Lune

Profiles2020-2026

Xu Sun

Peking University

55Papers
40Last author
80Unique co-authors
1First author
0Solo

First and last author come from byline position, which some venues assign alphabetically.

Recent papers55

  1. Conan: Progressive Learning to Reason Like a Detective over Multi-Scale Visual EvidenceAuthor 8 of 8CVPR 2026 · 17 citations
  2. TEMPLE: Incentivizing Temporal Understanding of Video Large Language Models via Progressive Pre-SFT AlignmentAuthor 10 of 10AAAI 2026 · 2 citations
  3. From Player to Master: Enhancing Test-Time Learning of LLM Agents via Reinforcement Learning over MemoryAuthor 11 of 11ICML 2026 · 1 citation
  4. Investigating Cross-Modal Skill Injection: Scenarios, Methods, and HyperparametersAuthor 8 of 8ACL 2026
  5. ReaForest: Fostering Generative Video Reasoning for Spatial PlanningAuthor 10 of 10ICML 2026
  6. TimeChat-Captioner: Scripting Multi-Scene Videos with Time-Aware and Structural Audio-Visual CaptionsAuthor 15 of 15ICML 2026
  7. Generative Frame Sampler for Long Video UnderstandingAuthor 7 of 8ACL Findings 2025 · 3 citations
  8. PunchBench: Benchmarking MLLMs in Multimodal Punchline ComprehensionAuthor 8 of 8ACL 2025 · 3 citations
  9. RICO: Improving Accuracy and Completeness in Image Recaptioning via Visual ReconstructionAuthor 9 of 9EMNLP 2025 · 1 citation
  10. ATLANTIS: Weak-to-Strong Learning via Importance SamplingAuthor 5 of 5ACL 2025
  11. Temporal Reasoning Transfer from Text to VideoAuthor 7 of 9ICLR 2025
  12. VidTwin: Video VAE with Decoupled Structure and DynamicsAuthor 5 of 6CVPR 2025
  13. Watch Out for Your Agents! Investigating Backdoor Threats to LLM-Based AgentsAuthor 6 of 6NeurIPS 2024 · 195 citations
  14. Towards Codable Watermarking for Injecting Multi-Bits Information to LLMsAuthor 8 of 8ICLR 2024 · 55 citations
  15. Edit As You Wish: Video Caption Editing with Multi-grained User ControlAuthor 7 of 8ACM MM 2024 · 4 citations
  16. Enhancing Byzantine-Resistant Aggregations with Client EmbeddingAuthor 5 of 5EMNLP Findings 2024 · 2 citations
  17. Label Words are Anchors: An Information Flow Perspective for Understanding In-Context LearningAuthor 8 of 8EMNLP 2023 · 26 citations
  18. Communication Efficient Federated Learning for Multilingual Neural Machine Translation with AdapterAuthor 6 of 6ACL Findings 2023 · 13 citations
  19. Delving into the Openness of CLIPAuthor 5 of 5ACL Findings 2023 · 10 citations
  20. Fed-FA: Theoretically Modeling Client Data Divergence for Federated Language Backdoor DefenseAuthor 6 of 6NeurIPS 2023 · 10 citations
  21. Diffusion Theory as a Scalpel: Detecting and Purifying Poisonous Dimensions in Pre-trained Language Models Caused by Backdoor or BiasAuthor 6 of 6ACL Findings 2023 · 4 citations
  22. How to Inject Backdoors with Better Consistency: Logit Anchoring on Clean DataAuthor 5 of 5ICLR 2022 · 41 citations
  23. Retrieve, Reason, and Refine: Generating Accurate and Faithful Patient InstructionsAuthor 7 of 9NeurIPS 2022 · 34 citations
  24. Dim-Krum: Backdoor-Resistant Federated Learning for NLP with Dimension-wise Krum-Based AggregationAuthor 3 of 3EMNLP Findings 2022 · 16 citations
  25. Fine-mixing: Mitigating Backdoors in Fine-tuned Language ModelsAuthor 5 of 5EMNLP Findings 2022 · 16 citations
  26. Well-Classified Examples Are Underestimated in Classification with Deep Neural NetworksAuthor 6 of 6AAAI 2022 · 16 citations
  27. Expose Backdoors on the Way: A Feature-Based Efficient Defense against Textual Backdoor AttacksAuthor 5 of 5EMNLP Findings 2022 · 15 citations
  28. GA-SAM: Gradient-Strength based Adaptive Sharpness-Aware Minimization for Improved GeneralizationAuthor 4 of 4EMNLP 2022 · 9 citations
  29. Holistic Sentence Embeddings for Better Out-of-Distribution DetectionAuthor 4 of 4EMNLP Findings 2022 · 9 citations
  30. Hierarchical Inductive Transfer for Continual Dialogue LearningAuthor 4 of 4ACL Findings 2022 · 2 citations

Showing the 30 most recent indexed papers.

Dusk over the sea between two cliffs drawn in fine vertical lines