Lune

Profiles2020-2026

Xipeng Qiu

Shanghai Innovation Institute

109Papers
68Last author
146Unique co-authors
1First author
0Solo

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

Recent papers109

  1. Sparse-dLLM: Accelerating Diffusion LLMs with Dynamic Cache EvictionAuthor 8 of 8AAAI 2026 · 43 citations
  2. Thinking with Video: Video Generation as a Promising Multimodal Reasoning ParadigmAuthor 14 of 14CVPR 2026 · 37 citations
  3. XY-Tokenizer: Mitigating the Semantic-Acoustic Conflict in Low-Bitrate Speech CodecsAuthor 9 of 9ACL 2026 · 35 citations
  4. LongLLaDA: Unlocking Long Context Capabilities in Diffusion LLMsAuthor 7 of 7AAAI 2026 · 30 citations
  5. Game-RL: Synthesizing Multimodal Verifiable Game Data to Boost VLMs' General ReasoningAuthor 22 of 24ICLR 2026 · 21 citations
  6. HERMES: KV Cache as Hierarchical Memory for Efficient Streaming Video UnderstandingAuthor 5 of 5ACL 2026 · 17 citations
  7. Towards True Speech-to-Speech Models Without Text GuidanceAuthor 15 of 15ICLR 2026 · 8 citations
  8. Dimensional Collapse in Transformer Attention Outputs: A Challenge for Sparse Dictionary LearningAuthor 5 of 5ICML 2026 · 5 citations
  9. Prism: Spectral-Aware Block-Sparse AttentionAuthor 7 of 7ICML 2026 · 3 citations
  10. BandPO: Bridging Trust Regions and Ratio Clipping via Probability-Aware Bounds for LLM Reinforcement LearningAuthor 7 of 7ICML 2026 · 2 citations
  11. Beyond Real: Imaginary Extension of Rotary Position Embeddings for Long-Context LLMsAuthor 9 of 9ICLR 2026 · 2 citations
  12. Sparser Block-Sparse Attention via Token PermutationAuthor 10 of 10ICML 2026 · 2 citations
  13. A Distributional View for Visual Mechanistic Interpretability: KL-Minimal Soft-Constraint PrincipleAuthor 8 of 8ICML 2026 · 1 citation
  14. ARISE: An Adaptive Resolution-Aware Metric for Test-Time Scaling Evaluation in Large Reasoning ModelsAuthor 7 of 7ACL Findings 2026
  15. Efficient KL Divergence Estimation via Truncated Top-K Integration for Large Language ModelsAuthor 6 of 6ACL 2026
  16. Faster Than Flash: Exploiting Attention Sparsity for Efficient Long-Context DecodingAuthor 8 of 8ICML 2026
  17. FASTer: Toward Powerful and Efficient Autoregressive Vision-Language-Action Models with Learnable Action Tokenizer and Block-wise DecodingAuthor 14 of 15ICLR 2026
  18. Scaling Transformers for End-to-End Discrete Audio TokenizationAuthor 12 of 12ICML 2026
  19. Implicit Reward as the Bridge: A Unified View of SFT and DPO ConnectionsAuthor 10 of 10NeurIPS 2025 · 23 citations
  20. Dynamic and Generalizable Process Reward ModelingAuthor 5 of 6ACL 2025 · 13 citations
  21. World-aware Planning Narratives Enhance Large Vision-Language Model PlannerAuthor 6 of 6NeurIPS 2025 · 11 citations
  22. R3-RAG: Learning Step-by-Step Reasoning and Retrieval for LLMs via Reinforcement LearningAuthor 10 of 10EMNLP Findings 2025 · 4 citations
  23. ConvSearch-R1: Enhancing Query Reformulation for Conversational Search with Reasoning via Reinforcement LearningAuthor 5 of 5EMNLP 2025 · 2 citations
  24. BitStack: Any-Size Compression of Large Language Models in Variable Memory EnvironmentsAuthor 6 of 6ICLR 2025
  25. Data Mixing Laws: Optimizing Data Mixtures by Predicting Language Modeling PerformanceAuthor 6 of 6ICLR 2025
  26. Decoupled Proxy Alignment: Mitigating Language Prior Conflict for Multimodal Alignment in MLLMsAuthor 9 of 9EMNLP Findings 2025
  27. Firewall Routing: Blocking Leads to Better Hybrid Inference for LLMsAuthor 6 of 6EMNLP 2025
  28. MCM-DPO: Multifaceted Cross-Modal Direct Preference Optimization for Alt-text GenerationAuthor 6 of 7ACM MM 2025
  29. ReAttention: Training-Free Infinite Context with Finite Attention ScopeAuthor 10 of 10ICLR 2025
  30. Revisiting the Test-Time Scaling of o1-like Models: Do they Truly Possess Test-Time Scaling Capabilities?Author 5 of 5ACL 2025

Showing the 30 most recent indexed papers.

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