Papers by Yunpu Ma
Self-Evolving Multi-Agent Systems via Textual Backpropagation (2026.findings-acl)
Copied to clipboard
Xiaowen Ma, Yunpu Ma, Chenyang Lin, Sikuan Yan, Jinhe Bi, Zixuan Cao, Yijun Tian, Volker Tresp, Hinrich Schuetze
| Challenge: | Large Language Models (LLMs) have proven effective for addressing complex, high-dimensional tasks, but current approaches rely on static, manually engineered multi-agent configurations. |
| Approach: | They propose a framework that conceptualizes multi-agent collaboration as a layered neural network architecture. |
| Outcome: | The proposed framework surpasses leading multi-agent baselines under the same configurations, showing consistent performance improvements. |
Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning (2026.acl-long)
Copied to clipboard
Sikuan Yan, Xiufeng Yang, Zuchao Huang, Ercong Nie, Zifeng Ding, Zonggen Li, Xiaowen Ma, Jinhe Bi, Kristian Kersting, Jeff Z. Pan, Hinrich Schuetze, Volker Tresp, Yunpu Ma
| Challenge: | Large Language Models (LLMs) are stateless and limited by a finite context window, preventing them from maintaining knowledge across long conversations or evolving tasks. |
| Approach: | They propose a reinforcement learning framework that empowers LLMs to actively manage external memory through two specialized agents. |
| Outcome: | The proposed framework outperforms baselines and benchmarks across diverse question types, three benchmarks, and multiple model scales. |
LLaVA Steering: Visual Instruction Tuning with 500x Fewer Parameters through Modality Linear Representation-Steering (2025.acl-long)
Copied to clipboard
| Challenge: | Multimodal Large Language Models (MLLMs) enhance visual tasks by integrating visual representations into large language models. |
| Approach: | They propose a method to re-balance modalities by steering visual representations . they propose LLaVA Steering, a platform that enables rapid customization of MLLMs a component-based architecture . |
| Outcome: | The proposed model re-balances the modalities of visual representations in large language models . the model requires 500 times fewer trainable parameters than LoRA while maintaining comparable performance . |
MINED: Probing and Updating with Multimodal Time-Sensitive Knowledge for Large Multimodal Models (2026.findings-acl)
Copied to clipboard
Kailin Jiang, Ning Jiang, Yuntao Du, Yuchen Ren, Yuchen Li, Yifan Gao, Jinhe Bi, Yunpu Ma, Bin Li, Lei Liu, Qing Li
| Challenge: | Existing benchmarks for Large Multimodal Models (LMMs) are constrained by static representations, inadequately evaluating their ability to understand time-sensitive knowledge. |
| Approach: | They propose a benchmark containing 2,104 time-sensitive knowledge samples spanning six knowledge types to evaluate temporal awareness along 6 key dimensions and 11 challenging tasks. |
| Outcome: | The proposed benchmark measures temporal awareness along 6 key dimensions and 11 tasks, while most open-source LMMs still lack time understanding ability. |
zrLLM: Zero-Shot Relational Learning on Temporal Knowledge Graphs with Large Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Existing methods to forecast links on temporal knowledge graphs are embedding-based . but they face a strong challenge in modeling the unseen zero-shot relations . |
| Approach: | They propose to embed knowledge graphs (TKGF) entities and relations based on observed contexts into embedding-based methods to model unseen zero-shot relations. |
| Outcome: | The proposed methods show strong performance on traditional TKG forecasting benchmarks, but they face a strong challenge in modeling unseen zero-shot relations that have no prior graph context. |
Parameter-Efficient Routed Fine-Tuning: Mixture-of-Experts Demands Mixture of Adaptation Modules (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing Parameter-Efficient Fine-Tuning (PEFT) strategies that focus on specialized experts are not effective for Mixture-of-Experts (MoE). |
| Approach: | They propose to integrate a dynamic routing mechanism among specialized experts in Mixture-of-Experts (MoE) . |
| Outcome: | Extensive experiments on commonsense and math reasoning tasks validate the performance and efficiency of the proposed routed approach. |
GenTKG: Generative Forecasting on Temporal Knowledge Graph with Large Language Models (2024.findings-naacl)
Copied to clipboard
| Challenge: | Existing methods for temporal relational forecasting are limited and require limited training data. |
| Approach: | They propose a retrieval-augmented generation framework that uses temporal logical rule-based retrieval and parameter-efficient instruction tuning to solve temporal knowledge forecasting challenges. |
| Outcome: | The proposed framework outperforms conventional methods in the temporal knowledge graph domain with low computation resources. |
SwarmAgentic: Towards Fully Automated Agentic System Generation via Swarm Intelligence (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing agentic system generation frameworks lack autonomy, autonomy, and functionality . current frameworks are too rigid, limiting adaptability and scalability. |
| Approach: | They propose a framework that fully automates agentic system generation, optimization, and collaboration . they construct agents from scratch and jointly refine functionality and coordination . |
| Outcome: | The proposed framework outperforms ADAS on six real-world, open-ended, and exploratory tasks on the TravelPlanner benchmark. |
DyERNIE: Dynamic Evolution of Riemannian Manifold Embeddings for Temporal Knowledge Graph Completion (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing embedding approaches for temporal knowledge graphs typically learn entity representations and their dynamic evolution in the Euclidean space. |
| Approach: | They propose a non-Euclidean embedding approach that learns evolving entity representations in a product of Riemannian manifolds. |
| Outcome: | The proposed model improves on three real-world datasets showing that the embeddings on Riemannian manifolds can capture the evolution of temporal KGs. |
KM-BART: Knowledge Enhanced Multimodal BART for Visual Commonsense Generation (2021.acl-long)
Copied to clipboard
| Challenge: | Existing models for visual and language understanding are not capable of multimodal reasoning. |
| Approach: | They propose a Transformer-based sequence-to-sequence model capable of reasoning about commonsense knowledge from multimodal inputs of images and texts. |
| Outcome: | The proposed model performs state-of-the-art on the Visual Commonsense Generation task. |
Mem2Evolve: Towards Self-Evolving Agents via Co-Evolutionary Capability Expansion and Experience Distillation (2026.acl-long)
Copied to clipboard
Zihao Cheng, Zeming Liu, Yingyu Shan, Xinyi Wang, Xiangrong Zhu, Yunpu Ma, Hongru Wang, Yuhang Guo, Wei Lin, Yunhong Wang
| Challenge: | Existing frameworks that focus on static tools and static assets are ineffective for self-evolving agents. |
| Approach: | They propose a paradigm of co-evolutionary Capability Expansion and Experience Distillation that leverages accumulated experience to guide dynamic creation of assets. |
| Outcome: | The proposed framework improves performance in single-task and cross-task settings by 18.53% over standard LLMs, 11.80% over agents evolving solely through experience, and 6.46% over those evolving solelly through asset creation. |
ImpliRet: Benchmarking the Implicit Fact Retrieval Challenge (2025.emnlp-main)
Copied to clipboard
| Challenge: | Retrieval systems rely on surface-level cues such as keyword overlap and semantic similarity to evaluate retrieval beyond these shallow signals. |
| Approach: | They propose a benchmark that shifts the reasoning challenge to query-side processing techniques that can help resolve complexity. |
| Outcome: | The proposed benchmarks show that document-side reasoning remains a challenge. |
VideoINSTA: Zero-shot Long Video Understanding via Informative Spatial-Temporal Reasoning with LLMs (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Long video understanding presents unique challenges due to the complexity of reasoning over extended timespans. |
| Approach: | They propose a framework VideoINSTA to leverage large language models for video understanding . they propose 'event-based temporalreasoning' and 'content-based spatial reasoning' |
| Outcome: | The proposed model significantly improves state-of-the-art on three long video question-answering benchmarks. |
METok: Multi-Stage Event-based Token Compression for Efficient Long Video Understanding (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in Video Large Language Models (VLLMs) have significantly enhanced their ability to understand video content. |
| Approach: | They propose a training-free, Multi-stage Event-based Token compression framework that eliminates redundant visual tokens across three critical stages . |
| Outcome: | The proposed framework reduces FLOPs and KV Cache memory consumption while maintaining comparable or even superior accuracy. |
Learning Neural Ordinary Equations for Forecasting Future Links on Temporal Knowledge Graphs (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing models for temporal knowledge graphs model the temporal KGs in discrete state spaces, whereas static models model the KG in discretized state spaces. |
| Approach: | They propose a continuum model that extends the idea of neural ordinary differential equations to multi-relational graph convolutional networks and encodes both temporal and structural information into continuous-time dynamic embeddings. |
| Outcome: | The proposed model outperforms existing models on five benchmark datasets showing it can predict future links on temporal knowledge graphs. |
TimeTraveler: Reinforcement Learning for Temporal Knowledge Graph Forecasting (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods focus on reasoning at past timestamps to complete the missing facts, and there are only a few works of reasoning on known TKGs to forecast future facts. |
| Approach: | They propose a time-shaped reward method that captures historical knowledge graph snapshots and a new representation method for unseen entities to improve the inductive inference ability of the model. |
| Outcome: | The proposed method improves on four benchmark datasets with higher explainability, less calculation, and fewer parameters when compared with existing state-of-the-art methods. |
Improving LLM Reasoning through Interpretable Role-Playing Steering (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for role-playing rely on prompt engineering, which lacks stability and interpretability. |
| Approach: | They propose a framework that extracts latent representations from role-play prompts and constructs a steering vector that can be injected into the model's residual stream with controllable intensity. |
| Outcome: | The proposed framework extracts latent representations from role-play prompts, selects the most relevant features based on activation patterns, and constructs a steering vector that can be injected into the model’s residual stream with controllable intensity. |
Language Mixing in Reasoning Language Models: Patterns, Impact, and Internal Causes (2025.emnlp-main)
Copied to clipboard
| Challenge: | Reasoning language models (RLMs) excel at complex tasks by leveraging a chain-of-thought process to generate structured intermediate steps. |
| Approach: | They present the first systematic study of language mixing in reasoning language models, examining its patterns, impact, and internal causes across 15 languages, 7 task difficulty levels, and 18 subject areas. |
| Outcome: | The proposed model generates reasoning steps that include a mixture of languages when prompted in one language, and this improves accuracy. |
Time-dependent Entity Embedding is not All You Need: A Re-evaluation of Temporal Knowledge Graph Completion Models under a Unified Framework (2021.emnlp-main)
Copied to clipboard
| Challenge: | Various temporal knowledge graph (KG) completion models have been proposed . knowledge graphs are typically static and store facts in their current state . |
| Approach: | They propose to use temporal embeddings and a score function to model temporal knowledge graphs . they classify the temporal embedded methods into two classes: timestamp and time-dependent . |
| Outcome: | The proposed models outperform current models on ICEWS datasets with 3000 experiments and 13159 GPU hours. |