home.social

#interpretability — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #interpretability, aggregated by home.social.

  1. Benjamin Han @[email protected] ·

    Can we read what #LLMs are thinking but not saying? A paper from #Anthropic finds a small set of word-linked patterns inside #Claude that it can report, control on request, and use to reason silently, like the global workspace theory in #neuroscience. How do these findings connect to #LatentReasoning, #AISafety, and #KnowledgeEditing?

    benjaminhan.net/posts/20260927

    #Paper #AI #Interpretability

    #llms #anthropic #claude #neuroscience #latentreasoning #aisafety
  2. Benjamin Han @[email protected] ·

    Can we read what #LLMs are thinking but not saying? A paper from #Anthropic finds a small set of word-linked patterns inside #Claude that it can report, control on request, and use to reason silently, like the global workspace theory in #neuroscience. How do these findings connect to #LatentReasoning, #AISafety, and #KnowledgeEditing?

    benjaminhan.net/posts/20260927

    #Paper #AI #Interpretability

    #llms #anthropic #claude #neuroscience #latentreasoning #aisafety
  3. Benjamin Han @[email protected] ·

    Can we read what #LLMs are thinking but not saying? A paper from #Anthropic finds a small set of word-linked patterns inside #Claude that it can report, control on request, and use to reason silently, like the global workspace theory in #neuroscience. How do these findings connect to #LatentReasoning, #AISafety, and #KnowledgeEditing?

    benjaminhan.net/posts/20260927

    #Paper #AI #Interpretability

    #llms #anthropic #claude #neuroscience #latentreasoning #aisafety
  4. Benjamin Han @[email protected] ·

    Can we read what #LLMs are thinking but not saying? A paper from #Anthropic finds a small set of word-linked patterns inside #Claude that it can report, control on request, and use to reason silently, like the global workspace theory in #neuroscience. How do these findings connect to #LatentReasoning, #AISafety, and #KnowledgeEditing?

    benjaminhan.net/posts/20260927

    #Paper #AI #Interpretability

    #interpretability #ai #paper #knowledgeediting #aisafety #latentreasoning
  5. Benjamin Han @[email protected] ·

    Can we read what #LLMs are thinking but not saying? A paper from #Anthropic finds a small set of word-linked patterns inside #Claude that it can report, control on request, and use to reason silently, like the global workspace theory in #neuroscience. How do these findings connect to #LatentReasoning, #AISafety, and #KnowledgeEditing?

    benjaminhan.net/posts/20260927

    #Paper #AI #Interpretability

    #llms #anthropic #claude #neuroscience #latentreasoning #aisafety