustunozgur.com/s/allenai#allenai-olmo3-model-flow-2025</link>
      <guid>https://foundationindex.ustunozgur.com/s/allenai#allenai-olmo3-model-flow-2025</guid>
      <pubDate>Thu, 20 Nov 2025 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Ai2 released the 7B and 32B Olmo 3 family with intermediate checkpoints, data, and recipes for successive training stages. The artifacts let researchers inspect how base models became instruction-tuned or reasoning models.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://allenai.org/blog/olmo3&quot;&gt;Olmo 3: Charting a path through the model flow&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Allen Institute for AI · Paper · 2024-11-22</title>
      <link>https://foundationindex.ustunozgur.com/s/allenai#allenai-tulu3-rlvr-post-training-2024</link>
      <guid>https://foundationindex.ustunozgur.com/s/allenai#allenai-tulu3-rlvr-post-training-2024</guid>
      <pubDate>Fri, 22 Nov 2024 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Ai2 published the Tülu 3 post-training recipe, combining supervised fine-tuning and direct preference optimization with reinforcement learning from verifiable rewards. For tasks with checkable answers or instructions, the recipe could derive rewards from a verifier.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://arxiv.org/abs/2411.15124&quot;&gt;Tülu 3: Pushing Frontiers in Open Language Model Post-Training&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Allen Institute for AI · Paper · 2024-09-25</title>
      <link>https://foundationindex.ustunozgur.com/s/allenai#allenai-molmo-pixmo-multimodal-2024</link>
      <guid>https://foundationindex.ustunozgur.com/s/allenai#allenai-molmo-pixmo-multimodal-2024</guid>
      <pubDate>Wed, 25 Sep 2024 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Following the September 3 release of OLMoE (a 7B-total, 1B-active open Mixture-of-Experts language model), Ai2 introduced Molmo and the PixMo dataset, training open vision-language models (1B–72B) from scratch on human speech-transcribed image descriptions and 2D pixel-coordinate pointing data without distilling from proprietary APIs.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://arxiv.org/abs/2409.17146&quot;&gt;Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Multimodal Models&lt;/a&gt;, &lt;a href=&quot;https://arxiv.org/abs/2409.02060&quot;&gt;OLMoE: Open Mixture-of-Experts Language Models&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Allen Institute for AI · Release · 2024-02-01</title>
      <link>https://foundationindex.ustunozgur.com/s/allenai#allenai-olmo-dolma-open-release-2024</link>
      <guid>https://foundationindex.ustunozgur.com/s/allenai#allenai-olmo-dolma-open-release-2024</guid>
      <pubDate>Thu, 01 Feb 2024 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Ai2 launched OLMo (1B and 7B) alongside the 3-trillion-token Dolma open corpus, training code, logs, and hundreds of intermediate Pre-Training Checkpoints under Apache 2.0, enabling researchers to audit data attribution, training dynamics, and memorization across the full pretraining lifecycle.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://allenai.org/blog/olmo-open-language-model-87ccfc95f580&quot;&gt;OLMo: Open Language Model&lt;/a&gt;, &lt;a href=&quot;https://arxiv.org/abs/2402.00838&quot;&gt;OLMo: Accelerating the Science of Language Models&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Allen Institute for AI · Paper · 2018-02-15</title>
      <link>https://foundationindex.ustunozgur.com/s/allenai#allenai-elmo-contextual-embeddings-2018</link>
      <guid>https://foundationindex.ustunozgur.com/s/allenai#allenai-elmo-contextual-embeddings-2018</guid>
      <pubDate>Thu, 15 Feb 2018 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Ai2 and University of Washington researchers (Matthew E. Peters et al.) published Deep contextualized word representations (ELMo), deriving token Embeddings from the internal states of a deep bidirectional LSTM language model so a word&amp;apos;s vector shifts dynamically with its surrounding sentence context—establishing the pretrain-then-adapt paradigm shortly before BERT.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://arxiv.org/abs/1802.05365&quot;&gt;Deep contextualized word representations&lt;/a&gt;&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
