Published in Meta Superintelligence Labs Technical Report, 2026
A 30B multimodal model for local autonomous agents, combining multi-step reasoning, tool use, and failure recovery. I am listed as a core contributor in Section 11.1.
Recommended citation: Muse Glimmer team. 2026. "Muse Glimmer 30B: An On-Device Agentic Model." Meta Superintelligence Labs. Technical report.
Download Paper
Published in ICML 2026
COVER accelerates revocable diffusion decoding while preserving context and generation quality.
Recommended citation: Yanzheng Xiang, Lan Wei, Yizhen Yao, Qinglin Zhu, Hanqi Yan, Chen Jin, Philip Alexander Teare, Dandan Zhang, Lin Gui, Amrutha Saseendran, Yulan He. 2026. "Stop the Flip-Flop: Context-Preserving Verification for Fast Revocable Diffusion Decoding." In ICML 2026.
Download Paper
Published in COLM 2025
Benchmarking agentic LLM systems on reproducing algorithms described in research papers.
Recommended citation: Yanzheng Xiang, Hanqi Yan, Shuyin Ouyang, Lin Gui, Yulan He. 2025. "SciReplicate-Bench: Benchmarking LLMs in Agent-driven Algorithmic Reproduction from Research Papers." In COLM 2025.
Download Paper
Published in Findings of ACL 2023
A graph-guided generate-and-rerank framework that advances cross-domain text-to-SQL generation.
Recommended citation: Yanzheng Xiang, Qian-Wen Zhang, Xu Zhang, Zejie Liu, Yunbo Cao, Deyu Zhou. 2023. "G3R: A Graph-Guided Generate-and-Rerank Framework for Cross-domain Text-to-SQL Generation." In Findings of ACL 2023.
Download Paper
Published in Findings of ACL 2024
Mitigating sensitivity to demonstration order in in-context learning with causal language models.
Recommended citation: Yanzheng Xiang, Hanqi Yan, Lin Gui, Yulan He. 2024. "Addressing Order Sensitivity of In-Context Demonstration Examples in Causal Language Models." In Findings of ACL 2024.
Download Paper
Published in NeurIPS 2026
A decoding framework that refines token beliefs and feeds predictions back to improve diffusion language model generation.
Recommended citation: Qinglin Zhu, Yizhen Yao, Runcong Zhao, Yanzheng Xiang, Amrutha Saseendran, Chen Jin, Philip Teare, Bin Liang, Yulan He, Lin Gui. 2026. "Latent Refinement Decoding: Enhancing Diffusion-Based Language Models by Refining Belief States." In NeurIPS 2026.
Download Paper
Published in EMNLP 2026 (Oral)
Anchor-supervised revocable decoding uses reliable tokens to guide generation and verify uncertain tokens in diffusion LLMs.
Recommended citation: Yizhen Yao, Qinglin Zhu, Runcong Zhao, Xiangxiang Dai, Yanzheng Xiang, Yulan He, Lin Gui. 2026. "Follow the Latent Roadmap: Navigating Revocable Decoding for Diffusion LLMs with Anchor Tokens." In EMNLP 2026 (Oral).
Download Paper
Published in EMNLP 2024
Analyzing how feature decorrelation influences monosemanticity in language models.
Recommended citation: Hanqi Yan, Yanzheng Xiang, Guangyi Chen, Yifei Wang, Lin Gui, Yulan He. 2024. "Encourage or Inhibit Monosemanticity? Revisit Monosemanticity from a Feature Decorrelation Perspective." In EMNLP 2024.
Download Paper
Published in EMNLP 2024
Surveying interpretability and analysis techniques for understanding in-context learning.
Recommended citation: Yuxiang Zhou, Jiazheng Li, Yanzheng Xiang, Hanqi Yan, Lin Gui, Yulan He. 2024. "The Mystery of In-Context Learning: A Comprehensive Survey on Interpretation and Analysis." In EMNLP 2024.
Download Paper
Published in Findings of ACL 2025
Dynamic progress control with hierarchical graph reasoning for long-document question answering.
Recommended citation: Xinyu Wang, Yanzheng Xiang, Lin Gui, Yulan He. 2025. "PECAN: LLM-Guided Dynamic Progress Control with Attention-Guided Hierarchical Weighted Graph for Long-Document QA." In Findings of ACL 2025.
Download Paper
Published in Findings of EMNLP 2021
Divide-and-conquer strategy for multi-label, multi-hop relation detection in knowledge-base question answering.
Recommended citation: Deyu Zhou, Yanzheng Xiang, Linhai Zhang, Chenchen Ye, Qian-Wen Zhang, Yunbo Cao. 2021. "A Divide-And-Conquer Approach for Multi-label Multi-hop Relation Detection in Knowledge Base QA." In Findings of EMNLP 2021.
Download Paper