arXiv:2608. 06714v1 Announce Type: new Abstract: Recent systems for optimizing prompts, programs, and ML workflows typically rely on explicit outer-loop controllers such as evolutionary search, bandits, or textual-gradient methods.
Recent systems for optimizing prompts, programs, and ML workflows typically rely on explicit outer-loop controllers such as evolutionary search, bandits, or textual-gradient methods. We ask a fundamentally different question: how much of this search policy can be internalized by a single tool-using agent?
arXiv:2608. 12841v1 Announce Type: cross Abstract: We study recursive self-improvement at the level of quantitative-investment research: whether an autonomous system can use evidence from earlier experiments to improve the hypotheses and candidates proposed in later iterations.
arXiv:2609.38445v1 Announce Type: new Abstract: Frontier LLMs are increasingly used to automate scientific research through iterative search. We distinguish idea-driven search from solution-driven se...
arXiv:2608. 10424v1 Announce Type: new Abstract: A slew of recent works develop agents for solving research problems end-to-end, a paradigm increasingly referred to as autoresearch.
AlphaDiverse is a framework that enhances large language model–based multi‑agent systems for alpha factor mining by addressing cost, availability, and confidentiality constraints. It generates diverse research paths through varied environments and post‑training local agents, then fine‑tunes these agents with supervised learning and optimizes them jointly using a GRPO method that balances predictive quality and diversity. The approach limits research feedback to inner‑period data and evaluates a frozen model on outer‑period data to avoid test‑set tuning, demonstrating competitive prediction and broader exploration across four Chinese stock universes.
The paper reports on applying AutoResearch—a large language model that iteratively edits training scripts—to optimize embedding systems for a book recommendation pipeline at production scale. Over twelve weeks, the authors ran 220+ experiments across two representation‑learning systems, uncovering five recurring failure modes (infrastructure fragility, agent memory decay, search‑direction stagnation, iteration‑cost asymmetry, and metric fixation) that were not present in smaller settings. They propose a three‑principle scaffolding (prevent, persist, redirect) to address these failures, achieving a 1.82× lift in Recall@6, a 2.1× lift in coherence, and an autonomous text‑only fallback that expanded catalog coverage by 5.8×.
arXiv:2603. 04873v3 Announce Type: replace Abstract: Accurate time series forecasting underpins decision-making in many domains, yetconventional ML development often faces data scarcity, distribution shift, anddiminishing returns from manual iteration.
The article argues that agentic auto‑research should be guided by dense, intermediate signals of epistemic progress rather than by sparse final benchmarks. It compares this approach to fuzz testing, where coverage provides continuous feedback that directs input mutation. The authors propose controlled experiments to test whether such signals improve discovery efficiency and reduce false positives, and demonstrate in a simulated physics setting that an AI agent using feedback‑driven search uncovers a hidden law while optimization‑driven baselines fail.
arXiv:2606. 05080v1 Announce Type: new Abstract: Scientific and engineering progress is fundamentally a long-horizon iterative process: proposing changes, running experiments, measuring outcomes, and continuously refining artifacts.
AutoFyn is a new agent framework based on the Expert Iteration algorithm that updates a frozen model’s persistent state using verified reward signals instead of changing model weights. Each iteration starts with a fresh model session, and only durable information—such as memory files, reports, and repository state—is carried over through explicit interfaces. The system orchestrates exploration, planning, and specialized agents, while a task‑grounded verifier supplies objective rewards that are distilled back into the persistent state to guide the next round. AutoFyn has been applied to olympiad mathematics, data science, and cybersecurity, outperforming baseline models on the 2026 International Mathematical Olympiad, topping the Spider 2.0 dbt benchmark, and generating sixteen maintainer‑confirmed vulnerability advisories across several popular software projects.
As coding agents move from supervised code completion to unattended, around-the-clock exploration, their work expands from isolated predictions into long trajectories of reasoning, tool use, and feedback. Token efficiency therefore becomes important for scaling recursive self-improvement.