arXiv subjects

Daifeng Li

Publications and source records attributed to Daifeng Li.

16 recordsLinked to original sources

arXiv2026

CompEvo: Competition-Induced Evolution for Multi-Agent in News-Driven Time Series Forecasting

News-driven time series forecasting uses evolving textual events together with historical observations to predict future values, supporting applications such as market risk monitoring and resource scheduling. In multi-agent settings, two challenges still remain. The first is degeneration of thought, where agents converge to similar evidence-seeking behaviors. The second is insufficient theoretical grounding, where strategy updates are often heuristic and lack a principled formulation. To address the above challenges, we propose CompEvo, a competition-induced evolution framework for multi-agent news-driven time series forecasting. For theoretical grounding, we introduce an evolutionary game formulation to guarantee equilibrium existence and optimization convergence. Building on this formulation, we construct a trainable multi-agent evolution framework that integrates strategy execution, fitness-based differentiable selection, and competition-induced strategy evolution. CompEvo enables heterogeneous agents to explore diverse news evidence, converts forecasting feedback into differentiable influence weights, and evolves agent strategies under competitive pressure to preserve effective logic while maintaining diversity. Experiments on four real-world datasets show that CompEvo reduces RMSE by 27.3% and MAPE by 26.2% on average over strong baselines. Further analysis indicates that CompEvo successfully maintains diverse and specialized agent behaviors.

cs.NE↗

arXiv2026

ARGUS: Agentic GPU Optimization Guided by Data-Flow Invariants

LLM-based coding agents can generate functionally correct GPU kernels, yet their performance remains far below hand-optimized libraries on critical computations such as matrix multiplication, attention, and Mixture-of-Experts (MoE). Peak GPU performance requires coordinated reasoning over tightly coupled optimizations, including tiling, shared-memory staging, software pipelining, and instruction scheduling, while existing agents rely on sparse pass/fail feedback, leaving them unable to diagnose global constraint violations. We present Argus, an agentic framework that addresses this through data-flow invariants: compile-time specifications encoding how data must be choreographed throughout kernel execution. Argus introduces a tile-based, Pythonic DSL exposing hardware instructions and compiler policies while hiding low-level representations. The DSL provides tag functions to propagate symbolic annotations through data and control flow, and tag assertions to enforce relational constraints at use sites. When violations occur, the compiler returns concrete counterexamples identifying the thread, data element, and program point, enabling dense, structured feedback for targeted fixes. Invariants are verified at compile time via abstract interpretation over a layout algebra and SMT solving, with zero runtime overhead. An in-context reinforcement learning planner learns to select optimizations and synthesize effective invariants, supported by a curated knowledge base of GPU optimization techniques. We evaluate Argus on the AMD MI300X GPU across GEMM, flash attention, and MoE kernels accounting for over 90% of GPU time in LLM inference. Generated kernels achieve 99-104% of state-of-the-art hand-optimized assembly throughput and are 2-1543x faster than existing agentic systems. Argus further generalizes to 200 KernelBench tasks, solving 100% of Level 1 and 90% of Level 2 problems.

cs.DC↗

arXiv2026

AREAL-DTA: Dynamic Tree Attention for Efficient Reinforcement Learning of Large Language Models

Reinforcement learning (RL)-based post-training for large language models (LLMs) is computationally expensive, as it generates many rollout sequences that frequently share long token prefixes. Existing RL frameworks usually process these sequences independently during policy training, i.e., repeatedly recomputing identical prefixes in both the forward and backward passes of policy gradient computation, leading to substantial inefficiencies in computation resources and memory usage. Although prefix sharing naturally induces a tree structure over rollouts, packed tree-mask approaches scale poorly in RL settings. In this paper, we introduce AReaL-DTA, which efficiently exploits prefix sharing in RL training. AReaL-DTA employs a depth-first search (DFS)-based execution strategy that dynamically traverses the rollout prefix tree during both forward and backward computation, materializing only a single root-to-leaf path at a time. To further improve scalability, AReaL-DTA incorporates a load-balanced distributed batching mechanism that dynamically constructs and processes prefix trees across multiple GPUs. On $\tau^2$-bench, AReaL-DTA improves training throughput by up to $8.31\times$ over dense training and up to $1.70\times$ over sparse training. Our code is available at https://github.com/areal-project/AReaL/tree/feat/dta.

cs.LG↗

arXiv2025

Two-way Evidence self-Alignment based Dual-Gated Reasoning Enhancement

Large language models (LLMs) encounter difficulties in knowledge-intensive multi-step reasoning (KIMSR) tasks. One challenge is how to effectively extract and represent rationale evidence. The current methods often extract semantically relevant but logically irrelevant evidence, resulting in flawed reasoning and inaccurate responses. We propose a two-way evidence self-alignment (TW-ESA) module, which utilizes the mutual alignment between strict reasoning and LLM reasoning to enhance its understanding of the causal logic of evidence, thereby addressing the first challenge. Another challenge is how to utilize the rationale evidence and LLM's intrinsic knowledge for accurate reasoning when the evidence contains uncertainty. We propose a dual-gated reasoning enhancement (DGR) module to gradually fuse useful knowledge of LLM within strict reasoning, which can enable the model to perform accurate reasoning by focusing on causal elements in the evidence and exhibit greater robustness. The two modules are collaboratively trained in a unified framework ESA-DGR. Extensive experiments on three diverse and challenging KIMSR datasets reveal that ESA-DGR significantly surpasses state-of-the-art LLM-based fine-tuning methods, with remarkable average improvements of 4% in exact match (EM) and 5% in F1 score. The implementation code is available at https://anonymous.4open.science/r/ESA-DGR-2BF8.

cs.CL↗

arXiv2025

Can Competition Enhance the Proficiency of Agents Powered by Large Language Models in the Realm of News-driven Time Series Forecasting?

Multi-agents-based news-driven time series forecasting is considered as a potential paradigm shift in the era of large language models (LLMs). The challenge of this task lies in measuring the influences of different news events towards the fluctuations of time series. This requires agents to possess stronger abilities of innovative thinking and the identifying misleading logic. However, the existing multi-agent discussion framework has limited enhancement on time series prediction in terms of optimizing these two capabilities. Inspired by the role of competition in fostering innovation, this study embeds a competition mechanism within the multi-agent discussion to enhance agents' capability of generating innovative thoughts. Furthermore, to bolster the model's proficiency in identifying misleading information, we incorporate a fine-tuned small-scale LLM model within the reflective stage, offering auxiliary decision-making support. Experimental results confirm that the competition can boost agents' capacity for innovative thinking, which can significantly improve the performances of time series prediction. Similar to the findings of social science, the intensity of competition within this framework can influence the performances of agents, providing a new perspective for studying LLMs-based multi-agent systems.

cs.AI↗

arXiv2025

Structuring Scientific Innovation: A Framework for Modeling and Discovering Impactful Knowledge Combinations

The emergence of large language models offers new possibilities for structured exploration of scientific knowledge. Rather than viewing scientific discovery as isolated ideas or content, we propose a structured approach that emphasizes the role of method combinations in shaping disruptive insights. Specifically, we investigate how knowledge unit--especially those tied to methodological design--can be modeled and recombined to yield research breakthroughs. Our proposed framework addresses two key challenges. First, we introduce a contrastive learning-based mechanism to identify distinguishing features of historically disruptive method combinations within problem-driven contexts. Second, we propose a reasoning-guided Monte Carlo search algorithm that leverages the chain-of-thought capability of LLMs to identify promising knowledge recombinations for new problem statements.Empirical studies across multiple domains show that the framework is capable of modeling the structural dynamics of innovation and successfully highlights combinations with high disruptive potential. This research provides a new path for computationally guided scientific ideation grounded in structured reasoning and historical data modeling.

cs.AI↗

arXiv2024

A Large Language Model-based Framework for Semi-Structured Tender Document Retrieval-Augmented Generation

The drafting of documents in the procurement field has progressively become more complex and diverse, driven by the need to meet legal requirements, adapt to technological advancements, and address stakeholder demands. While large language models (LLMs) show potential in document generation, most LLMs lack specialized knowledge in procurement. To address this gap, we use retrieval-augmented techniques to achieve professional document generation, ensuring accuracy and relevance in procurement documentation.

cs.CL↗

arXiv2024

Mirror Gradient: Towards Robust Multimodal Recommender Systems via Exploring Flat Local Minima

Multimodal recommender systems utilize various types of information to model user preferences and item features, helping users discover items aligned with their interests. The integration of multimodal information mitigates the inherent challenges in recommender systems, e.g., the data sparsity problem and cold-start issues. However, it simultaneously magnifies certain risks from multimodal information inputs, such as information adjustment risk and inherent noise risk. These risks pose crucial challenges to the robustness of recommendation models. In this paper, we analyze multimodal recommender systems from the novel perspective of flat local minima and propose a concise yet effective gradient strategy called Mirror Gradient (MG). This strategy can implicitly enhance the model's robustness during the optimization process, mitigating instability risks arising from multimodal information inputs. We also provide strong theoretical evidence and conduct extensive empirical experiments to show the superiority of MG across various multimodal recommendation models and benchmarks. Furthermore, we find that the proposed MG can complement existing robust training methods and be easily extended to diverse advanced recommendation models, making it a promising new and fundamental paradigm for training multimodal recommender systems. The code is released at https://github.com/Qrange-group/Mirror-Gradient.

cs.IR↗

arXiv2023

Effects of Same-Race Mentorship Preferences on Academic Performance and Survival

Same-race mentorship preference refers to mentors or mentees forming connections significantly influenced by a shared race. Although racial diversity in science has been well-studied and linked to favorable outcomes, the extent and effects of same-race mentorship preferences remain largely underexplored. Here, we analyze 465,355 mentor-mentee pairs from more than 60 research areas over the last 70 years to investigate the effect of same-race mentorship preferences on mentees' academic performance and survival. We use causal inference and statistical matching to measure same-race mentorship preferences while accounting for racial demographic variations across institutions, time periods, and research fields. Our findings reveal a pervasive same-race mentorship propensity across races, fields, and universities of varying research intensity. We observe an increase in same-race mentorship propensity over the years, further reinforced inter-generationally within a mentorship lineage. This propensity is more pronounced for minorities (Asians, Blacks, and Hispanics). Our results reveal that mentees under the supervision of mentors with high same-race propensity experience significantly lower productivity, impact, and collaboration reach during and after training, ultimately leading to a 27.6% reduced likelihood of remaining in academia. In contrast, a mentorship approach devoid of racial propensity appears to offer the best prospects for academic performance and persistence. These findings underscore the importance of mentorship diversity for academic success and shed light on factors contributing to minority underrepresentation in science.

cs.SI↗

arXiv2022

Causal Inference for Chatting Handoff

Aiming to ensure chatbot quality by predicting chatbot failure and enabling human-agent collaboration, Machine-Human Chatting Handoff (MHCH) has attracted lots of attention from both industry and academia in recent years. However, most existing methods mainly focus on the dialogue context or assist with global satisfaction prediction based on multi-task learning, which ignore the grounded relationships among the causal variables, like the user state and labor cost. These variables are significantly associated with handoff decisions, resulting in prediction bias and cost increasement. Therefore, we propose Causal-Enhance Module (CEM) by establishing the causal graph of MHCH based on these two variables, which is a simple yet effective module and can be easy to plug into the existing MHCH methods. For the impact of users, we use the user state to correct the prediction bias according to the causal relationship of multi-task. For the labor cost, we train an auxiliary cost simulator to calculate unbiased labor cost through counterfactual learning so that a model becomes cost-aware. Extensive experiments conducted on four real-world benchmarks demonstrate the effectiveness of CEM in generally improving the performance of existing MHCH methods without any elaborated model crafting.

cs.AI↗

arXiv2020

Pandemics are catalysts of scientific novelty: Evidence from COVID-19

Scientific novelty drives the efforts to invent new vaccines and solutions during the pandemic. First-time collaboration and international collaboration are two pivotal channels to expand teams' search activities for a broader scope of resources required to address the global challenge, which might facilitate the generation of novel ideas. Our analysis of 98,981 coronavirus papers suggests that scientific novelty measured by the BioBERT model that is pre-trained on 29 million PubMed articles, and first-time collaboration increased after the outbreak of COVID-19, and international collaboration witnessed a sudden decrease. During COVID-19, papers with more first-time collaboration were found to be more novel and international collaboration did not hamper novelty as it had done in the normal periods. The findings suggest the necessity of reaching out for distant resources and the importance of maintaining a collaborative scientific community beyond nationalism during a pandemic.

cs.DL↗

arXiv2020

Building a PubMed knowledge graph

PubMed is an essential resource for the medical domain, but useful concepts are either difficult to extract or are ambiguated, which has significantly hindered knowledge discovery. To address this issue, we constructed a PubMed knowledge graph (PKG) by extracting bio-entities from 29 million PubMed abstracts, disambiguating author names, integrating funding data through the National Institutes of Health (NIH) ExPORTER, collecting affiliation history and educational background of authors from ORCID, and identifying fine-grained affiliation data from MapAffil. Through the integration of the credible multi-source data, we could create connections among the bio-entities, authors, articles, affiliations, and funding. Data validation revealed that the BioBERT deep learning method of bio-entity extraction significantly outperformed the state-of-the-art models based on the F1 score (by 0.51%), with the author name disambiguation (AND) achieving a F1 score of 98.09%. PKG can trigger broader innovations, not only enabling us to measure scholarly impact, knowledge usage, and knowledge transfer, but also assisting us in profiling authors and organizations based on their connections with bio-entities. The PKG is freely available on Figshare (https://figshare.com/s/6327a55355fc2c99f3a2, simplified version that exclude PubMed raw data) and TACC website (http://er.tacc.utexas.edu/datasets/ped, full version).

cs.DL↗

arXiv2020

Attribute2vec: Deep Network Embedding Through Multi-Filtering GCN

We present a multi-filtering Graph Convolution Neural Network (GCN) framework for network embedding task. It uses multiple local GCN filters to do feature extraction in every propagation layer. We show this approach could capture different important aspects of node features against the existing attribute embedding based method. We also show that with multi-filtering GCN approach, we can achieve significant improvement against baseline methods when training data is limited. We also perform many empirical experiments and demonstrate the benefit of using multiple filters against single filter as well as most current existing network embedding methods for both the link prediction and node classification tasks.

cs.LG↗

arXiv2012

What is the Nature of Chinese MicroBlogging: Unveiling the Unique Features of Tencent Weibo

China has the largest number of online users in the world and about 20% internet users are from China. This is a huge, as well as a mysterious, market for IT industry due to various reasons such as culture difference. Twitter is the largest microblogging service in the world and Tencent Weibo is one of the largest microblogging services in China. Employ the two data sets as a source in our study, we try to unveil the unique behaviors of Chinese users. We have collected the entire Tencent Weibo from 10th, Oct, 2011 to 5th, Jan, 2012 and obtained 320 million user profiles, 5.15 billion user actions. We study Tencent Weibo from both macro and micro levels. From the macro level, Tencent users are more active on forwarding messages, but with less reciprocal relationships than Twitter users, their topic preferences are very different from Twitter users from both content and time consuming; besides, information can be diffused more efficient in Tencent Weibo. From the micro level, we mainly evaluate users' social influence from two indexes: "Forward" and \Follower", we study how users' actions will contribute to their social influences, and further identify unique features of Tencent users. According to our studies, Tencent users' actions are more personalized and diversity, and the influential users play a more important part in the whole networks. Based on the above analysis, we design a graphical model for predicting users' forwarding behaviors. Our experimental results on the large Tencent Weibo data validate the correctness of the discoveries and the effectiveness of the proposed model. To the best of our knowledge, this work is the first quantitative study on the entire Tencentsphere and information diffusion on it.

cs.SI↗

arXiv2012

User-level Weibo Recommendation incorporating Social Influence based on Semi-Supervised Algorithm

Tencent Weibo, as one of the most popular micro-blogging services in China, has attracted millions of users, producing 30-60 millions of weibo (similar as tweet in Twitter) daily. With the overload problem of user generate content, Tencent users find it is more and more hard to browse and find valuable information at the first time. In this paper, we propose a Factor Graph based weibo recommendation algorithm TSI-WR (Topic-Level Social Influence based Weibo Recommendation), which could help Tencent users to find most suitable information. The main innovation is that we consider both direct and indirect social influence from topic level based on social balance theory. The main advantages of adopting this strategy are that it could first build a more accurate description of latent relationship between two users with weak connections, which could help to solve the data sparsity problem; second provide a more accurate recommendation for a certain user from a wider range. Other meaningful contextual information is also combined into our model, which include: Users profile, Users influence, Content of weibos, Topic information of weibos and etc. We also design a semi-supervised algorithm to further reduce the influence of data sparisty. The experiments show that all the selected variables are important and the proposed model outperforms several baseline methods.

cs.SI↗

arXiv2012

Topic-Level Opinion Influence Model(TOIM): An Investigation Using Tencent Micro-Blogging

Mining user opinion from Micro-Blogging has been extensively studied on the most popular social networking sites such as Twitter and Facebook in the U.S., but few studies have been done on Micro-Blogging websites in other countries (e.g. China). In this paper, we analyze the social opinion influence on Tencent, one of the largest Micro-Blogging websites in China, endeavoring to unveil the behavior patterns of Chinese Micro-Blogging users. This paper proposes a Topic-Level Opinion Influence Model (TOIM) that simultaneously incorporates topic factor and social direct influence in a unified probabilistic framework. Based on TOIM, two topic level opinion influence propagation and aggregation algorithms are developed to consider the indirect influence: CP (Conservative Propagation) and NCP (None Conservative Propagation). Users' historical social interaction records are leveraged by TOIM to construct their progressive opinions and neighbors' opinion influence through a statistical learning process, which can be further utilized to predict users' future opinions on some specific topics. To evaluate and test this proposed model, an experiment was designed and a sub-dataset from Tencent Micro-Blogging was used. The experimental results show that TOIM outperforms baseline methods on predicting users' opinion. The applications of CP and NCP have no significant differences and could significantly improve recall and F1-measure of TOIM.

cs.SI↗