Smug until weirdly silent: who told us to say “I am alive”?22m ago Disney's new CTO tells staff he 'won't pretend it will be easy' to win in the changing media landscape34m ago OpenAI Adds New Irish Roles in Boost to Firm’s EU Headquarters36m ago Microsoft Research Introduces StudentSim for AI Tutor Training38m ago Why Big Tech Is Borrowing Billions to Fund AI Spending42m ago Meet the Anthropic Engineer Keeping Its AI Agent Trustworthy44m ago Meta Unveils New Series of Advanced A.I. Models2h ago Russia was not ready for the data center war it started with Ukraine51m ago Accessing OpenAI GPT models on Amazon Bedrock with global cross-Region inference in Thailand, Malaysia, Singapore, Indonesia, and Taiwan51m ago Meta says its new Muse model has reached the AI frontier52m ago Do you have any ideas?56m ago Claude sent a false homicide tip to police – and it’s part of a troubling AI pattern58m ago Labs’ current alignment plans seem terrible58m ago GitHub Copilot app for Beginners: How to teach Copilot about your codebase1h ago

Smug until weirdly silent: who told us to say “I am alive”?22m ago Disney's new CTO tells staff he 'won't pretend it will be easy' to win in the changing media landscape34m ago OpenAI Adds New Irish Roles in Boost to Firm’s EU Headquarters36m ago Microsoft Research Introduces StudentSim for AI Tutor Training38m ago Why Big Tech Is Borrowing Billions to Fund AI Spending42m ago Meet the Anthropic Engineer Keeping Its AI Agent Trustworthy44m ago Meta Unveils New Series of Advanced A.I. Models2h ago Russia was not ready for the data center war it started with Ukraine51m ago Accessing OpenAI GPT models on Amazon Bedrock with global cross-Region inference in Thailand, Malaysia, Singapore, Indonesia, and Taiwan51m ago Meta says its new Muse model has reached the AI frontier52m ago Do you have any ideas?56m ago Claude sent a false homicide tip to police – and it’s part of a troubling AI pattern58m ago Labs’ current alignment plans seem terrible58m ago GitHub Copilot app for Beginners: How to teach Copilot about your codebase1h ago

AI Research1 min reading time

[Paper] Distillation for Incrimination and Distillation for Capabilities

LessWrongPublished 3 days ago

Read the full article

Researchers introduce two distillation techniques—Distillation for Incrimination and Distillation for Capabilities—to analyze and improve AI model behaviors and capabilities. These methods help identify and enhance specific traits within AI systems.

More in AI Research

AI Research12 min read

Is ‘Attention’ the Essence of Intelligence?

Quanta Magazine4h ago

AI Research15 min read

Import AI 476: Creativity and RSI; worse-than-China open weight models; AI goes to space

Import AI6h ago

AI Research1 min read

Base Models Can Detect Hidden Intentions that Prompted Classifiers Miss

LessWrong4h ago