Siyuan Zhu

Kshitiz Bansal, Keshav Rungta, Siyuan Zhu et al.

Autonomous perception requires high-quality environment sensing in the form of 3D bounding boxes of dynamic objects. The primary sensors used in automotive systems are light-based cameras and LiDARs. However, they are known to fail in adverse weather conditions. Radars can potentially solve this problem as they are barely affected by adverse weather conditions. However, specular reflections of wireless signals cause poor performance of radar point clouds. We introduce Pointillism, a system that combines data from multiple spatially separated radars with an optimal separation to mitigate these problems. We introduce a novel concept of Cross Potential Point Clouds, which uses the spatial diversity induced by multiple radars and solves the problem of noise and sparsity in radar point clouds. Furthermore, we present the design of RP-net, a novel deep learning architecture, designed explicitly for radar's sparse data distribution, to enable accurate 3D bounding box estimation. The spatial techniques designed and proposed in this paper are fundamental to radars point cloud distribution and would benefit other radar sensing applications.

Siyuan Zhu, Chao Yu, Rongxin Yang et al.

Reinforcement learning has become a powerful paradigm for post-training large language model agents, yet credit assignment in multi-turn environments remains a challenge. Agents often receive sparse, trajectory-level rewards only at the end of an episode, making it difficult to determine which intermediate actions contributed to success or failure. As a result, propagating delayed outcomes back to individual decision steps without relying on costly auxiliary value models remains an open problem. We propose Generalized Advantage Grouped Policy Optimization (GAGPO), a critic-free reinforcement learning method for precise, step-aligned temporal credit assignment. GAGPO constructs a non-parametric grouped value proxy from sampled rollouts and uses it to compute TD/GAE-style temporal advantages, recursively propagating outcome supervision backward through time. Combined with group-wise advantage normalization and an action-level importance ratio, GAGPO extracts stable, localized optimization signals directly from multi-turn trajectories. Experiments on ALFWorld and WebShop show that GAGPO outperforms strong reinforcement learning baselines. Further analyses demonstrate faster early-stage learning, improved interaction efficiency, and smoother optimization dynamics, suggesting that GAGPO offers a simple yet effective framework for multi-turn agentic reinforcement learning.

Siyuan Zhu, Chengdong Xu, Kaiqiang Ke et al.

In long-context question answering, selecting the appropriate scope of context for a query remains a key and unresolved challenge. Insufficient context can lead to missing essential information, whereas excessive context often introduces noise and degrades answer quality. Conventional methods, such as retrieving a fixed number of passages or applying reranking, struggle to dynamically determine which context to include. This is especially problematic for factoid questions, which typically depend only on a few precise pieces of evidence. To overcome this limitation, we propose Context-Picker, a reasoning-aware framework that reframes context selection as the task of identifying a minimal sufficient evidence subset, moving beyond conventional similarity-based ranking. Context-Picker uses a human-inspired two-stage reinforcement learning schedule: stage 1 focuses on improving the recall rate of critical passages, and stage 2 prioritizes pruning redundancy to distill a compact evidence set. To resolve reward sparsity, we propose an offline evidence distillation pipeline that mines ``minimal sufficient sets" via a Leave-One-Out (LOO) procedure, providing dense and task-aligned supervision. Experiments on five long-context and multi-hop QA datasets demonstrate that our method outperforms strong RAG baselines and achieved higher answer accuracy. Ablation studies also indicate that our coarse-to-fine optimization schedule, the redundancy-aware reward shaping, along with the rationale generated by the policy, all contribute substantially to these gains.

JingWei Xu, Siyuan Zhu, Zenan Li et al.

Deep neural networks (DNNs) have become a key part of many modern software applications. After training and validating, the DNN is deployed as an irrevocable component and applied in real-world scenarios. Although most DNNs are built meticulously with huge volumes of training data, data in the real world still remain unknown to the DNN model, which leads to the crucial requirement of runtime out-of-distribution (OOD) detection. However, many existing approaches 1) need OOD data for classifier training or parameter tuning, or 2) simply combine the scores of each hidden layer as an ensemble of features for OOD detection. In this paper, we present a novel outlook on in-distribution data in a generative manner, which takes their latent features generated from each hidden layer as a joint distribution across representation spaces. Since only the in-distribution latent features are comprehensively understood in representation space, the internal difference between in-distribution and OOD data can be naturally revealed without the intervention of any OOD data. Specifically, We construct a generative model, called Latent Sequential Gaussian Mixture (LSGM), to depict how the in-distribution latent features are generated in terms of the trace of DNN inference across representation spaces. We first construct the Gaussian Mixture Model (GMM) based on in-distribution latent features for each hidden layer, and then connect GMMs via the transition probabilities of the inference traces. Experimental evaluations on popular benchmark OOD datasets and models validate the superiority of the proposed method over the state-of-the-art methods in OOD detection.