alphaXiv shares paper proposing Memory Foundational Model
The model compresses past interactions into dynamic layer states, enabling persistent memory without gradient updates, though still early and lossy.
See the latest news and media coverage for alphaXiv. We track all announcements, press releases, and industry mentions in real time, all in one place.
Platform for discussing and discovering arXiv papers
alphaxiv.orgLast updated
In short: alphaXiv raised $7 million in seed funding and launched its Assistant 2.0 and Pro Plan to enhance its AI-driven research platform.
The model compresses past interactions into dynamic layer states, enabling persistent memory without gradient updates, though still early and lossy.
The method uses Hilbert function space for data-free pruning and merging of neurons.
Examines data efficiency, base vs instruct models, population size on Countdown and GSM8K. ES excels in low-data regimes. GRPO superior with more data and base models.
Trains Qwen3-8B on arXivQA dataset from real queries. Achieves competitive recall against frontier models using multi-hop search.
Thanks for your great work! It significantly improves the efficiency and convenience of reading and managing paper. The AI assistant function is great, but sometimes...
AlphaXiv launched a Claude Skill that pulls structured paper overviews via API instead of parsing raw PDFs.
AlphaXiv, an AI research platform, raised a $7 million seed led by Menlo Ventures and Haystack, co-founder Raj Palleti tells Axios exclusively. Why it matters...
There is an inherent tension in the dissemination of research. On one hand, science thrives on openness and communication. On the other, ensuring high-quality scientific...
Track alphaXiv and your other target companies to get real-time alerts and weekly summaries delivered straight to your inbox.
Browse news for competitors to alphaXiv and other trending companies.