Compound LLM systems often solve a coordination problem by adding a higher-level LLM. The resulting meta-agent reads workers' outputs, writes the final answer, allocates later calls, and decides when to stop. It is expressive, but it also concentrates three control decisions in an opaque, order-sensitive model call. We ask whether the manager needs to be generative at all.
Is this you? As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.
Claim your profile
Articles by Yanwei Cui
Who verifies the verifier? Co-evolving inspectable graders with self-improving agents
We changed the agent: did it actually get better? Every self-improving agent loop answers this hundreds of times, and every answer comes from a verifier. On open-ended tasks none exists, so the loop is handed a hand-written rubric or a bare LLM judge grading output from a model like itself, inviting reward hacking and shared blind spots.
Metrics that write themselves: Evolving an evaluator from its own blind spots
Agents improve quickly against a reliable automatic metric and stall without one, and the applications that need them most, report generation among them, are the ones nobody knows how to score. Can the metric write itself? Saying what makes an answer good is hard; pointing at something wrong with one is easier, so the metric we evolve is a pool of small Python operators that each flag a candidate for one named defect, or abstain, and vote.
Breeding of high-yield cold-adapted lipase fungi, optimization of fermentation conditions and isolation and purification of lipase
Abstract Cold-adapted lipases have attracted extensive research attention in enzymology owing to their unique low-temperature adaptability and catalytic properties, and investigations in this area continue to deepen. This study aimed to efficiently produce a cold-adapted lipase to overcome the limitations of conventional lipases under low-temperature conditions, such as low activity, high cost, and restricted applicability.
The blind curator: How a biased judge silently disables skill retirement in self-evolving agents
A self-evolving agent retires its bad skills by watching them fail, so what happens when the judge cannot see the failures? Skill retirement is the structural constraint that keeps a growing library from drifting below the no-skill baseline, but its guarantee assumes an unbiased reward, which is false for the LLM judges that reference-free tasks require. We show that a biased judge does not merely add noise; it silently switches off the curator.
Library drift: Diagnosing and fixing a silent failure mode in self-evolving LLM skill libraries
Self-evolving skill libraries face a silent failure mode we term library drift: unbounded skill accumulation without outcome-driven lifecycle management causes retrieval degradation, false-positive injections, and performance stagnation. Recent evaluation confirms the symptom (LLM-authored skills deliver +0.0pp gain while human-curated ones deliver +16.2pp; SkillsBench (Li et al., 2026)), yet the underlying mechanism has not been isolated.
Closing the feedback loop: From experience extraction to insight governance in verbal reinforcement learning
Training-free verbal reinforcement learning enables LLM agents to learn from world feedback—objective signals such as dynamic task outcomes, market returns, or demand forecasts—by extracting verbal rules from experience and injecting them as context, updating the agent's behavior without parameter changes.
Optimizing enterprise AI assistants: How Crypto.com uses LLM reasoning and feedback for enhanced efficiency | Amazon Web Services
This post is co-written with Jessie Jiao from Crypto.com. Crypto.com is a crypto exchange and comprehensive trading service serving 140 million users in 90 countries. To improve the service quality of Crypto.com, the firm implemented generative AI-powered assistant services on AWS. Modern AI assistants—artificial intelligence systems designed to interact with users through natural language, answer questions, and even perform tasks—face increasingly complex challenges in production environments.
Hydrogen Leakage Simulation and Optimization of Sensor Layout for Hydrogen Supply Systems in Fuel Cell Truck
Conflict of Interest The authors declare no conflict of interest. References 1, , , , , , Int. J. Hydrogen Energy 2024, 68, 1090. 2, , , , , , , , , , , Int. J. Hydrogen Energy 2023, 48, 12935. 3, , , , , , , Comput. Chem. Eng. 2012, 47, 194. 4, , , , Ocean Eng. 2019, 187, 106223. 5, , Int. J. Energy Res. 2022, 46, 17720. 6, , , , , , , J. Loss Prev. Process Ind. 2022, 80, 104859. 7, , , , , Int. J. Greenhouse Gas Control 2020, 95, 102951. 8, , , , , Process Saf. Environ. Prot. 2022, 161, 571. 9, , , , , IEEE Sens. J.
How Apoidea Group enhances visual information extraction from banking documents with multimodal models using LLaMA-Factory on Amazon SageMaker HyperPod | AWS Machine Learning Blog
This post is co-written with Ken Tsui, Edward Tsoi and Mickey Yip from Apoidea Group. The banking industry has long struggled with the inefficiencies associated with repetitive processes such as information extraction, document review, and auditing. These tasks, which require significant human resources, slow down critical operations such as Know Your Customer (KYC) procedures, loan applications, and credit analysis.
W$_{4}$IrC$_{1-x}$: A new noncentrosymmetric superconductor with the cubic $\beta$-Mn type structure
W$_{4}$IrC$_{1-x}$: A new noncentrosymmetric superconductor with the cubic $\beta$-Mn type structure Qinqing Zhu, Guorui Xiao, Yanwei Cui, Wuzhang Yang, Siqi Wu, Guang-han Cao and Zhi Ren Abstract We report the discovery of a new noncentrosymmetric superconductor, W$_{4}$IrC$_{1-x}$, which adopts the cubic $\beta$-Mn type structure for $x$ in the range of 0 to 0.2. This compound is stabilized by interstitial carbon and becomes unstable by annealing or varying the W/Ir ratio.
Preparation of three-dimensional nanosheet-based molybdenum disulfide nanotubes as anode materials for lithium storage
Three dimensional (3D) hierarchical nanosheet-based MoS2 nanotubes are synthesized via a self-sacrificial template method. The MoO3 nanorods as self-sacrificial templates are directly converted into MoS2 nanotubes through a solvothermal sulfidation reaction. The MoS2 nanotubes which are in a diameter range of 250-400 nm are systematically characterized by scanning electron microscopy, high-resolution transmission electron microscopy, X-ray photo electron spectroscopy, and Raman spectroscopy, etc.
Show More
loading
Actions
Is this you?
As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.Get in touch with Yanwei
Contact Yanwei, search articles and posts on X, monitor coverage, and track replies from one place.
Learn more about Muck Rack