<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Reasoning &amp; Test-Time Compute — FoundationIndex</title>
    <link>https://foundationindex.ustunozgur.com/s/test-time-compute</link>
    <description>Training AI models to spend extra thinking time before answering—generating, verifying, and refining intermediate reasoning steps at inference time rather than blurting out the first token.</description>
    <language>en</language>
    <atom:link href="https://foundationindex.ustunozgur.com/s/test-time-compute/feed.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Reasoning &amp; Test-Time Compute · Release · 2025-03-25</title>
      <link>https://foundationindex.ustunozgur.com/s/test-time-compute#test-time-compute-gemini-25-2025</link>
      <guid>https://foundationindex.ustunozgur.com/s/test-time-compute#test-time-compute-gemini-25-2025</guid>
      <pubDate>Tue, 25 Mar 2025 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Google DeepMind launched Gemini 2.5 Pro as a thinking-by-default frontier model, merging test-time reasoning with a 1M-token native multimodal context window—signaling industry convergence between general-purpose and reasoning models.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/&quot;&gt;Gemini 2.5: Our most intelligent AI model&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Reasoning &amp; Test-Time Compute · Paper · 2025-01-22</title>
      <link>https://foundationindex.ustunozgur.com/s/test-time-compute#test-time-compute-deepseek-r1-2025</link>
      <guid>https://foundationindex.ustunozgur.com/s/test-time-compute#test-time-compute-deepseek-r1-2025</guid>
      <pubDate>Wed, 22 Jan 2025 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;DeepSeek published DeepSeek-R1, showing that pure reinforcement learning with verifiable rewards (GRPO) spontaneously elicits self-reflection, backtracking, and long-horizon reasoning without human-written reasoning traces, while open-sourcing distilled reasoning Checkpoints.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://arxiv.org/abs/2501.12948&quot;&gt;DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Reasoning &amp; Test-Time Compute · Release · 2024-09-12</title>
      <link>https://foundationindex.ustunozgur.com/s/test-time-compute#test-time-compute-o1-2024</link>
      <guid>https://foundationindex.ustunozgur.com/s/test-time-compute#test-time-compute-o1-2024</guid>
      <pubDate>Thu, 12 Sep 2024 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;OpenAI launched o1-preview, establishing that large-scale reinforcement learning over private chain-of-thought tokens produces smooth performance scaling with both RL training compute and inference-time thinking tokens.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://openai.com/index/learning-to-reason-with-llms/&quot;&gt;Learning to Reason with LLMs&lt;/a&gt;&lt;/p&gt;</description>
    </item>
    <item>
      <title>Reasoning &amp; Test-Time Compute · Paper · 2024-08-06</title>
      <link>https://foundationindex.ustunozgur.com/s/test-time-compute#test-time-compute-snell-2024</link>
      <guid>https://foundationindex.ustunozgur.com/s/test-time-compute#test-time-compute-snell-2024</guid>
      <pubDate>Tue, 06 Aug 2024 00:00:00 GMT</pubDate>
      <description>&lt;p&gt;Snell et al. published Scaling LLM Test-Time Compute Optimally, demonstrating that adaptive inference-time search and revision guided by process reward models can beat a 14× larger model on math reasoning under matched compute budgets.&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://arxiv.org/abs/2408.03314&quot;&gt;Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters&lt;/a&gt;&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
