Wangtao Sun

Wangtao Sun, Xuanqing Yu, Shizhu He et al.

Black-box Large Language Models (LLMs) have shown great power in solving various tasks and are considered general problem solvers. However, LLMs still fail in many specific tasks although understand the task instruction. In this paper, we focus on the problem of boosting the ability of black-box LLMs to solve downstream tasks. We propose ExpNote, an automated framework to help LLMs better adapt to unfamiliar tasks through reflecting and noting experiences from training data and retrieving them from external memory during testing. We evaluate ExpNote on multiple tasks and the experimental results demonstrate that the proposed method significantly improves the performance of black-box LLMs. The data and code are available at https://github.com/forangel2014/ExpNote

Wangtao Sun, Chenxiang Zhang, XueYou Zhang et al.

Although Large Language Models (LLMs) have demonstrated strong ability, they are further supposed to be controlled and guided by in real-world scenarios to be safe, accurate, and intelligent. This demands the possession of capability of LLMs. However, no prior work has made a clear evaluation of the inferential rule-following capability of LLMs. Previous studies that try to evaluate the inferential rule-following capability of LLMs fail to distinguish the inferential rule-following scenarios from the instruction-following scenarios. Therefore, this paper first clarifies the concept of inferential rule-following and proposes a comprehensive benchmark, RuleBench, to evaluate a diversified range of inferential rule-following abilities. Our experimental results on a variety of LLMs show that they are still limited in following rules. Our analysis based on the evaluation results provides insights into the improvements for LLMs toward a better inferential rule-following intelligent agent. We further propose Inferential Rule-Following Tuning (IRFT). The experimental results show that through IRFT, LLMs can learn abstract rule-following abilities from purely synthetic data and then generalize to RuleBench. The data and code can be found at: https://anonymous.4open.science/r/llm-rule-following-B3E3/

Xuanqing Yu, Wangtao Sun, Jingwei Li et al.

In the realm of event prediction, temporal knowledge graph forecasting (TKGF) stands as a pivotal technique. Previous approaches face the challenges of not utilizing experience during testing and relying on a single short-term history, which limits adaptation to evolving data. In this paper, we introduce the Online Neural-Symbolic Event Prediction (ONSEP) framework, which innovates by integrating dynamic causal rule mining (DCRM) and dual history augmented generation (DHAG). DCRM dynamically constructs causal rules from real-time data, allowing for swift adaptation to new causal relationships. In parallel, DHAG merges short-term and long-term historical contexts, leveraging a bi-branch approach to enrich event prediction. Our framework demonstrates notable performance enhancements across diverse datasets, with significant Hit@k (k=1,3,10) improvements, showcasing its ability to augment large language models (LLMs) for event prediction without necessitating extensive retraining. The ONSEP framework not only advances the field of TKGF but also underscores the potential of neural-symbolic approaches in adapting to dynamic data environments.

Wangtao Sun, Xiang Cheng, Jialin Fan et al.

We study whether self-learning can scale LLM-based agents without relying on human-curated datasets or predefined rule-based rewards. Through controlled experiments in a search-agent setting, we identify two key determinants of scalable agent training: the source of reward signals and the scale of agent task data. We find that rewards from a Generative Reward Model (GRM) outperform rigid rule-based signals for open-domain learning, and that co-evolving the GRM with the policy further boosts performance. Increasing the volume of agent task data-even when synthetically generated-substantially enhances agentic capabilities. Building on these insights, we propose \textbf{Agentic Self-Learning} (ASL), a fully closed-loop, multi-role reinforcement learning framework that unifies task generation, policy execution, and evaluation within a shared tool environment and LLM backbone. ASL coordinates a Prompt Generator, a Policy Model, and a Generative Reward Model to form a virtuous cycle of harder task setting, sharper verification, and stronger solving. Empirically, ASL delivers steady, round-over-round gains, surpasses strong RLVR baselines (e.g., Search-R1) that plateau or degrade, and continues improving under zero-labeled-data conditions, indicating superior sample efficiency and robustness. We further show that GRM verification capacity is the main bottleneck: if frozen, it induces reward hacking and stalls progress; continual GRM training on the evolving data distribution mitigates this, and a small late-stage injection of real verification data raises the performance ceiling. This work establishes reward source and data scale as critical levers for open-domain agent learning and demonstrates the efficacy of multi-role co-evolution for scalable, self-improving agents. The data and code of this paper are released at https://github.com/forangel2014/Towards-Agentic-Self-Learning

Wangtao Sun, Haotian Xu, Xuanqing Yu et al.

Although Large Language Models (LLMs) are showing impressive performance on a wide range of Natural Language Processing tasks, researchers have found that they still have limited ability to conduct induction. Recent works mainly adopt ``post processes'' paradigms to improve the performance of LLMs on induction (e.g., the hypothesis search & refinement methods), but their performance is still constrained by the inherent inductive capability of the LLMs. In this paper, we propose a novel framework, Induction through Deduction (ItD), to enable the LLMs to teach themselves induction through deduction. The ItD framework is composed of two main components: a Deductive Data Generation module to generate induction data and a Naive Bayesian Induction module to optimize the fine-tuning and decoding of LLMs. Our empirical results showcase the effectiveness of ItD on two induction benchmarks, achieving relative performance improvement of 36% and 10% compared with previous state-of-the-art, respectively. Our ablation study verifies the effectiveness of two key modules of ItD. We also verify the effectiveness of ItD across different LLMs and deductors. The data and code of this paper can be found at https://anonymous.4open.science/r/ItD-E844.

Yao Xu, Mingyu Xu, Fangyu Lei et al.

Recently, models such as OpenAI-o1 and DeepSeek-R1 have demonstrated remarkable performance on complex reasoning tasks through Long Chain-of-Thought (Long-CoT) reasoning. Although distilling this capability into student models significantly enhances their performance, this paper finds that fine-tuning LLMs with full parameters or LoRA with a low rank on long CoT data often leads to Cyclical Reasoning, where models repeatedly reiterate previous inference steps until the maximum length limit. Further analysis reveals that smaller differences in representations between adjacent tokens correlates with a higher tendency toward Cyclical Reasoning. To mitigate this issue, this paper proposes Shift Feedforward Networks (Shift-FFN), a novel approach that edits the current token's representation with the previous one before inputting it to FFN. This architecture dynamically amplifies the representation differences between adjacent tokens. Extensive experiments on multiple mathematical reasoning tasks demonstrate that LoRA combined with Shift-FFN achieves higher accuracy and a lower rate of Cyclical Reasoning across various data sizes compared to full fine-tuning and standard LoRA. Our data and code are available at https://anonymous.4open.science/r/Shift-FFN

Wangtao Sun, Xiang Cheng, Xing Yu et al.

Reinforcement learning from human feedback (RLHF) is a critical technique for training large language models. However, conventional reward models based on the Bradley-Terry model (BTRM) often suffer from overconfidence when faced with inconsistent labels or out-of-distribution samples, leading to reward hacking, where the policy model blindly optimizes for proxy rewards while degrading true performance. This paper proposes the Probabilistic Uncertain Reward Model (PURM), which generalizes the Bradley-Terry model to learn the reward distributions that emerged from the preference data. We theoretically derive the loss function of PURM and introduce a novel method that uses the overlap between distributions to quantify uncertainty. Empirical results show that PURM outperforms existing methods with more accurate reward and sound uncertainty estimations, and sustains effective learning for more optimization steps and obtain higher maximum win rate in RLHF. The data and code of this paper are released at https://anonymous.4open.science/r/Probabilistic-Uncertain-Reward-Model/

Wangtao Sun, Shizhu He, Jun Zhao et al.

With good explanatory power and controllability, rule-based methods play an important role in many tasks such as knowledge reasoning and decision support. However, existing studies primarily focused on learning chain-like rules, which limit their semantic expressions and accurate prediction abilities. As a result, chain-like rules usually fire on the incorrect grounding values, producing inaccurate or even erroneous reasoning results. In this paper, we propose the concept of tree-like rules on knowledge graphs to expand the application scope and improve the reasoning ability of rule-based methods. Meanwhile, we propose an effective framework for refining chain-like rules into tree-like rules. Experimental comparisons on four public datasets show that the proposed framework can easily adapt to other chain-like rule induction methods and the refined tree-like rules consistently achieve better performances than chain-like rules on link prediction. The data and code of this paper can be available at https://anonymous.4open.science/r/tree-rule-E3CD/.

Wangtao Sun, Haotian Xu, Huanxuan Liao et al.

The interaction with Large Language Models (LLMs) through instructions has been extensively investigated in the research community. While instructions have been widely used as the guidelines for task solving, this paper further notices that both instructions and parameters are the compression of task data. Therefore, they could be strongly correlated and can be learned to predict one from the other. This paper proposes a novel neural network framework, SHIP (\textbf{Sh}uttle between the \textbf{I}nstructions and the \textbf{P}arameters), to model and learn the mutual mappings between the instructions and the parameters of LLMs. We verify that SHIP can effectively map one of the instructions/parameters to the other by evaluating it on the tasks of instruction deduction and induction. The results show that SHIP performs better than existing baseline methods in terms of deductive capabilities while significantly surpassing them in inductive capabilities. Moreover, SHIP can effectively combine the two mapping processes to perform excellent inductive reasoning. The code and data for this paper are released at https://anonymous.4open.science/r/Shuttle-Between-Instructions-Parameters/.