Searcharxiv⌕ Search

arXiv subjects

Wentao Guo

Publications and source records attributed to Wentao Guo.

At least 19 recordsLinked to original sources
arXiv2026

Social-JEPA: Emergent Geometric Isomorphism

World models compress rich sensory streams into compact latent codes that anticipate future observations. We let separate agents acquire such models from distinct viewpoints of the same environment without any parameter sharing or coordination. After training, their internal representations exhibit a striking emergent property: the two latent spaces are related by an approximate linear isometry, enabling transparent translation between them. This geometric consensus survives large viewpoint shifts and scant overlap in raw pixels. Leveraging the learned alignment, a classifier trained on one agent can be ported to the other with no additional gradient steps, while distillation-like migration accelerates later learning and markedly reduces total compute. The findings reveal that predictive learning objectives impose strong regularities on representation geometry, suggesting a lightweight path to interoperability among decentralized vision systems. The code is available at https://anonymous.4open.science/r/Social-JEPA-5C57.

cs.CV↗
arXiv2026

MLS-Bench: A Holistic and Rigorous Assessment of AI Systems on Building Better AI

Modern AI progress has been driven by ML methods that are generalizable across settings and scalable to larger regimes. As large language models demonstrate advanced capabilities in reasoning, coding, and engineering tasks, it is increasingly important to understand whether they can discover such methods rather than only apply existing ones. We introduce MLS-Bench, a benchmark for evaluating whether AI systems can invent generalizable and scalable ML methods. MLS-Bench contains 140 tasks across 12 domains, each requiring an agent to improve one targeted component of an ML system or algorithm and demonstrate that the improvement generalizes across controlled settings and scales. We find that current agents remain far from reliably surpassing human-designed methods, and that engineering-style tuning is easier for them than genuine method invention. We further study the effects of test-time scaling, adaptive compute allocation, and context provision on agents' discovery performance, together with case studies of their behavior. Our analyses suggest that the bottleneck is not only in proposing new methods, but also in the scientific insight needed to plan, validate, and scale claims about them. More search, compute, or context alone does not remove this bottleneck. We build and maintain a community platform for cumulative and comparable iteration, and release the data and code at https://mls-bench.com.

cs.LG↗
arXiv2026

RoboLineage: Agent-Native Data Lifecycle Governance Across Robot Policy Iterations

We present RoboLineage, an agent-native data lifecycle governance system for robot policy iteration. Modern robot policies improve through repeated data collection, review, retraining, evaluation, and release decisions, but the evidence connecting these steps is often scattered across local tools, scripts, and expert memory. RoboLineage makes this lifecycle explicit by representing rollouts, reviews, dataset decisions, training runs, policy metadata, evaluations, deployment recommendations, and next-collection plans as typed lineage artifacts. Agents interpret embodied rollout evidence, adapt accepted data to existing training stacks, maintain data health, and summarize cross-iteration state under explicit artifact boundaries. In real-robot manipulation workflows, RoboLineage makes routine policy iteration faster and more auditable while maintaining downstream policy performance. We open source RoboLineage as a lightweight lifecycle layer for different robot embodiments and training families. Project page: https://robolineage.github.io/

cs.RO↗
arXiv2026

A Conformation-Centric Generative Foundation Model for Linear Polymer Modeling and Design

Linear polymers, macromolecules formed from monomers covalently bonded into continuous chains, underpin countless technologies and are indispensable to modern life. While deep learning is advancing polymer science, existing methods typically represent the whole linear polymer solely through monomer-level descriptors, overlooking the global structural information inherent in polymer conformations, which ultimately limits their practical performance. Moreover, this important field still lacks a dedicated foundation model that can effectively support diverse downstream tasks, thereby severely constraining progress. To address these challenges, we introduce PolyConFM, a foundation model tailored for modeling and designing linear polymers through conformation-centric generative pretraining. Recognizing that each linear polymer is essentially a continuous chain whose conformation can be naturally decomposed into a sequence of local conformations (i.e., those of its repeating units), we pretrain PolyConFM under the conditional generation paradigm, reconstructing these local conformations via masked autoregressive (MAR) modeling and further generating their orientation transformations to recover the corresponding polymer conformation. Meanwhile, we construct a linear polymer conformation dataset via molecular dynamics simulations to mitigate data sparsity, thereby enabling conformation-centric pretraining. Experiments demonstrate that PolyConFM consistently outperforms representative task-specific methods across diverse downstream tasks, thereby equipping polymer science with a powerful tool targeting linear polymers.

cs.LG↗
arXiv2026

MuCO: Generative Peptide Cyclization Empowered by Multi-stage Conformation Optimization

Modeling peptide cyclization is critical for the virtual screening of candidate peptides with desirable physical and pharmaceutical properties. This task is challenging because a cyclic peptide often exhibits diverse, ring-shaped conformations, which cannot be well captured by deterministic prediction models derived from linear peptide folding. In this study, we propose MuCO (Multi-stage Conformation Optimization), a generative peptide cyclization method that models the distribution of cyclic peptide conformations conditioned on the corresponding linear peptide. In principle, MuCO decouples the peptide cyclization task into three stages: topology-aware backbone design, generative side-chain packing, and physics-aware all-atom optimization, thereby generating and optimizing conformations of cyclic peptides in a coarse-to-fine manner. This multi-stage framework enables an efficient parallel sampling strategy for conformation generation and allows for rapid exploration of diverse, low-energy conformations. Experiments on the large-scale CPSea dataset demonstrate that MuCO significantly and consistently outperforms state-of-the-art methods in physical stability, structural diversity, secondary structure recovery, and computational efficiency, making it a promising computational tool for exploring and designing cyclic peptides. The demo of the proposed method can be found at https://github.com/mianqiu00/MuCO.

q-bio.BM↗
arXiv2026

FLUID: From Ephemeral IDs to Multimodal Semantic Codes for Industrial-Scale Livestreaming Recommendation

Modern recommender systems rely heavily on ID-based collaborative filtering: each item is represented by a unique ID embedding that accumulates collaborative signals from user interactions. Livestreaming recommendation, however, faces a unique challenge in this paradigm: a live room typically broadcasts for only tens of minutes, so its item ID remains poorly learned in a persistent cold-start state and ID-centric ranking models fail to generalize. We present FLUID, the first framework to fully retire the candidate-side item ID from a production-scale livestreaming ranker. FLUID introduces a cross-domain multimodal encoder, jointly trained on short videos and livestreams, to produce discrete hierarchical semantic codes, called LUCID, for content-based item characterization. To adapt the ranker to LUCID, FLUID further employs a staged warmup scheme: it first incorporates cold, slice-level LUCID as an independent token alongside the ID embedding, and then replaces the ID embedding with warm, room-level LUCID before online incremental training. Deployed on our industrial livestreaming recommenders with a cross-platform combined user base of over one billion globally, FLUID delivers significant online gains of +0.55% Quality Watch Duration, +2.05% Cold-Start Room Views, and +0.05% Active Hours.

cs.AI↗
arXiv2026

Uncertainty-Calibrated Recommendations for Low-Active Users

A fundamental challenge in recommender systems is balancing reliability for Low-Active Users (LAUs) with diversity for High-Active Users (HAUs). The key to this balance lies in quantifying model uncertainty, which approximates the risk of prediction errors and reveals the limits of the model's current knowledge. On large-scale short-video and livestream platforms, model uncertainty can warn of low-quality recommendations that may lead to disengagement of LAUs and at the same time identify opportunities to diversify content recommendation for HAUs. To leverage this dichotomy, we introduce a unified, production-ready framework that calibrates uncertainty to drive differentiated strategies. Specifically, we implement a model-uncertainty-based risk-averse deboosting policy for LAUs to suppress unreliable recommendations, while employing a risk-seeking Upper Confidence Bound (UCB) strategy for HAUs to encourage exploration. Validated on a major livestream platform, our framework demonstrates significant improvements in retention (active hours) and satisfaction (quality watch time ratio) for LAUs as well as remarkable increases in interest diversity and category coverage for HAUs, proving the value of uncertainty-aware recommendation in industrial settings.

cs.IR↗
arXiv2026

TwinRouterBench: Fast Static and Live Dynamic Evaluation for Realistic Agentic LLM Routing

LLM routing matters most in long-horizon applications such as coding agents, deep research systems, and computer-use agents, where a single user request triggers many model calls. Routing each call to the cheapest sufficient model can cut costs without sacrificing quality, yet existing router benchmarks evaluate routers only on one-shot prompts. They never expose the router-visible prefix at an intermediate agent step, never test whether a cheaper replacement preserves downstream task success, and often rely on online LLM judges at evaluation time. We introduce TwinRouterBench, a step-level routing benchmark with two tracks. The static track provides 970 router-visible prefixes from 520 instances across SWE-bench, BFCL, mtRAG, QMSum, and PinchBench, each paired with an execution-verified target tier estimated under a released downgrade-and-cascade protocol; scoring is deterministic arithmetic over tier labels, trajectory membership, and token costs, with no online evaluator-side LLM judge. The dynamic track supplies a harness that runs routers on the full 500-case SWE-bench Verified suite; in this paper we report a 100-case held-out evaluation disjoint from the static SWE supervision split. At each LLM call the router selects a concrete model from a locked pool, and success is measured by official task resolution and realized API spend. The two tracks support fast offline iteration followed by end-to-end validation under live agent execution. Code and data are available at https://github.com/CommonstackAI/TwinRouterBench.

cs.LG↗
arXiv2026

PEARL: Unbiased Percentile Estimation via Contrastive Learning for Industrial-Scale Livestream Recommendation

Recommender systems trained on user interaction data are susceptible to behavioral intensity imbalance--a systematic distortion arising from heterogeneous engagement patterns across users. This imbalance skews feedback signals such that observed interactions no longer faithfully reflect true preferences, causing models to disproportionately amplify signals from highly active users while underrepresenting others, which ultimately degrades recommendation quality and robustness at scale. To address this issue, we propose a nonparametric contrastive percentile approximation framework, PEARL, that models relative preference signals instead of absolute engagement magnitudes. Building upon relative advantage debiasing, PEARL leverages real contrastive interaction samples to approximate percentile relationships directly, without relying on auxiliary distribution estimation models. We provide theoretical justification demonstrating that such pairwise comparisons yield unbiased estimates of percentile-based preference signals. For broader applicability, we introduce a prediction-based bootstrapping mechanism for percentile smoothing to handle sparse and discrete feedback, alongside a generalized value-weighted formulation and a co-training strategy to enhance both modeling flexibility and representation learning. Extensive offline experiments demonstrate that PEARL effectively mitigates behavioral bias and consistently improves recommendation performance across multiple ranking targets. Deployed in a production livestream platform with a combined user base of billions, online A/B testing confirms substantial real-world gains: +2.10% Watch Duration, +0.80% Consumption Amount, +1.49% Interaction Rate, and -6.91% Report Rate.

cs.LG↗
arXiv2026

SonicMoE: Accelerating MoE with IO and Tile-aware Optimizations

Mixture of Experts (MoE) models have emerged as the de facto architecture for scaling up language models without significantly increasing the computational cost. Recent MoE models demonstrate a clear trend towards high expert granularity (smaller expert intermediate dimension) and higher sparsity (constant number of activated experts with a higher number of total experts), which improve model quality per FLOP. However, fine-grained MoEs suffer from increased activation memory footprint and reduced hardware efficiency due to higher IO costs, while sparser MoEs suffer from wasted computations due to padding in Grouped GEMM kernels. In response, we propose a memory-efficient algorithm to compute the forward and backward passes of MoEs with minimal activation caching for the backward pass. We also design GPU kernels that overlap memory IO with computation, benefiting all MoE architectures. Finally, we propose a novel "token rounding" method that minimizes the wasted compute due to padding in Grouped GEMM kernels. As a result, our method SonicMoE reduces activation memory by 45% and achieves a 1.86x compute throughput improvement on Hopper GPUs compared to ScatterMoE's BF16 MoE kernel for a fine-grained 7B MoE. Concretely, SonicMoE on 64 H100s achieves a training throughput of 213 billion tokens per day, comparable to ScatterMoE's 225 billion tokens per day on 96 H100s for a 7B MoE model training with FSDP-2 using the lm-engine codebase. On Blackwell GPUs, SonicMoE also achieves a 25% and 15% relative speedup on the forward and backward pass respectively compared to a highly optimized DeepGEMM baseline on OLMoE-sized 7B MoE models. Under high MoE sparsity settings, our tile-aware token rounding algorithm yields an additional 1.16x speedup on kernel execution time compared to vanilla top-K routing while maintaining similar downstream performance on Hopper GPUs. We open-source all our kernels.

cs.LG↗
arXiv2026

MolReasoner: Toward Effective and Interpretable Reasoning for Molecular LLMs

Large Language Models (LLMs) have shown impressive performance across various domains, but their ability to perform molecular reasoning remains underexplored. Existing methods mostly rely on general-purpose prompting, which lacks domain-specific molecular semantics, or fine-tuning, which faces challenges in interpretability and reasoning depth, often leading to structural and textual hallucinations. To address these issues, we introduce MolReasoner, a two-stage framework that transitions LLMs from memorization to high-fidelity chemical reasoning. In the Mol-SFT stage, knowledge-enhanced Chain-of-Thought (CoT) data provides a strong foundation, while the Mol-RL stage refines reasoning using a novel, task-adaptive reward system to mitigate hallucinations. Extensive evaluations demonstrate that MolReasoner significantly outperforms a wide range of strong baselines in both molecule generation and captioning tasks. Further analyses highlight the framework's synergistic design and its ability to produce more interpretable outputs. Our work presents a principled and effective new approach for advancing high-fidelity molecular reasoning.

cs.LG↗
arXiv2026

Zenith: Scaling up Ranking Models for Billion-scale Livestreaming Recommendation

Accurately capturing feature interactions is essential in recommender systems, and recent trends show that scaling up model capacity could be a key driver for next-level predictive performance. While prior work has explored various model architectures to capture multi-granularity feature interactions, relatively little attention has been paid to efficient feature handling and scaling model capacity without incurring excessive inference latency. In this paper, we address this by presenting Zenith, a scalable and efficient ranking architecture that learns complex feature interactions with minimal runtime overhead. Zenith is designed to handle a few high-dimensional Prime Tokens with Token Fusion and Token Boost modules, which exhibits superior scaling laws compared to other state-of-the-art ranking methods, thanks to its improved token heterogeneity. Its real-world effectiveness is demonstrated by deploying the architecture to TikTok Live, a leading online livestreaming platform that attracts billions of users globally. Our A/B test shows that Zenith achieves +1.05%/-1.10% in online CTR AUC and Logloss, and realizes +9.93% gains in Quality Watch Session / User and +8.11% in Quality Watch Duration / User.

cs.LG↗
arXiv2025

SCAL for Pinch-Lifting: Complementary Rotational and Linear Prototypes for Environment-Adaptive Grasping

This paper presents environment-adaptive pinch-lifting built on a slot-constrained adaptive linkage (SCAL) and instantiated in two complementary fingers: SCAL-R, a rotational-drive design with an active fingertip that folds inward after contact to form an envelope, and SCAL-L, a linear-drive design that passively opens on contact to span wide or weak-feature objects. Both fingers convert surface following into an upward lifting branch while maintaining fingertip orientation, enabling thin or low-profile targets to be raised from supports with minimal sensing and control. Two-finger grippers are fabricated via PLA-based 3D printing. Experiments evaluate (i) contact-preserving sliding and pinch-lifting on tabletops, (ii) ramp negotiation followed by lift, and (iii) handling of bulky objects via active enveloping (SCAL-R) or contact-triggered passive opening (SCAL-L). Across dozens of trials on small parts, boxes, jars, and tape rolls, both designs achieve consistent grasps with limited tuning. A quasi-static analysis provides closed-form fingertip-force models for linear parallel pinching and two-point enveloping, offering geometry-aware guidance for design and operation. Overall, the results indicate complementary operating regimes and a practical path to robust, environment-adaptive grasping with simple actuation.

cs.RO↗
arXiv2025

Hoecken-D Hand: A Novel Robotic Hand for Linear Parallel Pinching and Self-Adaptive Grasping

This paper presents the Hoecken-D Hand, an underactuated robotic gripper that combines a modified Hoecken linkage with a differential spring mechanism to achieve both linear parallel pinching and a mid-stroke transition to adaptive envelope. The original Hoecken linkage is reconfigured by replacing one member with differential links, preserving straight-line guidance while enabling contact-triggered reconfiguration without additional actuators. A double-parallelogram arrangement maintains fingertip parallelism during conventional pinching, whereas the differential mechanism allows one finger to wrap inward upon encountering an obstacle, improving stability on irregular or thin objects. The mechanism can be driven by a single linear actuator, minimizing complexity and cost; in our prototype, each finger is driven by its own linear actuator for simplicity. We perform kinematic modeling and force analysis to characterize grasp performance, including simulated grasping forces and spring-opening behavior under varying geometric parameters. The design was prototyped using PLA-based 3D printing, achieving a linear pinching span of approximately 200 mm. Preliminary tests demonstrate reliable grasping in both modes across a wide range of object geometries, highlighting the Hoecken-D Hand as a compact, adaptable, and cost-effective solution for manipulation in unstructured environments.

cs.RO↗
arXiv2025

A Novel Robot Hand with Hoeckens Linkages and Soft Phalanges for Scooping and Self-Adaptive Grasping in Environmental Constraints

This paper presents a novel underactuated adaptive robotic hand, Hockens-A Hand, which integrates the Hoeckens mechanism, a double-parallelogram linkage, and a specialized four-bar linkage to achieve three adaptive grasping modes: parallel pinching, asymmetric scooping, and enveloping grasping. Hockens-A Hand requires only a single linear actuator, leveraging passive mechanical intelligence to ensure adaptability and compliance in unstructured environments. Specifically, the vertical motion of the Hoeckens mechanism introduces compliance, the double-parallelogram linkage ensures line contact at the fingertip, and the four-bar amplification system enables natural transitions between different grasping modes. Additionally, the inclusion of a mesh-textured silicone phalanx further enhances the ability to envelop objects of various shapes and sizes. This study employs detailed kinematic analysis to optimize the push angle and design the linkage lengths for optimal performance. Simulations validated the design by analyzing the fingertip motion and ensuring smooth transitions between grasping modes. Furthermore, the grasping force was analyzed using power equations to enhance the understanding of the system's performance.Experimental validation using a 3D-printed prototype demonstrates the three grasping modes of the hand in various scenarios under environmental constraints, verifying its grasping stability and broad applicability.

cs.RO↗
arXiv2025

Mitigating Non-IID Drift in Zeroth-Order Federated LLM Fine-Tuning with Transferable Sparsity

Federated Learning enables collaborative fine-tuning of Large Language Models (LLMs) across decentralized Non-Independent and Identically Distributed (Non-IID) clients, but such models' massive parameter sizes lead to significant memory and communication challenges. This work introduces Meerkat, a sparse zeroth-order optimization (ZO) method designed for federated LLM fine-tuning. By limiting fine-tuning to a transferable, static, extremely sparse subset of parameters, Meerkat achieves remarkable communication efficiency, enabling cost-effective high-frequency synchronization. With theoretical analysis and experiments, we show that this high-frequency communication effectively mitigates Non-IID data challenges and leads to superior performance compared to full-parameter ZO. Furthermore, experiment results show that Meerkat outperforms existing sparsity baselines with better performance at the same communication frequency. To further handle Non-IID drift, Meerkat leverages traceable local updates and forms a virtual path for each client. This virtual path mechanism reveals the GradIP phenomenon: the inner products between LLM pre-training gradients maintained by server and client gradients estimated via ZO converges for extreme Non-IID clients but oscillates for IID ones. This distinct behavior provides a signal for identifying clients with extreme data heterogeneity. Using this signal, Meerkat-vp is proposed to analyze GradIP trajectories to identify extreme Non-IID clients and applies early stopping to enhance aggregated model quality. Experiments confirm that Meerkat and Meerkat-vp significantly improve the efficiency and effectiveness of ZO federated LLM fine-tuning.

cs.LG↗
arXiv2025

PolyConf: Unlocking Polymer Conformation Generation through Hierarchical Generative Models

Polymer conformation generation is a critical task that enables atomic-level studies of diverse polymer materials. While significant advances have been made in designing conformation generation methods for small molecules and proteins, these methods struggle to generate polymer conformations due to their unique structural characteristics. Meanwhile, the scarcity of polymer conformation datasets further limits the progress, making this important area largely unexplored. In this work, we propose PolyConf, a pioneering tailored polymer conformation generation method that leverages hierarchical generative models to unlock new possibilities. Specifically, we decompose the polymer conformation into a series of local conformations (i.e., the conformations of its repeating units), generating these local conformations through an autoregressive model, and then generating their orientation transformations via a diffusion model to assemble them into the complete polymer conformation. Moreover, we develop the first benchmark with a high-quality polymer conformation dataset derived from molecular dynamics simulations to boost related research in this area. The comprehensive evaluation demonstrates that PolyConf consistently outperforms existing conformation generation methods, thus facilitating advancements in polymer modeling and simulation.

cond-mat.soft↗
arXiv2025

Sensitivity Meets Sparsity: The Impact of Extremely Sparse Parameter Patterns on Theory-of-Mind of Large Language Models

This paper investigates the emergence of Theory-of-Mind (ToM) capabilities in large language models (LLMs) from a mechanistic perspective, focusing on the role of extremely sparse parameter patterns. We introduce a novel method to identify ToM-sensitive parameters and reveal that perturbing as little as 0.001% of these parameters significantly degrades ToM performance while also impairing contextual localization and language understanding. To understand this effect, we analyze their interaction with core architectural components of LLMs. Our findings demonstrate that these sensitive parameters are closely linked to the positional encoding module, particularly in models using Rotary Position Embedding (RoPE), where perturbations disrupt dominant-frequency activations critical for contextual processing. Furthermore, we show that perturbing ToM-sensitive parameters affects LLM's attention mechanism by modulating the angle between queries and keys under positional encoding. These insights provide a deeper understanding of how LLMs acquire social reasoning abilities, bridging AI interpretability with cognitive science. Our results have implications for enhancing model alignment, mitigating biases, and improving AI systems designed for human interaction.

cs.CL↗