Institution profile

University of California

Academic institutionnorthamerica · us
Official website
Research library133linked papers
Opportunities0open roles
Selected work

Representative Papers

Provably safe and human-like car-following behaviors: Part 1. Analysis of phases and dynamics in standard models

May 15, 2025

Existing car-following models lack rigorous safety guarantees and provable consistency with human driving behavior. Method: We systematically analyze the phase-plane dynamics of mainstream models (e.g., IDM, Gipps, Newell), formulate a novel multi-order framework—spanning zeroth-order (minimum spacing, comfort), first-order (speed, time headway), and second-order (acceleration/deceleration bounds, braking curves)—and conduct phase-plane modeling, multi-phase dynamical systems analysis, and formal stability and safety proofs. We also rigorously derive the Newell simplified model. Results: Our analysis exposes fundamental deficiencies in existing models regarding safety boundary adherence and human-like behavior consistency. Numerical simulations and empirical validation confirm these insights and establish a solid theoretical foundation for our subsequent multi-phase projection-based car-following model.

1 citations1 influentialRead paper

Differential Voting: Loss Functions For Axiomatically Diverse Aggregation of Heterogeneous Preferences

Jan 25, 2026

This work addresses a critical limitation in current reinforcement learning from human feedback (RLHF) methods, which implicitly aggregate heterogeneous preferences without explicit control over social choice axioms, leading to opaque normative assumptions in the learned reward functions. To remedy this, the authors propose the Differential Voting framework, which for the first time reformulates classical voting rules—such as Copeland and Kemeny—as instance-level differentiable loss functions, ensuring that the optimization objective precisely aligns with a specified voting mechanism at the population level. Through consistency analysis, gradient field modeling, and asymptotic studies of smoothing parameters, the paper systematically uncovers the geometric structure of these losses and their correspondence to foundational social choice axioms, enabling principled axiom-based trade-offs in RLHF. Experiments confirm the alignment between the proposed method and its target voting rules, and the implementation is publicly released.

1 citationsRead paper

DIML: Differentiable Inverse Mechanism Learning from Behaviors of Multi-Agent Learning Trajectories

Jan 25, 2026

This study addresses the problem of inferring unknown and unstructured incentive mechanisms—such as those represented by neural networks—from observed learning trajectories of self-interested multi-agent systems. To this end, the authors propose the Differentiable Inverse Multi-Agent Learning (DIML) framework, which models differentiable multi-agent learning dynamics and integrates a conditional Logit response model with maximum likelihood estimation to enable end-to-end recovery of the underlying incentives. This work presents the first differentiable inverse learning approach capable of handling unstructured incentive functions, establishes identifiability conditions for incentive differences, and proves the statistical consistency of the resulting estimator. Empirical evaluations demonstrate that DIML accurately recovers identifiable incentive structures, supports counterfactual prediction, matches the performance of an enumeration oracle in small-scale settings, and scales effectively to games involving hundreds of agents.

1 citationsRead paper

SpANNS: Optimizing Approximate Nearest Neighbor Search for Sparse Vectors Using Near Memory Processing

Jan 06, 2026arXiv.org

This work addresses the challenge of achieving both high scalability and efficiency in approximate nearest neighbor search (ANNS) over sparse vectors on conventional CPU architectures. To this end, we propose SpANNS—the first near-memory computing architecture tailored for sparse ANNS—built upon the CXL Type-2 platform. SpANNS integrates a hybrid inverted index, query parsing, clustering-based filtering, and a compute-enabled DIMM co-processing mechanism to perform index traversal and distance computation efficiently near the data. Evaluated against the state-of-the-art CPU baseline, SpANNS achieves a speedup of 15.2× to 21.6×, substantially enhancing both performance and scalability for sparse vector retrieval.

1 citationsRead paper

Provably safe and human-like car-following behaviors: Part 2. A parsimonious multi-phase model with projected braking

May 15, 2025

Autonomous car-following under real-world uncertainty must simultaneously guarantee formal safety and human-like driving behavior, yet existing models fail to satisfy both requirements. Method: We propose the first provably safe and human-like multi-phase car-following model, integrating an extended Newell model with a novel projection-based braking control law. We introduce the first modeling framework for projection-based braking, rigorously defining leader-follower braking profiles and establishing phase-transition criteria. Within a unified framework, we jointly achieve formal safety verification and human-like acceleration/deceleration dynamics. Results: Experiments demonstrate 100% collision-free stopping when following a stationary lead vehicle; safety stopping distance error is below 5%; and acceleration profiles achieve a Pearson correlation coefficient of 0.92 with human drivers—significantly outperforming baseline models.

1 citationsRead paper
Recent publications

Latest Papers