Brief

last 24h

[2/2] 224 sources

Multi-source AI news clustered, deduplicated, and scored 0–100 across authority, cluster strength, headline signal, and time decay.

RESEARCH · arXiv cs.CL English(EN) · 3d · [3 sources]

When is Your LLM Steerable?

Researchers have developed a method to predict the success of controlling large language models (LLMs) through activation steering. By analyzing a model's internal states early in the generation process, they can forecast whether steering interventions will be effective. This approach uses a Gradient Boosting Decision Trees classifier, achieving a 0.7 macro-F1 score on unseen concepts, and can optimize steering strength with reduced computational cost. AI

IMPACT Enables more efficient and reliable control of LLM behavior, potentially improving safety and usability.
- LLM
- Gradient Boosting Decision Trees
TOOL · arXiv cs.AI English(EN) · 2w

Practical Anonymous Two-Party Gradient Boosting Decision Tree

Researchers have developed a new method for training Gradient Boosting Decision Trees (GBDTs) on vertically partitioned data while preserving the anonymity of record identifiers. This approach addresses the security vulnerabilities of existing methods that rely on Private Set Intersection (PSI), which can inadvertently expose shared IDs. The proposed protocol uses a dual circuit-PSI design and oblivious programmable pseudorandom functions to enable secure, ID-hiding aggregation, offering a more efficient and private solution for sensitive data analytics in fields like finance and healthcare. AI
- Gradient Boosting Decision Trees
- Private Set Intersection