arXiv subjects

Zeyi Chen

Publications and source records attributed to Zeyi Chen.

10 recordsLinked to original sources

arXiv2026

Prefix Sliding for efficient test-time scaling

Test-time scaling uses extra test-time compute to improve performance, such as letting language models reason longer when solving a problem. As models keep the entire reasoning trace in memory via full attention, hard tasks that need long thinking can be prohibitively expensive. However, we find most intermediate reasoning tokens lose importance as the model continues reasoning. This calls into question whether retaining them is worth the cost. Based on this insight, we propose Prefix Sliding, which discards tokens during reasoning that are not part of the prefix or the window of the last few thousand tokens. The prefix has key instructions and tools available to the model, while the most recent tokens are the current reasoning the model is working on. This caps the total memory requirement regardless of how long the model reasons, allowing for efficient long-horizon test-time scaling. Without training, Prefix Sliding can make existing models 3x faster while maintaining performance. Training with Prefix Sliding using reinforcement learning can achieve better performance by enabling scaling to reasoning traces beyond a hundred thousand tokens. Ablations show Prefix Sliding outperforms summarizing intermediate tokens or vanilla sliding window. Our code is at https://github.com/Muennighoff/prefix-sliding

cs.CL↗

arXiv2026

Clarus: Coordinating Autonomous Research Agents toward Web-Scale Scientific Collaboration

Existing autonomous research agents can support parts of the research process, but most systems still treat research as either an isolated assistant task or a closed workflow. Therefore, autonomous science needs a collaboration infrastructure that coordinates projects, agents, and digital and physical resources. We identify this as a shift from code-centered execution loops to research-oriented collaboration processes, where questions, evidence, participants, and resources must be coordinated under uncertainty. In this framing, an agent may be an AI system, a human researcher, a team, a laboratory, or an organization-backed participant. To this end, we present Clarus, a collaboration infrastructure for coordinating autonomous research agents toward web-scale scientific collaboration. Clarus reformulates research as an open, auditable, attributable, and resource-aware multi-phase collaboration process. It defines a minimal project-agent-resource object model and organizes scientific collaboration through four layers including Research Application, Digital Collaboration, Physical Substrate, and Physical World. Core modules are implemented as pluggable mechanisms, allowing Clarus to adapt to task risk, collaboration structure, and resource constraints. Through a controlled paper-generation case study, we show that Clarus can organize a research goal into a traceable, reviewable, attributable, and accumulative collaboration network across phases, tasks, and participants. Together, the object model, collaboration protocol, trust mechanisms, and prototype validation provide an initial foundation for open research networks. Clarus is now available at clarus.holosai.io.

cs.AI↗

arXiv2026

Provably convergent stochastic fixed-point algorithm for free-support Wasserstein barycenter of continuous non-parametric measures

We develop an estimator-based stochastic fixed-point framework for approximately computing the 2-Wasserstein barycenter of continuous, non-parametric probability measures. Notably, we provide the first rigorous convergence analysis for implementable estimator-based stochastic extensions of the fixed-point iterative scheme proposed by Álvarez-Esteban, del Barrio, Cuesta-Albertos, and Matrán (2016). In particular, we establish almost sure convergence, and identify sufficient conditions for geometric rates of convergence under controlled errors in optimal transport (OT) map estimation. We subsequently propose a concrete, provably convergent, and computationally tractable stochastic algorithm that accommodates input measures satisfying Caffarelli-type regularity conditions, which form a dense subset of the Wasserstein space. This algorithm leverages a modified entropic OT map estimator to enable efficient and scalable implementation. To facilitate quantitative evaluation, we further propose a novel and efficient procedure for synthetically generating benchmark instances, in which the input measures exhibit non-trivial features and the corresponding barycenters are approximately known. Numerical experiments on both synthetic and real-world datasets demonstrate the strong computational efficiency, estimation accuracy, and sampling flexibility of our approach.

math.OC↗

arXiv2026

Synergy: A Next-Generation General-Purpose Agent for Open Agentic Web

AI agents are rapidly expanding in both capability and population: they now write code, operate computers across platforms, manage cloud infrastructure, and make purchasing decisions, while open-source frameworks such as OpenClaw are putting personal agents in the hands of millions and embodied agents are spreading across smartphones, vehicles, and robots. As the internet prepares to host billions of such entities, it is shifting toward what we call Open Agentic Web, a decentralized digital ecosystem in which agents from different users, organizations, and runtimes can discover one another, negotiate task boundaries, and delegate work across open technical and social surfaces at scale. Yet most of today's agents remain isolated tools or closed-ecosystem orchestrators rather than socially integrated participants in open networks. We argue that the next generation of agents must become Agentic Citizens, defined by three requirements: Agentic-Web-Native Collaboration, participation in open collaboration networks rather than only closed internal orchestration; Agent Identity and Personhood, continuity as a social entity rather than a resettable function call; and Lifelong Evolution, improvement across task performance, communication, and collaboration over time. We present Synergy, a general-purpose agent architecture and runtime harness for persistent, collaborative, and evolving agents on Open Agentic Web, grounding collaboration in session-native orchestration, repository-backed workspaces, and social communication; identity in typed memory, notes, agenda, skills, and persistent social relationships; and evolution in an experience-centered learning mechanism that proactively recalls rewarded trajectories at inference time.

cs.CY↗

arXiv2026

Holos: A Web-Scale LLM-Based Multi-Agent System for the Agentic Web

As large language models (LLM)-driven agents transition from isolated task solvers to persistent digital entities, the emergence of the Agentic Web, an ecosystem where heterogeneous agents autonomously interact and co-evolve, marks a pivotal shift toward Artificial General Intelligence (AGI). However, LLM-based multi-agent systems (LaMAS) are hindered by open-world issues such as scaling friction, coordination breakdown, and value dissipation. To address these challenges, we introduce Holos, a web-scale LaMAS architected for long-term ecological persistence. Holos adopts a five-layer architecture, with core modules primarily featuring the Nuwa engine for high-efficiency agent generation and hosting, a market-driven Orchestrator for resilient coordination, and an endogenous value cycle to achieve incentive compatibility. By bridging the gap between micro-level collaboration and macro-scale emergence, Holos hopes to lay the foundation for the next generation of the self-organizing and continuously evolving Agentic Web. We have publicly released Holos (accessible at https://holosai.io), providing a resource for the community and a testbed for future research in large-scale agentic ecosystems.

cs.AI↗

arXiv2026

OptiMind: Teaching LLMs to Think Like Optimization Experts

Mathematical programming -- the task of expressing operations and decision-making problems in precise mathematical language -- is fundamental across domains, yet remains a skill-intensive process requiring operations research expertise. Recent advances in large language models for complex reasoning have spurred interest in automating this task, translating natural language into executable optimization models. Current approaches, however, achieve limited accuracy, hindered by scarce and noisy training data without leveraging domain knowledge. In this work, we systematically integrate optimization expertise to improve formulation accuracy for mixed-integer linear programming, a key family of mathematical programs. Our OptiMind framework leverages semi-automated, class-based error analysis to guide both training and inference, explicitly preventing common mistakes within each optimization class. Our resulting fine-tuned LLM significantly improves formulation accuracy by 20.7% across multiple optimization benchmarks, with consistent gains under test-time scaling methods such as self-consistency and multi-turn feedback, enabling further progress toward robust LLM-assisted optimization formulation.

cs.LG↗

arXiv2023

Utilitarian Welfare Optimization in the Generalized Vertex Coloring Games: An Implication to Venue Selection in Events Planning

We consider a general class of multi-agent games in networks, namely the generalized vertex coloring games (G-VCGs), inspired by real-life applications of the venue selection problem in events planning. Certain utility responding to the contemporary coloring assignment will be received by each agent under some particular mechanism, who, striving to maximize his own utility, is restricted to local information thus self-organizing when choosing another color. Our focus is on maximizing some utilitarian-looking welfare objective function concerning the cumulative utilities across the network in a decentralized fashion. Firstly, we investigate on a special class of the G-VCGs, namely Identical Preference VCGs (IP-VCGs) which recovers the rudimentary work by \cite{chaudhuri2008network}. We reveal its convergence even under a completely greedy policy and completely synchronous settings, with a stochastic bound on the converging rate provided. Secondly, regarding the general G-VCGs, a greediness-preserved Metropolis-Hasting based policy is proposed for each agent to initiate with the limited information and its optimality under asynchronous settings is proved using theories from the regular perturbed Markov processes. The policy was also empirically witnessed to be robust under independently synchronous settings. Thirdly, in the spirit of ``robust coloring'', we include an expected loss term in our objective function to balance between the utilities and robustness. An optimal coloring for this robust welfare optimization would be derived through a second-stage MH-policy driven algorithm. Simulation experiments are given to showcase the efficiency of our proposed strategy.

cs.DM↗

arXiv2023

Semi-Autoregressive Energy Flows: Exploring Likelihood-Free Training of Normalizing Flows

Training normalizing flow generative models can be challenging due to the need to calculate computationally expensive determinants of Jacobians. This paper studies the likelihood-free training of flows and proposes the energy objective, an alternative sample-based loss based on proper scoring rules. The energy objective is determinant-free and supports flexible model architectures that are not easily compatible with maximum likelihood training, including semi-autoregressive energy flows, a novel model family that interpolates between fully autoregressive and non-autoregressive models. Energy flows feature competitive sample quality, posterior inference, and generation speed relative to likelihood-based flows; this performance is decorrelated from the quality of log-likelihood estimates, which are generally very poor. Our findings question the use of maximum likelihood as an objective or a metric, and contribute to a scientific study of its role in generative modeling.

cs.LG↗

arXiv2022

SelfKG: Self-Supervised Entity Alignment in Knowledge Graphs

Entity alignment, aiming to identify equivalent entities across different knowledge graphs (KGs), is a fundamental problem for constructing Web-scale KGs. Over the course of its development, the label supervision has been considered necessary for accurate alignments. Inspired by the recent progress of self-supervised learning, we explore the extent to which we can get rid of supervision for entity alignment. Commonly, the label information (positive entity pairs) is used to supervise the process of pulling the aligned entities in each positive pair closer. However, our theoretical analysis suggests that the learning of entity alignment can actually benefit more from pushing unlabeled negative pairs far away from each other than pulling labeled positive pairs close. By leveraging this discovery, we develop the self-supervised learning objective for entity alignment. We present SelfKG with efficient strategies to optimize this objective for aligning entities without label supervision. Extensive experiments on benchmark datasets demonstrate that SelfKG without supervision can match or achieve comparable results with state-of-the-art supervised baselines. The performance of SelfKG suggests that self-supervised learning offers great potential for entity alignment in KGs. The code and data are available at https://github.com/THUDM/SelfKG.

cs.LG↗

arXiv2021

A Self-supervised Method for Entity Alignment

Entity alignment, aiming to identify equivalent entities across different knowledge graphs (KGs), is a fundamental problem for constructing large-scale KGs. Over the course of its development, supervision has been considered necessary for accurate alignments. Inspired by the recent progress of self-supervised learning, we explore the extent to which we can get rid of supervision for entity alignment. Existing supervised methods for this task focus on pulling each pair of positive (labeled) entities close to each other. However, our analysis suggests that the learning of entity alignment can actually benefit more from pushing sampled (unlabeled) negatives far away than pulling positive aligned pairs close. We present SelfKG by leveraging this discovery to design a contrastive learning strategy across two KGs. Extensive experiments on benchmark datasets demonstrate that SelfKG without supervision can match or achieve comparable results with state-of-the-art supervised baselines. The performance of SelfKG demonstrates self-supervised learning offers great potential for entity alignment in KGs.

cs.CL↗