arXiv subjects

Jiawei Fu

Publications and source records attributed to Jiawei Fu.

18 recordsLinked to original sources

arXiv2026

Covariate Adjustment in Randomized Experiments: A Unified Framework for Decision and Practice

Should researchers adjust for covariates in randomized experiments, and if so, how? The literature offers three distinct prescriptions: do not adjust because randomization guarantees unbiasedness; adjust for outcome-prognostic covariates to improve precision; or adjust for covariates imbalanced between treatment arms. These competing prescriptions create confusion and uncertainty. We develop a unified framework for decision and practice. Given available information, we show that the optimal correction is what we call ex-post bias. The only relevant criterion for adjustment is prognosticity for ex-post bias; neither raw covariate imbalance nor outcome prognosticity is sufficient by itself. We also show that correcting imbalance and improving precision are two sides of the same decision problem. We develop two estimation approaches, one of which recovers familiar adjustment estimators and provides a new theoretical justification for them. Simulations compare alternative covariate-selection and adjustment strategies. Overall, our framework provides a unified foundation for covariate adjustment in randomized experiments.

stat.ME↗

arXiv2026

Inference for Group Interaction Experiments

A common experimental research design is one in which individuals are randomly allocated into groups that then interact under different group-level treatment conditions. We develop design-based inference for such "group interaction" experiments, covering scenarios in which groups are either fixed or randomly formed and in which potential outcomes are either fixed relative to others' group assignments or subject to interference. For each scenario, we characterize the causal estimand that the design targets and the inferential strategy appropriate to it. Working in a sparse-sampling asymptotic regime, we show that cluster-robust inference remains consistent and accounts for dependencies from various sources when interference is present, delivering valid inference on marginalized exposure effects. When interference is absent and groups are formed randomly, the design reduces to an individually randomized experiment, and individual-level heteroskedasticity-robust inference suffices for the average treatment effect. Our results on the asymptotic distribution of commonly used estimators rely on a novel coupling strategy that may be useful for design-based inference in other complex experiments.

stat.ME↗

arXiv2026

AFUN: Towards an Affordance Foundation Model for Functionality Understanding

Affordance understanding bridges visual perception and physical action, serving as an explainable interface for robot manipulation in open and unstructured real-world environments. Yet, building an affordance foundation model that not only understands where and how the interaction should happen, but also generalizes across diverse environments, objects, and tasks, remains a long-standing research challenge. Existing methods typically address only part of this challenge, either localizing task-relevant regions without specifying executable motion, or predicting motion but with limited scalability. In this paper, we present ourmodel, a step towards an affordance foundation model for functionality understanding. From a single RGB-D observation and a language task description, ourmodel predicts a task-conditional functional mask (where to interact) and a 3D post-contact motion curve (how to interact). To support open-world generalization, we build a large-scale standardized data pipeline that converts heterogeneous robot, human, simulation, and real-world scan data into a shared affordance schema with language, masks, and object-centric 3D motion labels. We evaluate ourmodel from three aspects: for affordance segmentation, ourmodel outperforms all baselines by a large margin across 8 test sets from 4 benchmarks, improving mean gIoU/cIoU by +23.9/+26.3; for contact-point prediction, it predicts substantially more accurate points, with a 12.7--61.3% hit-rate gain over the best baseline; and for 3D motion, it achieves the best performance on all three test sets. ourmodel can be deployed for real-world robot manipulation without finetuning for robot embodiment or using task-specific heuristics, demonstrating the ability to adapt to open-world affordance tasks. Project page: https://www.zhaoningwang.com/AFUN

cs.RO↗

arXiv2026

How Should LLMs Consume High-Quality Data? Optimal Data Scheduling via Quality-Aware Functional Scaling Laws

High-quality data is scarce in large language model (LLM) training, yet how to schedule its use with optimization dynamics lacks theoretical guidance. We extend functional scaling laws with time-varying data quality and derive asymptotically optimal joint data-quality and batch-size schedules within a feature-space regression model. The solution reveals two regimes and dual uses of high-quality data: in the noise-limited regime, a smaller batch converts cleaner data into more signal at comparable noise; in the signal-limited regime, late placement suppresses terminal noise without sacrificing signal accumulation. This explains why conventional decay schedules can conflict with curriculum-style pipelines. Motivated by the theoretical structure, we propose Drop-Stable-Rampup for LLM midtraining: drop the batch size at the quality transition, keep it low to accumulate signal, then ramp up to suppress noise. On a 15B MoE model midtrained on 108B tokens of general-domain proprietary data, Drop-Stable-Rampup improves average accuracy over Warmup-Stable-Decay by +1.70 and Cosine-decay by +2.98, including +4.23 on GSM8K and +2.80 on MATH. On a public math-and-code mixture, it leads all reported STEM, mathematics, and code benchmarks, improving the overall mean over the strongest baseline by +3.27 on a 600M dense model and +5.25 on the same MoE architecture.

cs.LG↗

arXiv2026

Nonparametric Identification and Estimation of Causal Effects on Latent Outcomes

How should researchers conduct causal inference when the outcome of interest is latent and measured imperfectly by multiple indicators? We develop a general nonparametric framework for identifying and estimating average treatment effects on latent outcomes in randomized experiments. We show that latent-outcome estimation faces two distinct noncomparability challenges. First, across studies, different measurement systems may cause estimators to target different empirical quantities even when the underlying latent treatment effect is the same. Second, within a study, different indicators may have different and possibly nonlinear relationships with the same latent outcome, making them not directly comparable. To address these challenges, we propose a design-based approach built around nonparametric bridge functions. We show that these bridge functions can be characterized and identified. Estimation relies on a debiasing procedure that permits valid inference even when the bridge functions are weakly identified. Simulations demonstrate that standard methods, such as principal components analysis and inverse covariance weighting, can generate spurious cross-study differences, whereas our approach recovers comparable latent treatment effects. Overall, the framework provides both a general strategy for causal inference with latent outcomes and practical guidance for designing measurements that support identification, comparability, and efficient estimation.

stat.ME↗

arXiv2026

HE-SNR: Uncovering Latent Logic via Entropy for Guiding Mid-Training on SWE-bench

SWE-bench has emerged as the premier benchmark for evaluating Large Language Models on complex software engineering tasks. While these capabilities are fundamentally acquired during the mid-training phase and subsequently elicited during Supervised Fine-Tuning (SFT), there remains a critical deficit in metrics capable of guiding mid-training effectively. Standard metrics such as Perplexity (PPL) are compromised by the "Long-Context Tax" and exhibit weak correlation with downstream SWE performance. In this paper, we bridge this gap by first introducing a rigorous data filtering strategy. Crucially, we propose the Entropy Compression Hypothesis, redefining intelligence not by scalar Top-1 compression, but by the capacity to structure uncertainty into Entropy-Compressed States of low orders ("reasonable hesitation"). Grounded in this fine-grained entropy analysis, we formulate a novel metric, HE-SNR (High-Entropy Signal-to-Noise Ratio). We validate our approach on models with up to 560B parameters across different context windows (32K/128K). This work provides both the theoretical foundation and practical tools for optimizing the latent potential of LLMs in complex engineering domains.

cs.LG↗

arXiv2025

Scaling Cross-Embodiment World Models for Dexterous Manipulation

Cross-embodiment learning seeks to build generalist robots that learn from and operate across diverse morphologies, but differences in kinematics and action spaces hinder data sharing and control transfer. We ask: What structure can be shared across embodiments despite these differences? We argue that the physical interactions they induce can be modeled in a shared geometric space, allowing world models to provide a common interface for learning and control. To realize this idea, we represent human and robot hands as sets of 3D particles and define actions as end-effector particle displacement fields. This representation abstracts away embodiment-specific joint spaces while preserving the geometry and motion relevant to physical interaction. We train a graph-based world model on random interaction data from diverse simulated robot hands and real human hands, and integrate it with model-predictive control for deployment on new hardware. Experiments on rigid and deformable manipulation reveal three findings: increasing the diversity of training embodiments improves generalization to unseen hands; appropriately combining simulated and real-world data outperforms either source alone; and the same learned model enables effective control on robotic hands with distinct kinematics and degrees of freedom. These results position particle-based world models as a shared interface for learning from and for heterogeneous embodiments.

cs.RO↗

arXiv2025

LongCat-Flash Technical Report

We introduce LongCat-Flash, a 560-billion-parameter Mixture-of-Experts (MoE) language model designed for both computational efficiency and advanced agentic capabilities. Stemming from the need for scalable efficiency, LongCat-Flash adopts two novel designs: (a) Zero-computation Experts, which enables dynamic computational budget allocation and activates 18.6B-31.3B (27B on average) per token depending on contextual demands, optimizing resource usage. (b) Shortcut-connected MoE, which enlarges the computation-communication overlap window, demonstrating notable gains in inference efficiency and throughput compared to models of a comparable scale. We develop a comprehensive scaling framework for large models that combines hyperparameter transfer, model-growth initialization, a multi-pronged stability suite, and deterministic computation to achieve stable and reproducible training. Notably, leveraging the synergy among scalable architectural design and infrastructure efforts, we complete model training on more than 20 trillion tokens within 30 days, while achieving over 100 tokens per second (TPS) for inference at a cost of \$0.70 per million output tokens. To cultivate LongCat-Flash towards agentic intelligence, we conduct a large-scale pre-training on optimized mixtures, followed by targeted mid- and post-training on reasoning, code, and instructions, with further augmentation from synthetic data and tool use tasks. Comprehensive evaluations demonstrate that, as a non-thinking foundation model, LongCat-Flash delivers highly competitive performance among other leading models, with exceptional strengths in agentic tasks. The model checkpoint of LongCat-Flash is open-sourced to foster community research. LongCat Chat: https://longcat.ai Hugging Face: https://huggingface.co/meituan-longcat GitHub: https://github.com/meituan-longcat

cs.CL↗

arXiv2025

LodeStar: Long-horizon Dexterity via Synthetic Data Augmentation from Human Demonstrations

Developing robotic systems capable of robustly executing long-horizon manipulation tasks with human-level dexterity is challenging, as such tasks require both physical dexterity and seamless sequencing of manipulation skills while robustly handling environment variations. While imitation learning offers a promising approach, acquiring comprehensive datasets is resource-intensive. In this work, we propose a learning framework and system LodeStar that automatically decomposes task demonstrations into semantically meaningful skills using off-the-shelf foundation models, and generates diverse synthetic demonstration datasets from a few human demos through reinforcement learning. These sim-augmented datasets enable robust skill training, with a Skill Routing Transformer (SRT) policy effectively chaining the learned skills together to execute complex long-horizon manipulation tasks. Experimental evaluations on three challenging real-world long-horizon dexterous manipulation tasks demonstrate that our approach significantly improves task performance and robustness compared to previous baselines. Videos are available at lodestar-robot.github.io.

cs.RO↗

arXiv2025

Causal Inference for Experiments with Latent Outcomes: Key Results and Their Implications for Design and Analysis

How should researchers analyze randomized experiments in which the main outcome is latent and measured in multiple ways but each measure contains some degree of error? We first identify a critical study-specific noncomparability problem in existing methods for handling multiple measurements, which often rely on strong modeling assumptions or arbitrary standardization. Such approaches render the resulting estimands noncomparable across studies. To address the problem, we describe design-based approaches that enable researchers to identify causal parameters of interest, suggest ways that experimental designs can be augmented so as to make assumptions more credible, and discuss empirical tests of key assumptions. We show that when experimental researchers invest appropriately in multiple outcome measures, an optimally weighted scaled index of these measures enables researchers to obtain efficient and interpretable estimates of causal parameters by applying standard regression. An empirical application illustrates the gains in precision and robustness that multiple outcome measures can provide.

econ.EM↗

arXiv2024

Solving Vision Tasks with Simple Photoreceptors Instead of Cameras

A de facto standard in solving computer vision problems is to use a common high-resolution camera and choose its placement on an agent (i.e., position and orientation) based on human intuition. On the other hand, extremely simple and well-designed visual sensors found throughout nature allow many organisms to perform diverse, complex behaviors. In this work, motivated by these examples, we raise the following questions: 1. How effective simple visual sensors are in solving vision tasks? 2. What role does their design play in their effectiveness? We explore simple sensors with resolutions as low as one-by-one pixel, representing a single photoreceptor First, we demonstrate that just a few photoreceptors can be enough to solve many tasks, such as visual navigation and continuous control, reasonably well, with performance comparable to that of a high-resolution camera. Second, we show that the design of these simple visual sensors plays a crucial role in their ability to provide useful information and successfully solve these tasks. To find a well-performing design, we present a computational design optimization algorithm and evaluate its effectiveness across different tasks and domains, showing promising results. Finally, we perform a human survey to evaluate the effectiveness of intuitive designs devised manually by humans, showing that the computationally found design is among the best designs in most cases.

cs.CV↗

arXiv2024

Multi-objective Cross-task Learning via Goal-conditioned GPT-based Decision Transformers for Surgical Robot Task Automation

Surgical robot task automation has been a promising research topic for improving surgical efficiency and quality. Learning-based methods have been recognized as an interesting paradigm and been increasingly investigated. However, existing approaches encounter difficulties in long-horizon goal-conditioned tasks due to the intricate compositional structure, which requires decision-making for a sequence of sub-steps and understanding of inherent dynamics of goal-reaching tasks. In this paper, we propose a new learning-based framework by leveraging the strong reasoning capability of the GPT-based architecture to automate surgical robotic tasks. The key to our approach is developing a goal-conditioned decision transformer to achieve sequential representations with goal-aware future indicators in order to enhance temporal reasoning. Moreover, considering to exploit a general understanding of dynamics inherent in manipulations, thus making the model's reasoning ability to be task-agnostic, we also design a cross-task pretraining paradigm that uses multiple training objectives associated with data from diverse tasks. We have conducted extensive experiments on 10 tasks using the surgical robot learning simulator SurRoL~\cite{long2023human}. The results show that our new approach achieves promising performance and task versatility compared to existing methods. The learned trajectories can be deployed on the da Vinci Research Kit (dVRK) for validating its practicality in real surgical robot settings. Our project website is at: https://med-air.github.io/SurRoL.

cs.RO↗

arXiv2024

A Formal Theory of Survey Experiment Generalizability: Attention and Salience

Survey experiments are widely used to identify causal effects in political science and the social sciences. Yet researchers are typically interested in more than the internal validity of an experimentally induced contrast. They also want to know whether the estimated effect corresponds to the effect in the real world. We develop a formal theory of survey experiment generalizability grounded in behavioral microfoundations. The theory highlights two mechanisms. First, the survey environment shapes attention: it determines which considerations enter the respondent's active consideration set. Second, it shapes salience: conditional on consideration, it influences the relative weight assigned to those considerations. This framework yields two main results. Consideration-set compression generates amplification: survey-experimental effects can be larger in magnitude than their real-world counterparts, even for the same individuals, treatment content, and outcome. Context-dependent salience generates sign instability: the direction of the survey effect need not coincide with the direction of the corresponding real-world effect. The theory clarifies what survey experiments identify, when those effects are likely to generalize, and how survey designs can be modified to improve decision-environment transportability.

econ.EM↗

arXiv2024

Heterogeneous Treatment Effects and Causal Mechanisms

The credibility revolution advances the use of research designs that permit identification and estimation of causal effects. However, understanding which mechanisms produce measured causal effects remains a challenge. The dominant current approach to the quantitative evaluation of mechanisms relies on the detection of heterogeneous treatment effects (HTEs) with respect to pre-treatment covariates. This paper develops a framework to understand when the existence of such heterogeneous treatment effects can support inferences about the activation of a mechanism. We show first that this design cannot provide evidence of mechanism activation without additional, generally implicit, exclusion assumptions. Further, even when these assumptions are satisfied, the presence of HTEs supports the inference that mechanism is active but the absence of HTEs is generally uninformative about mechanism activation. We provide novel guidance for interpretation and research design in light of these findings.

econ.EM↗

arXiv2024

Extracting Mechanisms from Heterogeneous Effects: An Identification Strategy for Mediation Analysis

Understanding causal mechanisms is crucial for explaining and generalizing empirical phenomena. Causal mediation analysis offers statistical techniques to quantify the mediation effects. Although numerous methods have been developed for causal inference more broadly, the methodological toolkit for causal mediation analysis remains limited. Current methods often require multiple ignorability assumptions or sophisticated research designs. In this paper, we introduce an alternative identification strategy that enables the simultaneous identification and estimation of treatment and mediation effects. By combining explicit and implicit mediation analysis, this strategy leverages heterogeneous treatment effects and does not require addressing some unobserved confounders. Monte Carlo simulations demonstrate that the method is more accurate and precise across various scenarios. To illustrate the efficiency and efficacy of our method, we apply it to estimate the causal mediation effects in two studies with distinct data structures, focusing on common pool resource governance and voting information.

stat.ME↗

arXiv2023

InteractionNet: Joint Planning and Prediction for Autonomous Driving with Transformers

Planning and prediction are two important modules of autonomous driving and have experienced tremendous advancement recently. Nevertheless, most existing methods regard planning and prediction as independent and ignore the correlation between them, leading to the lack of consideration for interaction and dynamic changes of traffic scenarios. To address this challenge, we propose InteractionNet, which leverages transformer to share global contextual reasoning among all traffic participants to capture interaction and interconnect planning and prediction to achieve joint. Besides, InteractionNet deploys another transformer to help the model pay extra attention to the perceived region containing critical or unseen vehicles. InteractionNet outperforms other baselines in several benchmarks, especially in terms of safety, which benefits from the joint consideration of planning and forecasting. The code will be available at https://github.com/fujiawei0724/InteractionNet.

cs.RO↗

arXiv2023

Complementing Onboard Sensors with Satellite Map: A New Perspective for HD Map Construction

High-definition (HD) maps play a crucial role in autonomous driving systems. Recent methods have attempted to construct HD maps in real-time using vehicle onboard sensors. Due to the inherent limitations of onboard sensors, which include sensitivity to detection range and susceptibility to occlusion by nearby vehicles, the performance of these methods significantly declines in complex scenarios and long-range detection tasks. In this paper, we explore a new perspective that boosts HD map construction through the use of satellite maps to complement onboard sensors. We initially generate the satellite map tiles for each sample in nuScenes and release a complementary dataset for further research. To enable better integration of satellite maps with existing methods, we propose a hierarchical fusion module, which includes feature-level fusion and BEV-level fusion. The feature-level fusion, composed of a mask generator and a masked cross-attention mechanism, is used to refine the features from onboard sensors. The BEV-level fusion mitigates the coordinate differences between features obtained from onboard sensors and satellite maps through an alignment module. The experimental results on the augmented nuScenes showcase the seamless integration of our module into three existing HD map construction methods. The satellite maps and our proposed module notably enhance their performance in both HD map semantic segmentation and instance detection tasks.

cs.CV↗

arXiv2022

Learning Deep Sensorimotor Policies for Vision-based Autonomous Drone Racing

Autonomous drones can operate in remote and unstructured environments, enabling various real-world applications. However, the lack of effective vision-based algorithms has been a stumbling block to achieving this goal. Existing systems often require hand-engineered components for state estimation, planning, and control. Such a sequential design involves laborious tuning, human heuristics, and compounding delays and errors. This paper tackles the vision-based autonomous-drone-racing problem by learning deep sensorimotor policies. We use contrastive learning to extract robust feature representations from the input images and leverage a two-stage learning-by-cheating framework for training a neural network policy. The resulting policy directly infers control commands with feature representations learned from raw images, forgoing the need for globally-consistent state estimation, trajectory planning, and handcrafted control design. Our experimental results indicate that our vision-based policy can achieve the same level of racing performance as the state-based policy while being robust against different visual disturbances and distractors. We believe this work serves as a stepping-stone toward developing intelligent vision-based autonomous systems that control the drone purely from image inputs, like human pilots.

cs.RO↗