Twitxy Saved

Decart

Verified account This account is verified because it's an official organization on X. Learn more

@DecartAI

We are an AI research lab building Live AI models and an ultra-fast infrastructure to power them
decart.ai Joined August 2023
0 Following 28K Followers 495 Posts

Latest posts 5

Decart Verified account @DecartAI
What if an AI model could take back a harmful answer before it finishes it?

Diffusion language models can.

New Decart research accepted to #EMNLP2026 found harmful responses in just 9–21% of attacks, vs. 48–62% for comparable autoregressive models.

We turned that behavior into SRI Guard - detecting harmful generations with just 0.01–0.04% added latency.

Safety shouldn’t come at the cost of speed.

Read the paper: research.decart.ai/step-wise-refu…

@INSIGHTLabBGU @TechnionLive @bengurionu @CS_Uni_of_Haifa
0:08
3 replies 4 reposts 52 likes 12K views View on X