<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Synthetic Data &amp; Self-Play — FoundationIndex</title>
    <link>https://foundationindex.ustunozgur.com/s/synthetic-data</link>
    <description>Using AI models to generate, filter, and verify new training examples—textbooks, code solutions, or reasoning traces—when high-quality human data runs short or needs targeted coverage.</description>
    <language>en</language>
    <atom:link href="https://foundationindex.ustunozgur.com/s/synthetic-data/feed.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Synthetic Data &amp; Self-Play · Release · 2024-11-28</title>
      <link>https://foundationindex.ustunozgur.com/s/synthetic-data#synthetic-data-cosmopedia-2024</link>
      <guid>https://foundationindex.ustunozgur.com/s/synthetic-data#synthetic-data-cosmopedia-2024</guid>
      <pubDate>Thu, 28 Nov 2024 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Hugging Face&amp;apos;s HuggingFaceTB team published Cosmopedia, an open Apache-2.0 synthetic pre-training corpus of over 30 million files and 25 billion tokens of generated textbooks, stories, and guides.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://huggingface.co/datasets/HuggingFaceTB/cosmopedia&quot;&gt;Cosmopedia Dataset on Hugging Face&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Synthetic Data &amp; Self-Play · Paper · 2024-06-17</title>
      <link>https://foundationindex.ustunozgur.com/s/synthetic-data#synthetic-data-nemotron4-2024</link>
      <guid>https://foundationindex.ustunozgur.com/s/synthetic-data#synthetic-data-nemotron4-2024</guid>
      <pubDate>Mon, 17 Jun 2024 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;NVIDIA released Nemotron-4 340B with a permissive open license specifically aimed at synthetic data generation and reward-model filtering for enterprise Post-Training (SFT, RLHF, DPO &amp;amp; GRPO) pipelines.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://arxiv.org/abs/2406.11704&quot;&gt;Nemotron-4 340B Technical Report&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Synthetic Data &amp; Self-Play · Paper · 2023-06-20</title>
      <link>https://foundationindex.ustunozgur.com/s/synthetic-data#synthetic-data-phi1-textbooks-2023</link>
      <guid>https://foundationindex.ustunozgur.com/s/synthetic-data#synthetic-data-phi1-textbooks-2023</guid>
      <pubDate>Tue, 20 Jun 2023 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Gunasekar et al. at Microsoft Research published Textbooks Are All You Need (Phi-1), demonstrating that a 1.3B model trained on curated synthetic textbooks and exercises matched much larger models on Python coding benchmarks.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://arxiv.org/abs/2306.11644&quot;&gt;Textbooks Are All You Need&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Synthetic Data &amp; Self-Play · Paper · 2022-12-20</title>
      <link>https://foundationindex.ustunozgur.com/s/synthetic-data#synthetic-data-self-instruct-2022</link>
      <guid>https://foundationindex.ustunozgur.com/s/synthetic-data#synthetic-data-self-instruct-2022</guid>
      <pubDate>Tue, 20 Dec 2022 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Wang et al. published Self-Instruct, showing that a pretrained language model can bootstrap its own instruction-following dataset through iterative generation and filtering—sparking the open-weight instruction-tuning wave.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://arxiv.org/abs/2212.10560&quot;&gt;Self-Instruct: Aligning Language Models with Self-Generated Instructions&lt;/a&gt;&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
