AIHacker Newsabout 2 hours ago
Astra Improves Qwen Math Scores
17 min read
Researchers demonstrate that the Astra model can natively read and write neural network weights to improve other models. By editing 0.016% of Qwen3.5-0.8B weights, they raise its MATH-500 score from 7.9% to 48.1% and boost IFEval by 3x without using gradients.
Level
Hype check
Tap to vote and see what everyone thinks.