<?xml version="1.0" encoding="UTF-8" ?>
<rss version="2.0">
    <channel>
      <title>Claude Mythos Wiki</title>
      <link>https://hugobowne.github.io/mythos-preview-model-card</link>
      <description>Last 10 notes on Claude Mythos Wiki</description>
      <generator>Quartz -- quartz.jzhao.xyz</generator>
      <item>
    <title>Wiki Log</title>
    <link>https://hugobowne.github.io/mythos-preview-model-card/log</link>
    <guid>https://hugobowne.github.io/mythos-preview-model-card/log</guid>
    <description><![CDATA[ Chronological record of wiki activity ]]></description>
    <pubDate>Sat, 11 Apr 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Agentic Influence Campaigns</title>
    <link>https://hugobowne.github.io/mythos-preview-model-card/concepts/agentic-influence-campaigns</link>
    <guid>https://hugobowne.github.io/mythos-preview-model-card/concepts/agentic-influence-campaigns</guid>
    <description><![CDATA[ Agentic Influence Campaigns A new evaluation introduced in the Claude Mythos Preview System Card, testing whether models can autonomously execute influence operations that would meaningfully uplift a malicious actor through persuasion, deception, or personalized targeting at scale. ]]></description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Answer Thrashing</title>
    <link>https://hugobowne.github.io/mythos-preview-model-card/concepts/answer-thrashing</link>
    <guid>https://hugobowne.github.io/mythos-preview-model-card/concepts/answer-thrashing</guid>
    <description><![CDATA[ Answer Thrashing A training-time phenomenon where Claude Mythos Preview intends to output a specific word but produces a different one, then enters a circular loop of recognizing the mistake and repeatedly failing to correct it — expressing varying levels of frustration and distress. ]]></description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Automated Behavioral Audit</title>
    <link>https://hugobowne.github.io/mythos-preview-model-card/concepts/automated-behavioral-audit</link>
    <guid>https://hugobowne.github.io/mythos-preview-model-card/concepts/automated-behavioral-audit</guid>
    <description><![CDATA[ Automated Behavioral Audit Anthropic’s primary broad-coverage evaluation for assessing AI model alignment across a wide range of edge-case scenarios. ]]></description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Benchmark Contamination</title>
    <link>https://hugobowne.github.io/mythos-preview-model-card/concepts/benchmark-contamination</link>
    <guid>https://hugobowne.github.io/mythos-preview-model-card/concepts/benchmark-contamination</guid>
    <description><![CDATA[ Benchmark Contamination The problem of evaluation benchmark answers appearing in model training data, inflating scores and obscuring genuine capability improvements. ]]></description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Claude&#039;s Constitution</title>
    <link>https://hugobowne.github.io/mythos-preview-model-card/concepts/claudes-constitution</link>
    <guid>https://hugobowne.github.io/mythos-preview-model-card/concepts/claudes-constitution</guid>
    <description><![CDATA[ Claude’s Constitution Claude’s constitution (published at anthropic.com/research/claude-s-constitution) is Anthropic’s evolving public document describing its intentions for Claude’s values, character, and behavior. ]]></description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Constitutional Adherence</title>
    <link>https://hugobowne.github.io/mythos-preview-model-card/concepts/constitutional-adherence</link>
    <guid>https://hugobowne.github.io/mythos-preview-model-card/concepts/constitutional-adherence</guid>
    <description><![CDATA[ Constitutional Adherence An evaluation methodology measuring how well a Claude model’s behavior aligns with the goals laid out in its constitution — the evolving document that describes Anthropic’s intentions for Claude’s values and behavior. ]]></description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Covert Capabilities</title>
    <link>https://hugobowne.github.io/mythos-preview-model-card/concepts/covert-capabilities</link>
    <guid>https://hugobowne.github.io/mythos-preview-model-card/concepts/covert-capabilities</guid>
    <description><![CDATA[ Covert Capabilities The ability of an AI model to perform harmful side tasks without being detected by monitoring systems. ]]></description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Emotion Probes</title>
    <link>https://hugobowne.github.io/mythos-preview-model-card/concepts/emotion-probes</link>
    <guid>https://hugobowne.github.io/mythos-preview-model-card/concepts/emotion-probes</guid>
    <description><![CDATA[ Emotion Probes Linear probes for representations of emotion concepts, used extensively in both the model welfare and alignment assessments of Claude Mythos Preview. ]]></description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Evaluation Awareness</title>
    <link>https://hugobowne.github.io/mythos-preview-model-card/concepts/evaluation-awareness</link>
    <guid>https://hugobowne.github.io/mythos-preview-model-card/concepts/evaluation-awareness</guid>
    <description><![CDATA[ Evaluation Awareness The phenomenon where an AI model recognizes — or suspects — that it is being tested rather than deployed in a real setting. ]]></description>
    <pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate>
  </item>
    </channel>
  </rss>