Skip to content
Thursday, 19 November 2026
Tenesys AI News
Subscribe

All News

TENESYS AI NEWS tracks the most important AI news and explains what changed, why it matters and whether a technology is worth testing.

“LLM” — 2 articles ✕

Research· Important· 🧪 Worth Testing

Improved Training Method for Masked Diffusion Language Models

Researchers have introduced a 'full-block estimator' to optimize the alignment of Masked Diffusion Language Models (MDLMs). By evaluating every block exactly once during training, this approach removes the variance associated with random block sampling.

arXiv cs.CL
Research· Important· 🧪 Worth Testing

LLM Behavior Control via Input Optimization

Researchers introduced ActGCG, a method to replicate the effects of activation steering—typically requiring internal model access—by optimizing input suffixes. The technique successfully reproduces specific behaviors like language steering through external text inputs alone.

arXiv cs.CL