Startup's new mechanistic interpretability tool lets you debug LLMs(technologyreview.com)
technologyreview.com
Startup's new mechanistic interpretability tool lets you debug LLMs
https://www.technologyreview.com/2026/04/30/1136721/this-startups-new-mechanistic-interpretability-tool-lets-you-debug-llms/
2 comments
https://archive.ph/XQQHf
The urgency of reliable automated reasoning over natural language is more and more and more urgent. The LLMs have started the timer (or countdown), collective reality assessment makes the urgency obvious. From the article, for example:
> For example, many models will tell you that 9.11 is greater than 9.9. Looking inside a model to see what’s going on might reveal that it is being influenced by neurons associated with the Bible, in which verse 9.9 comes before 9.11, or by code repositories where consecutive updates are numbered 9.9, 9.10, 9.11 and so on. Using this information, the model can be retrained to make it avoid its “Bible” neurons when doing math
...See, that's not how it works (you do not "exclude golfing movements" when you "pilot a helicopter").
> For example, many models will tell you that 9.11 is greater than 9.9. Looking inside a model to see what’s going on might reveal that it is being influenced by neurons associated with the Bible, in which verse 9.9 comes before 9.11, or by code repositories where consecutive updates are numbered 9.9, 9.10, 9.11 and so on. Using this information, the model can be retrained to make it avoid its “Bible” neurons when doing math
...See, that's not how it works (you do not "exclude golfing movements" when you "pilot a helicopter").