AI · Article
AI Tunes Another AI's Weights to Fix Repetitive Loops and Lift Math Scores
GeekNews ·
At a glance
Researchers gave an AI called GPT-6 Astra the task of improving the math problem-solving of Qwen3.5-0.8B, a publicly released model. Astra found the parts that cause the solutions to repeat the same phrases and made very small changes to the weights (the internal numerical values of an AI), and the score rose sharply.
Key points
- Before the fix, Qwen often repeated the same phrases while solving problems and ran into its character limit before reaching an answer.
- Astra ran Qwen directly, examined the values in its neural network layers, and changed the weights in two layers so that the patterns linked to repetition weakened.
- Only 0.016% of the weights were changed, and the MATH-500 score rose from 7.9% to 48.1%.
- The gain does not appear to come from new math ability; it seems to come from the model finishing more of the problems it could already solve.
Why it matters
It shows that people may be able to look inside an AI and fix the causes of its wrong answers.
Original source
[GN#383] AI가 다른 AI의 머릿속을 고쳤다
GeekNews · Nov 9, 09:00 KST
Read on GeekNews ↗AI summarized the original article. Check the original for accuracy. Copyright belongs to GeekNews.
More from this field
- Korean Firms Build Three-Way AI Push: Anthropic Stake, HBM Chips, Claude Rollout
Korean companies are expanding their roles in the global AI market around Anthropic's IPO. SK Telecom earned large paper gains from an early investment, while Samsung Electronics and SK hynix are strengthening chip supply ties. Samsung SDS and LG CNS are bringing Anthropic's Claude model into the business AI market.
뉴스웨이 · Nov 21, 06:10 KST - MIT and IBM Team Probes Gaps in Multimodal AI Understanding
MIT and IBM researchers are studying where today's multimodal AI models fall short, even when they handle text, images, and sound fluently. Their team, co-led by MIT's Jim Glass, builds new benchmarks and training methods that need less labeled data. This Q&A walks through how the project started and how it has evolved.
MIT News · Nov 21, 06:00 KST - Leaky AI Supply Chain Exposes Secret User Data at Alarming Rate
A research test sent nearly two million prompts with fake but working secrets through a network of AI services to see where they ended up. Many of the secrets were stolen, and attackers used them to reach cloud accounts. The results point to weak protection of customer data along the chain.
Ars Technica · Nov 21, 05:59 KST - OpenAI Shows How to Claim a Shopify Store in ChatGPT
This video shows how to claim your business in ChatGPT by connecting your Shopify store. You can start in Business Hub or the Shopify App Store. After connecting, you can manage your business information to control how your store appears in ChatGPT.
OpenAI 공식 YouTube · Nov 21, 05:48 KST