Selected work

Representative Papers

Distilling Future Temporal Knowledge with Masked Feature Reconstruction for 3D Object Detection

Dec 09, 2025

To address the challenge of incorporating future-frame information in online 3D object detection, this paper proposes a Future Temporal Knowledge Distillation (FTKD) framework that relaxes the strict frame-wise alignment constraint inherent in conventional knowledge distillation. Methodologically, FTKD introduces a sparse query mechanism and a future-aware feature reconstruction strategy, jointly optimized with foreground-background contextual modeling to efficiently extract and transfer future temporal knowledge from an offline teacher model to an online student. Additionally, future-guided logit distillation is incorporated to enhance the student’s modeling capability for motion dynamics and temporal consistency. Evaluated on the nuScenes dataset, FTKD achieves consistent improvements of +1.3 mAP and +1.3 NDS over strong baselines, with notable gains in velocity estimation accuracy—while incurring no additional computational overhead during online inference.

0 citationsRead paper

HMVLM: Multistage Reasoning-Enhanced Vision-Language Model for Long-Tailed Driving Scenarios

Jun 06, 2025

To address the trade-off between generalization and real-time performance in long-tail driving scenarios, this paper proposes a dual-branch end-to-end autonomous driving framework: a “fast” branch for low-latency control output and a “slow” branch leveraging multi-view images and 4-second ego-motion history to invoke a vision-language model (VLM) for selective five-view prompting and multi-stage chain-of-reasoning—comprising scene understanding, decision-making, and trajectory inference—to generate high-level driving intent. A spline-based trajectory post-processing module further enhances action smoothness. The method is supervised on the Waymo Open Dataset and achieved second place in the 2025 Waymo Vision-Only End-to-End Challenge, attaining a Rater Feedback Score of 7.7367—2.77% higher than the baseline—demonstrating significantly improved generalization to long-tail scenarios and enhanced decision rationality.

0 citationsRead paper

HMAD: Advancing E2E Driving with Anchored Offset Proposals and Simulation-Supervised Multi-target Scoring

May 29, 2025

End-to-end autonomous driving faces two key challenges: insufficient trajectory diversity and weak safety evaluation. To address these, we propose a generation-evaluation decoupled framework. First, we generate a large set of compliant and stable candidate trajectories via iterative offset decoding, leveraging BEVFormer and learnable anchor queries. Second, we introduce a simulation-driven multi-objective scoring module that jointly optimizes fault-free collision rate, drivable-area coverage, ride comfort, and an extended Planning Decision Metric (PDM). Our method pioneers an anchor-based offset trajectory proposal mechanism and a multi-task scorer network. Evaluated on the CVPR 2025 private test set, it achieves a 44.5% driving score—significantly outperforming baseline methods. This paradigm synergistically enhances both trajectory generation quality and evaluation robustness, establishing a novel, interpretable, and verifiable approach to safe path selection for end-to-end driving decision-making.

0 citationsRead paper

CogAD: Cognitive-Hierarchy Guided End-to-End Autonomous Driving

May 27, 2025

To address the fundamental misalignment between perception-planning decoupling in end-to-end autonomous driving and human cognitive principles, this paper proposes a human-inspired cognitive hierarchy. Our method features a global-local dual-level perception model for contextual awareness and an intention-conditioned multi-stage planning framework enabling intention-driven multimodal trajectory generation. We innovatively introduce dual uncertainty modeling—jointly quantifying perceptual confidence and planning intention entropy—to significantly enhance robustness in long-tail scenarios and cross-scenario generalization. Technically, the approach employs hierarchical Transformer encoding and intention-embedding-guided decoding. Evaluated on nuScenes and Bench2Drive, it achieves state-of-the-art end-to-end planning performance, outperforming existing methods notably under complex real-world traffic conditions.

0 citationsRead paper

Quality over Quantity: Boosting Data Efficiency Through Ensembled Multimodal Data Curation

Feb 12, 2025

Web-scraped datasets commonly suffer from low quality, high redundancy, and class imbalance, rendering existing heuristic filtering methods inadequate for modeling complex multimodal features—often introducing bias or erroneously discarding relevant samples. To address this, we propose EcoDatum, the first quality-centric multimodal collaborative filtering framework. Its core innovations are: (1) a quality-guided multimodal deduplication mechanism that jointly leverages visual, linguistic, and cross-modal embeddings for fine-grained similarity assessment; and (2) a weakly supervised ensemble optimization framework integrating automated hyperparameter search with multi-operator collaborative scoring. Evaluated on the DataComp benchmark, EcoDatum achieves a mean score of 0.182—outperforming prior baselines by 28%—and ranks first overall. Empirical results demonstrate substantial improvements in downstream model training efficiency and generalization performance.

0 citationsRead paper

Recent publications

Latest Papers

Distilling Future Temporal Knowledge with Masked Feature Reconstruction for 3D Object Detection

Dec 09, 2025

To address the challenge of incorporating future-frame information in online 3D object detection, this paper proposes a Future Temporal Knowledge Distillation (FTKD) framework that relaxes the strict frame-wise alignment constraint inherent in conventional knowledge distillation. Methodologically, FTKD introduces a sparse query mechanism and a future-aware feature reconstruction strategy, jointly optimized with foreground-background contextual modeling to efficiently extract and transfer future temporal knowledge from an offline teacher model to an online student. Additionally, future-guided logit distillation is incorporated to enhance the student’s modeling capability for motion dynamics and temporal consistency. Evaluated on the nuScenes dataset, FTKD achieves consistent improvements of +1.3 mAP and +1.3 NDS over strong baselines, with notable gains in velocity estimation accuracy—while incurring no additional computational overhead during online inference.

0 citationsRead paper

HMVLM: Multistage Reasoning-Enhanced Vision-Language Model for Long-Tailed Driving Scenarios

Jun 06, 2025

To address the trade-off between generalization and real-time performance in long-tail driving scenarios, this paper proposes a dual-branch end-to-end autonomous driving framework: a “fast” branch for low-latency control output and a “slow” branch leveraging multi-view images and 4-second ego-motion history to invoke a vision-language model (VLM) for selective five-view prompting and multi-stage chain-of-reasoning—comprising scene understanding, decision-making, and trajectory inference—to generate high-level driving intent. A spline-based trajectory post-processing module further enhances action smoothness. The method is supervised on the Waymo Open Dataset and achieved second place in the 2025 Waymo Vision-Only End-to-End Challenge, attaining a Rater Feedback Score of 7.7367—2.77% higher than the baseline—demonstrating significantly improved generalization to long-tail scenarios and enhanced decision rationality.

0 citationsRead paper

HMAD: Advancing E2E Driving with Anchored Offset Proposals and Simulation-Supervised Multi-target Scoring

May 29, 2025

End-to-end autonomous driving faces two key challenges: insufficient trajectory diversity and weak safety evaluation. To address these, we propose a generation-evaluation decoupled framework. First, we generate a large set of compliant and stable candidate trajectories via iterative offset decoding, leveraging BEVFormer and learnable anchor queries. Second, we introduce a simulation-driven multi-objective scoring module that jointly optimizes fault-free collision rate, drivable-area coverage, ride comfort, and an extended Planning Decision Metric (PDM). Our method pioneers an anchor-based offset trajectory proposal mechanism and a multi-task scorer network. Evaluated on the CVPR 2025 private test set, it achieves a 44.5% driving score—significantly outperforming baseline methods. This paradigm synergistically enhances both trajectory generation quality and evaluation robustness, establishing a novel, interpretable, and verifiable approach to safe path selection for end-to-end driving decision-making.

0 citationsRead paper

CogAD: Cognitive-Hierarchy Guided End-to-End Autonomous Driving

May 27, 2025

To address the fundamental misalignment between perception-planning decoupling in end-to-end autonomous driving and human cognitive principles, this paper proposes a human-inspired cognitive hierarchy. Our method features a global-local dual-level perception model for contextual awareness and an intention-conditioned multi-stage planning framework enabling intention-driven multimodal trajectory generation. We innovatively introduce dual uncertainty modeling—jointly quantifying perceptual confidence and planning intention entropy—to significantly enhance robustness in long-tail scenarios and cross-scenario generalization. Technically, the approach employs hierarchical Transformer encoding and intention-embedding-guided decoding. Evaluated on nuScenes and Bench2Drive, it achieves state-of-the-art end-to-end planning performance, outperforming existing methods notably under complex real-world traffic conditions.

0 citationsRead paper

Quality over Quantity: Boosting Data Efficiency Through Ensembled Multimodal Data Curation

Feb 12, 2025

Web-scraped datasets commonly suffer from low quality, high redundancy, and class imbalance, rendering existing heuristic filtering methods inadequate for modeling complex multimodal features—often introducing bias or erroneously discarding relevant samples. To address this, we propose EcoDatum, the first quality-centric multimodal collaborative filtering framework. Its core innovations are: (1) a quality-guided multimodal deduplication mechanism that jointly leverages visual, linguistic, and cross-modal embeddings for fine-grained similarity assessment; and (2) a weakly supervised ensemble optimization framework integrating automated hyperparameter search with multi-operator collaborative scoring. Evaluated on the DataComp benchmark, EcoDatum achieves a mean score of 0.182—outperforming prior baselines by 28%—and ranks first overall. Empirical results demonstrate substantial improvements in downstream model training efficiency and generalization performance.

0 citationsRead paper