Epi: Towards a Persistent, Self-Improving Agent for Autonomous Research

arXiv 2026 59.6 method

TLDR

Epi is an autonomous research agent with persistent memory and self-improvement, producing record mathematical results, a computer-assisted proof, and improved algorithmic coding solvers.

Reasoning

The paper presents a strong empirical demonstration of an autonomous research agent achieving notable mathematical and algorithmic results, with clear methodology involving persistent history, memory, and verification. Weaknesses include limited evidence of generality beyond specific problem families and no mention of literature review or paper writing capabilities.

Read-first score

Read-first score 59.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 70.

Recency 8%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,experiment,result

Topical relevance 42%

58.3

Uses existing LLM keyword relevance scores normalized to 0-100. AI scientist,automated scientific discovery,autonomous research agent,automated research,literature review agent,survey generation,automated experimentation,experiment design agent,AI for scientific research,paper writing agent,research automation,scientific discovery agent

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 52.

Keyword Scores

autonomous research agent

automated scientific discovery

scientific discovery agent

automated research

AI for scientific research

research automation

AI scientist

automated experimentation

experiment design agent

literature review agent

survey generation

paper writing agent

Deep Analysis

Innovations

Persistent, self-improving autonomous research agent that sustains investigations across model sessions and evolves methods through accumulated experience.
Combines persistent research history, hierarchical memory, branching, and independent review to maintain continuity, learn from experiments, and explore or revisit alternatives.
Runs experiments across session boundaries with outputs evaluated by a fixed verifier.
Extends the research process to algorithmic coding via Épi-turbo, a lightweight configuration that develops executable solvers.

Methodology

Épi maintains persistent research history and hierarchical memory, uses branching and independent review, and runs experiments across session boundaries with a fixed verifier evaluating outputs. Épi-turbo is evaluated on 30 FrontierOR tasks against Codex using the same underlying model and reasoning effort, comparing total benchmark score and model cost.

Key Results

Épi produced 120 record-setting results: 20 numerical bound records across 12 families and 100 construction records across eight families. It achieved an approximately 84-fold narrower interval for the n-queens constant and a more than 3.6-million-fold narrower rigorous interval for the quartic Feigenbaum attractor dimension, established via exact dual certificates that the Cohn-Elkies linear-programming bound cannot be sharp in dimensions 10 and 11, and gave a computer-assisted proof of Tuza's conjecture for split graphs with an eight-vertex clique part and at most three triangle-active neighborhood types; Épi-turbo achieved a 5.4% higher total benchmark score than Codex at 64% of model cost on 30 FrontierOR tasks.

Resources

Paper