Ju Li

Zhenting Qi, Huangyuan Su, Ao Qu et al.

How can a population of agents self-orchestrate and self-adapt into stronger collective intelligence without centralized control? Inspired by Friedrich Hayek's economic theory of decentralized coordination in markets, we study this question through an agent economy in which agents compete via auctions for the right to act, exchange payments, and accumulate wealth from environmental rewards. These simple economic signals induce decentralized credit assignment, driving planning without global orchestration or explicit communication protocols. The population evolves through economic selection: effective agents accumulate wealth and are mutated via exploitation, while ineffective ones go bankrupt and are replaced via exploration. We show that, initialized with weak agents, the economy produces emergent multi-step reasoning strategies and outperforms stronger monolithic baselines across five agentic tasks, including mathematical reasoning, financial research, scientific research, accelerator design, and distributed-system optimization. We further provide theoretical insights into how economic dynamics shape agent behaviors, linking local incentives to long-term global performance. Our results suggest a new path to multi-agent intelligence: rather than engineering coordination, we can design decentralized incentive structures under which it automatically emerges.

Bowen Deng, Bohan Li, Matthew Cox et al.

Computational materials science and chemistry span vast knowledge domains and fractured software ecosystems. Although large language models (LLMs) have demonstrated research capabilities, scaling monolithic agents to manage the rigor and complexity of atomistic research remains a challenge. Here, we introduce AtomisticSkills, an open-source harness framework that empowers general-purpose AI coding agents to conduct atomistic research across materials science, chemistry, and drug discovery. By hierarchically decomposing scientific workflows into agent skills and tools, AtomisticSkills provides agents with modular, extensible, and plug-and-play research capabilities. The framework integrates more than 100 human-curated multidisciplinary skills, including database access, thermodynamics and kinetics modeling, and diverse simulation engines employing machine learning interatomic potentials (MLIPs) and density functional theory (DFT). We validate its functional coverage against scientific literature and demonstrate robust orchestration capabilities across diverse scientific campaigns: generative design of Li-ion solid-state electrolytes, high-throughput screening of metal-organic frameworks for CO2 capture, autonomous MLIP benchmarking and fine-tuning, multi-stage structure-based virtual screening for drug design, multimodal X-ray diffraction pattern analysis, and screening of Fe-oxide catalysts for oxygen evolution reaction. AtomisticSkills provides a critical agent infrastructure towards building fully autonomous AI scientists.

Yang Jeong Park, Daniel Kaplan, Zhichu Ren et al.

We investigate whether large language models can perform the creative hypothesis generation that human researchers regularly do. While the error rate is high, generative AI seems to be able to effectively structure vast amounts of scientific knowledge and provide interesting and testable hypotheses. The future scientific enterprise may include synergistic efforts with a swarm of "hypothesis machines", challenged by automated experimentation and adversarial peer reviews.

Changhao Li, Boning Li, Omar Amer et al.

Distributed quantum computing is a promising computational paradigm for performing computations that are beyond the reach of individual quantum devices. Privacy in distributed quantum computing is critical for maintaining confidentiality and protecting the data in the presence of untrusted computing nodes. In this work, we introduce novel blind quantum machine learning protocols based on the quantum bipartite correlator algorithm. Our protocols have reduced communication overhead while preserving the privacy of data from untrusted parties. We introduce robust algorithm-specific privacy-preserving mechanisms with low computational overhead that do not require complex cryptographic techniques. We then validate the effectiveness of the proposed protocols through complexity and privacy analysis. Our findings pave the way for advancements in distributed quantum computing, opening up new possibilities for privacy-aware machine learning applications in the era of quantum technologies.

Yang Jeong Park, Sung Eun Jerng, Jin-Sung Park et al.

The advent of artificial intelligence (AI) has enabled a comprehensive exploration of materials for various applications. However, AI models often prioritize frequently encountered materials in the scientific literature, limiting the selection of suitable candidates based on inherent physical and chemical properties. To address this imbalance, we have generated a dataset of 1,494,017 natural language-material paragraphs based on combined OQMD, Materials Project, JARVIS, COD and AFLOW2 databases, which are dominated by ab initio calculations and tend to be much more evenly distributed on the periodic table. The generated text narratives were then polled and scored by both human experts and ChatGPT-4, based on three rubrics: technical accuracy, language and structure, and relevance and depth of content, showing similar scores but with human-scored depth of content being the most lagging. The merger of multi-modality data sources and large language model (LLM) holds immense potential for AI frameworks to help the exploration and discovery of solid-state materials for specific applications.

Tianyu Wu, Yu Yao, Zhenting Qi et al.

Speculative decoding accelerates LLM inference by having a small drafter propose tokens that a larger target model verifies in parallel. Recent diffusion-based parallel drafters such as DFlash predict the full B-token block in one forward pass, enabling deeper drafters and longer accepted blocks. However, existing multi-token drafter objectives often use fixed position-dependent weighting schedules, such as head-dependent weights or block-position decays, which do not adapt as the positions limiting acceptance change during training. To address this, we derive per-position training weights from a differentiable surrogate of expected accepted draft length, matching the weight of each position to its log-probability gradient contribution. The resulting loss, D-PACE (Dynamic Position-Aware Cross-Entropy), shifts training signal toward positions that currently limit acceptance as the drafter improves. Across six benchmarks, two Qwen3-4B draft depths, two decoding temperatures, and two additional target models, D-PACE consistently improves both wall-clock speedup and average emitted length, with 2.3\% measured training-time overhead and no changes to the drafter architecture or inference procedure.

Chia-Wei Hsu, Nien-Ti Tsou, Yu-Cheng Chen et al.

Gradient-based optimization drives the unprecedented performance of modern deep neural network models across diverse applications. Adaptive algorithms have accelerated neural network training due to their rapid convergence rates; however, they struggle to find ``flat minima" reliably, resulting in suboptimal generalization compared to stochastic gradient descent (SGD). By revisiting various adaptive algorithms' mechanisms, we propose the Frankenstein optimizer, which combines their advantages. The proposed Frankenstein dynamically adjusts first- and second-momentum coefficients according to the optimizer's current state to directly maintain consistent learning dynamics and immediately reflect sudden gradient changes. Extensive experiments across several research domains such as computer vision, natural language processing, few-shot learning, and scientific simulations show that Frankenstein surpasses existing adaptive algorithms and SGD empirically regarding convergence speed and generalization performance. Furthermore, this research deepens our understanding of adaptive algorithms through centered kernel alignment analysis and loss landscape visualization during the learning process. Code is available at https://github.com/acctouhou/Frankenstein_optimizer

Sathya Edamadaka, Soojung Yang, Ju Li et al.

Machine learning models of vastly different modalities and architectures are being trained to predict the behavior of molecules, materials, and proteins. However, it remains unclear whether they learn similar internal representations of matter. Understanding their latent structure is essential for building scientific foundation models that generalize reliably beyond their training domains. Although representational convergence has been observed in language and vision, its counterpart in the sciences has not been systematically explored. Here, we show that representations learned by nearly sixty scientific models, spanning string-, graph-, 3D atomistic, and protein-based modalities, are highly aligned across a wide range of chemical systems. Models trained on different datasets have highly similar representations of small molecules, and machine learning interatomic potentials converge in representation space as they improve in performance, suggesting that foundation models learn a common underlying representation of physical reality. We then show two distinct regimes of scientific models: on inputs similar to those seen during training, high-performing models align closely and weak models diverge into local sub-optima in representation space; on vastly different structures from those seen during training, nearly all models collapse onto a low-information representation, indicating that today's models remain limited by training data and inductive bias and do not yet encode truly universal structure. Our findings establish representational alignment as a quantitative benchmark for foundation-level generality in scientific models. More broadly, our work can track the emergence of universal representations of matter as models scale, and for selecting and distilling models whose learned representations transfer best across modalities, domains of matter, and scientific tasks.

Jaeha Oh, Byungchan Kim, Ju Li et al.

Modern scientific discovery is bottlenecked not by data scarcity, but by the inability to synthesize fragmented knowledge into actionable hypotheses. This challenge is especially acute in battery materials research, where electrochemical performance, interfacial behavior, and manufacturing feasibility must be optimized simultaneously. Here, we present the Multi-Persona Debate System (MPDS), a literature-grounded framework for automated scientific hypothesis generation that combines literature retrieval, long-context large language model reasoning, corpus-driven persona induction, and structured multi-agent debate. MPDS constructs literature snapshots of up to 500 papers, grounds agents in role-specific evidence pools, and conducts a three-round citation-aware debate followed by moderator synthesis, enabling negotiation between personas while preserving evidence traceability. We evaluate MPDS using a temporally controlled protocol excluding direct access to target papers, including two held-out battery-materials case studies and a blinded comparison across 30 matched cases. In sodium-ion anode and all-solid-state battery cathode design tasks, MPDS recovered design logics aligned with experimentally validated solution spaces and generated more mechanistically explicit, process-aware proposals than simpler baselines. To assess the impact of personas and debate, we introduce Integrative Hypothesis Quality scoring. In ablation studies, MPDS achieved the highest mean score among five conditions, with its largest advantage in cross-perspective integration. A laboratory follow-up suggests utility as a diagnostic aid for identifying practical bottlenecks in workflows. These results indicate that structured debate over literature snapshots improves hypothesis formation under coupled engineering constraints and provides a reusable workflow for text-intensive scientific discovery.

So Yeon Kim, Yang Jeong Park, Ju Li

Large-scale foundation models, including neural network interatomic potentials (NIPs) in computational materials science, have demonstrated significant potential. However, despite their success in accelerating atomistic simulations, NIPs face challenges in directly predicting electronic properties and often require coupling to higher-scale models or extensive simulations for macroscopic properties. Machine learning (ML) offers alternatives for structure-to-property mapping but faces trade-offs: feature-based methods often lack generalizability, while deep neural networks require significant data and computational power. To address these trade-offs, we introduce HackNIP, a two-stage pipeline that leverages pretrained NIPs. This method first extracts fixed-length feature vectors (embeddings) from NIP foundation models and then uses these embeddings to train shallow ML models for downstream structure-to-property predictions. This study investigates whether such a hybridization approach, by ``hacking" the NIP, can outperform end-to-end deep neural networks, determines the dataset size at which this transfer learning approach surpasses direct fine-tuning of the NIP, and identifies which NIP embedding depths yield the most informative features. HackNIP is benchmarked on Matbench, evaluated for data efficiency, and tested on diverse tasks including \textit{ab initio}, experimental, and molecular properties. We also analyze how embedding depth impacts performance. This work demonstrates a hybridization strategy to overcome ML trade-offs in materials science, aiming to democratize high-performance predictive modeling.

Yang Jeong Park, Mayank Kumaran, Chia-Wei Hsu et al.

Artificial intelligence (AI) is increasingly used for the inverse design of materials, such as crystals and molecules. Existing AI research on molecules has integrated chemical structures of molecules with textual knowledge to adapt to complex instructions. However, this approach has been unattainable for crystals due to data scarcity from the biased distribution of investigated crystals and the lack of semantic supervision in peer-reviewed literature. In this work, we introduce a contrastive language-crystals model (CLaC) pre-trained on a newly synthesized dataset of 126k crystal structure-text pairs. To demonstrate the advantage of using synthetic data to overcome data scarcity, we constructed a comparable dataset extracted from academic papers. We evaluate CLaC's generalization ability through various zero-shot cross-modal tasks and downstream applications. In experiments, CLaC achieves state-of-the-art zero-shot generalization performance in understanding crystal structures, surpassing latest large language models.

Yunwei Mao, Qi He, Ju Li

Networked computing power is a critical utility in the era of artificial intelligence. This paper presents a novel Physical Infrastructure Finance (PinFi) protocol designed to facilitate the distribution of computing power within networks in a decentralized manner. Addressing the core challenges of coordination, pricing, and liquidity in decentralized physical infrastructure networks (DePIN), the PinFi protocol introduces a distinctive dynamic pricing mechanism. It enables providers to allocate excess computing resources to a "dissipative" PinFi liquidity pool, distinct from traditional DeFi liquidity pools, ensuring seamless access for clients at equitable, market-based prices. This approach significantly reduces the costs of accessing computing power, potentially to as low as 1% compared to existing services, while simultaneously enhancing security and dependability. The PinFi protocol is poised to transform the dynamics of supply and demand in computing power networks, setting a new standard for efficiency and accessibility.

Yu Yao, Jiayi Dong, Ju Li et al.

Large language models (LLMs) have demonstrated remarkable capabilities not only in language generation but also in advancing scientific discovery. A growing body of work has explored ways to improve their reasoning, from self-consistency and chain-of-thought to multi-agent debate. Inspired by the dynamics of scientific committees and the "Society of Mind," we introduce Roundtable Policy, a complementary inference-time reasoning framework that performs inference through the weighted consensus of multiple LLMs. Our findings indicate that this approach significantly enhances reasoning in complex heterogeneous scientific tasks and improves scientific narratives in terms of creativity, rigor, and logical coherence, while reducing hallucinations that single models are prone to. Our approach emphasizes structured and interpretable consensus rather than opaque convergence, while requiring only black-box access and uniform procedures, making it broadly applicable to multi-LLM reasoning.

Gang Cheng, Xin Gao, Li Hu et al.

We introduce Wan-Animate, a unified framework for character animation and replacement. Given a character image and a reference video, Wan-Animate can animate the character by precisely replicating the expressions and movements of the character in the video to generate high-fidelity character videos. Alternatively, it can integrate the animated character into the reference video to replace the original character, replicating the scene's lighting and color tone to achieve seamless environmental integration. Wan-Animate is built upon the Wan model. To adapt it for character animation tasks, we employ a modified input paradigm to differentiate between reference conditions and regions for generation. This design unifies multiple tasks into a common symbolic representation. We use spatially-aligned skeleton signals to replicate body motion and implicit facial features extracted from source images to reenact expressions, enabling the generation of character videos with high controllability and expressiveness. Furthermore, to enhance environmental integration during character replacement, we develop an auxiliary Relighting LoRA. This module preserves the character's appearance consistency while applying the appropriate environmental lighting and color tone. Experimental results demonstrate that Wan-Animate achieves state-of-the-art performance. We are committed to open-sourcing the model weights and its source code.

Mouyang Cheng, Weiliang Luo, Hao Tang et al.

Diffusion-based deep generative models have emerged as powerful tools for inverse materials design. Yet, many existing approaches overlook essential chemical constraints such as oxidation state balance, which can lead to chemically invalid structures. Here we introduce CrysVCD (Crystal generator with Valence-Constrained Design), a modular framework that integrates chemical rules directly into the generative process. CrysVCD first employs a transformer-based elemental language model to generate valence-balanced compositions, followed by a diffusion model to generate crystal structures. The valence constraint enables orders-of-magnitude more efficient chemical valence checking, compared to pure data-driven approaches with post-screening. When fine-tuned on stability metrics, CrysVCD achieves 85% thermodynamic stability and 68% phonon stability. Moreover, CrysVCD supports conditional generation of functional materials, enabling discovery of candidates such as high thermal conductivity semiconductors and high-$κ$ dielectric compounds. Designed as a general-purpose plugin, CrysVCD can be integrated into diverse generative pipeline to promote chemical validity, offering a reliable, scientifically grounded path for materials discovery.

Ping Tuo, Jiale Chen, Ju Li

Elucidating reaction mechanisms hinges on efficiently generating transition states (TSs), products, and complete reaction networks. Recent generative models, such as diffusion models for TS sampling and sequence-based architectures for product generation, offer faster alternatives to quantum-chemistry searches. But diffusion models remain constrained by their stochastic differential equation (SDE) dynamics, which suffer from inefficiency and limited controllability. We show that flow matching, a deterministic ordinary differential (ODE) formulation, can replace SDE-based diffusion for molecular and reaction generation. We introduce MolGEN, a conditional flow-matching framework that learns an optimal transport path to transport Gaussian priors to target chemical distributions. On benchmarks used by TSDiff and OA-ReactDiff, MolGEN surpasses TS geometry accuracy and barrier-height prediction while reducing sampling to sub-second inference. MolGEN also supports open-ended product generation with competitive top-k accuracy and avoids mass/electron-balance violations common to sequence models. In a realistic test on the $γ$-ketohydroperoxide decomposition network, MolGEN yields higher fractions of valid and intended TSs with markedly fewer quantum-chemistry evaluations than string-based baselines. These results demonstrate that deterministic flow matching provides a unified, accurate, and computationally efficient foundation for molecular generative modeling, signaling that flow matching is the future for molecular generation across chemistry.

Ruoyan Avery Yin, Zhichu Ren, Zongyou Yin et al.

The Copilot for Real-world Experimental Scientist (CRESt) system empowers researchers to control autonomous laboratories through conversational AI, providing a seamless interface for managing complex experimental workflows. We have enhanced CRESt by integrating a multi-agent collaboration mechanism that utilizes the complementary strengths of the ChatGPT and Gemini models for precise image analysis in materials science. This innovative approach significantly improves the accuracy of experimental outcomes by fostering structured debates between the AI models, which enhances decision-making processes in materials phase analysis. Additionally, to evaluate the generalizability of this approach, we tested it on a quantitative task of counting particles. Here, the collaboration between the AI models also led to improved results, demonstrating the versatility and robustness of this method. By harnessing this dual-AI framework, this approach stands as a pioneering method for enhancing experimental accuracy and efficiency in materials research, with applications extending beyond CRESt to broader scientific experimentation and analysis.

Hao Tang, Brian Xiao, Wenhao He et al.

Machine learning (ML) plays an important role in quantum chemistry, providing fast-to-evaluate predictive models for various properties of molecules. However, most existing ML models for molecular electronic properties use density functional theory (DFT) databases as ground truth in training, and their prediction accuracy cannot surpass that of DFT. In this work, we developed a unified ML method for electronic structures of organic molecules using the gold-standard CCSD(T) calculations as training data. Tested on hydrocarbon molecules, our model outperforms DFT with the widely-used hybrid and double hybrid functionals in computational costs and prediction accuracy of various quantum chemical properties. As case studies, we apply the model to aromatic compounds and semiconducting polymers on both ground state and excited state properties, demonstrating its accuracy and generalization capability to complex systems that are hard to calculate using CCSD(T)-level methods.

Qi He, Yunwei Mao, Ju Li

PinFi is a class of novel protocols for decentralized pricing of dissipative assets, whose value naturally declines over time. Central to the protocol's functionality and its market efficiency is the role of liquidity providers (LPs). This study addresses critical stability and sustainability challenges within the protocol, namely: the propensity of LPs to prefer selling in external markets over participation in the protocol; a similar inclination towards selling within the PinFi system rather than contributing as LPs; and a scenario where LPs are disinclined to sell within the protocol. Employing a game-theoretic approach, we explore PinFi's mechanisms and its broader ramifications. Our findings reveal that, under a variety of common conditions and with an assumption of participant integrity, PinFi is capable of fostering a dynamic equilibrium among LPs, sellers, and buyers. This balance is maintained through a carefully calibrated range of block rewards for LPs, ensuring the protocol's long-term stability and utility.

Akihide Hayashi, So Takamoto, Ju Li et al.

Mapping the chemical reaction pathways and their corresponding activation barriers is a significant challenge in molecular simulation. Given the inherent complexities of 3D atomic geometries, even generating an initial guess of these paths can be difficult for humans. This paper presents an innovative approach that utilizes neural networks to generate initial guesses for reaction pathways based on the initial state and learning from a database of low-energy transition paths. The proposed method is initiated by inputting the coordinates of the initial state, followed by progressive alterations to its structure. This iterative process culminates in the generation of the guess reaction path and the coordinates of the final state. The method does not require one-the-fly computation of the actual potential energy surface, and is therefore fast-acting. The application of this geometry-based method extends to complex reaction pathways illustrated by organic reactions. Training was executed on the Transition1x dataset of organic reaction pathways. The results revealed the generation of reactions that bore substantial similarities with the test set of chemical reaction paths. The method's flexibility allows for reactions to be generated either to conform to predetermined conditions or in a randomized manner.

Huiping Shen, Yinan Zhao, Ju Li et al.

This paper presents a novel three-degree-of-freedom (3-DOF) translational parallel manipulator (TPM) by using a topological design method of parallel mechanism (PM) based on position and orientation characteristic (POC) equations. The proposed PM is only composed of lower-mobility joints and actuated prismatic joints, together with the investigations on three kinematic issues of importance. The first aspect pertains to geometric modeling of the TPM in connection with its topological characteristics, such as the POC, degree of freedom and coupling degree, from which its symbolic direct kinematic solutions are readily obtained. Moreover, the decoupled properties of input-output motions are directly evaluated without Jacobian analysis. Sequentially, based upon the inverse kinematics, the singular configurations of the TPM are identified, wherein the singular surfaces are visualized by means of a Gr{ö}bner based elimination operation. Finally, the workspace of the TPM is evaluated with a geometric approach. This 3-DOF TPM features less joints and links compared with the well-known Delta robot, which reduces the structural complexity. Its symbolic direct kinematics and partially-decoupled property will ease path planning and dynamic analysis. The TPM can be used for manufacturing large work pieces.

Huiping Shen, Damien Chablat, Boxiong Zeng et al.

According to the topological design theory and method of parallel mechanism (PM) based on position and orientation characteristic (POC) equations, this paper studied a 3-DOF translational PM that has three advantages, i.e., (i) it consists of three fixed actuated prismatic joints, (ii) the PM has analytic solutions to the direct and inverse kinematic problems, and (iii) the PM is of partial motion decoupling property. Firstly, the main topological characteristics, such as the POC, degree of freedom and coupling degree were calculated for kinematic modeling. Thanks to these properties, the direct and inverse kinematic problems can be readily solved. Further, the conditions of the singular configurations of the PM were analyzed which corresponds to its partial motion decoupling property.

So Takamoto, Satoshi Izumi, Ju Li

A universal interatomic potential for an arbitrary set of chemical elements is urgently needed in computational materials science. Graph convolution neural network (GCN) has rich expressive power, but previously was mainly employed to transport scalars and vectors, not rank $\ge 2$ tensors. As classic interatomic potentials were inspired by tight-binding electronic relaxation framework, we want to represent this iterative propagation of rank $\ge 2$ tensor information by GCN. Here we propose an architecture called the tensor embedded atom network (TeaNet) where angular interaction is translated into graph convolution through the incorporation of Euclidean tensors, vectors and scalars. By applying the residual network (ResNet) architecture and training with recurrent GCN weights initialization, a much deeper (16 layers) GCN was constructed, whose flow is similar to an iterative electronic relaxation. Our traning dataset is generated by density functional theory calculation of mostly chemically and structurally randomized configurations. We demonstrate that arbitrary structures and reactions involving the first 18 elements on the periodic table (H to Ar) can be realized satisfactorily by TeaNet, including C-H molecular structures, metals, amorphous SiO${}_2$, and water, showing surprisingly good performance (energy mean absolute error 19 meV/atom) and robustness for arbitrary chemistries involving elements from H to Ar.