Épi agent reports 120 record-setting autonomous research results
Original title: Epi: Towards a Persistent, Self-Improving Agent for Autonomous Research
Authors say Épi-turbo beat Codex by 5.4% on 30 FrontierOR tasks at 64% of the model cost.
Épi is an autonomous research agent that continues investigations across sessions using persistent history, hierarchical memory, branching, and independent review. A fixed verifier evaluates experiment outputs. The authors report 120 record-setting results and a computer-assisted proof of a restricted case of Tuza's conjecture. Épi-turbo applies the process to coding tasks such as routing and scheduling.
Unconfirmed: Record counts and the Codex comparison are author-reported and not independently confirmed here.
- Reported results include 20 numerical bound records and 100 construction records.
- Authors cite an about 84-fold narrower interval for the n-queens constant.
- They also cite a more than 3.6-million-fold narrower Feigenbaum interval.
- Dual certificates are said to show the Cohn-Elkies bound is not sharp in dimensions 10 and 11.
- Products
- FrontierOR
From the original article
1 min read at huggingface.co
We present Épi, an autonomous research agent that sustains investigations across model sessions and evolves its methods through accumulated experience. Épi combines persistent research history, hierarchical memory, branching and independent review to maintain research continuity, learn from experiments, and explore and revisit alternative approaches.
By Yue Huang, Zhangchen Xu, Xiaonan Luo · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.