1 comments

  • giang_at_glai13 hours ago
    Author here.<p>This post shows “concept algebra” on language model: inject, suppress, and compose human-understandable concepts at inference time (no retraining, no prompt engineering).<p>There’s an interactive demo on the post.<p>Would love feedback on: (1) what steering tasks you’d benchmark, (2) failure cases you’d want to see, (3) whether this kind of compositional control is useful in real products.<p>Related: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=47131225">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=47131225</a>