PretrainRL: Alleviating Factuality Hallucination of Large Language Models at the Beginning

Large language models (LLMs), despite their powerful capabilities, suffer from factual hallucinations where they generate verifiable falsehoods. We identify a root of this issue: the imbalanced data distribution in the pretraining corpus, which leads to a state of "low-probability truth" and "high-probability falsehood". Recent approaches, such as teaching models to say "I don't know" or post-hoc knowledge editing, either evade the problem or face catastrophic forgetting. To address this issue from its root, we propose \textbf{PretrainRL}, a novel framework that integrates reinforcement learning into the pretraining phase to consolidate factual knowledge. The core principle of PretrainRL is "\textbf{debiasing then learning}." It actively reshapes the model's probability distribution by down-weighting high-probability falsehoods, thereby making "room" for low-probability truths to be learned effectively. To enable this, we design an efficient negative sampling strategy to discover these high-probability falsehoods and introduce novel metrics to evaluate the model's probabilistic state concerning factual knowledge. Extensive experiments on three public benchmarks demonstrate that PretrainRL significantly alleviates factual hallucinations and outperforms state-of-the-art methods.

Enhancing Conversational Agents via Task-Oriented Adversarial Memory Adaptation

Conversational agents struggle to handle long conversations due to context window limitations. Therefore, memory systems are developed to leverage essential historical information. Existing memory systems typically follow a pipeline of offline memory construction and update, and online retrieval. Despite the flexible online phase, the offline phase remains fixed and task-independent. In this phase, memory construction operates under a predefined workflow and fails to emphasize task relevant information. Meanwhile, memory updates are guided by generic metrics rather than task specific supervision. This leads to a misalignment between offline memory preparation and task requirements, which undermines downstream task performance. To this end, we propose an Adversarial Memory Adaptation mechanism (AMA) that aligns memory construction and update with task objectives by simulating task execution. Specifically, first, a challenger agent generates question answer pairs based on the original dialogues. The constructed memory is then used to answer these questions, simulating downstream inference. Subsequently, an evaluator agent assesses the responses and performs error analysis. Finally, an adapter agent analyzes the error cases and performs dual level updates on both the construction strategy and the content. Through this process, the memory system receives task aware supervision signals in advance during the offline phase, enhancing its adaptability to downstream tasks. AMA can be integrated into various existing memory systems, and extensive experiments on long dialogue benchmark LoCoMo demonstrate its effectiveness.

Attention Needs to Focus: A Unified Perspective on Attention Allocation

The Transformer architecture, a cornerstone of modern Large Language Models (LLMs), has achieved extraordinary success in sequence modeling, primarily due to its attention mechanism. However, despite its power, the standard attention mechanism is plagued by well-documented issues: representational collapse and attention sink. Although prior work has proposed approaches for these issues, they are often studied in isolation, obscuring their deeper connection. In this paper, we present a unified perspective, arguing that both can be traced to a common root -- improper attention allocation. We identify two failure modes: 1) Attention Overload, where tokens receive comparable high weights, blurring semantic features that lead to representational collapse; 2) Attention Underload, where no token is semantically relevant, yet attention is still forced to distribute, resulting in spurious focus such as attention sink. Building on this insight, we introduce Lazy Attention, a novel mechanism designed for a more focused attention distribution. To mitigate overload, it employs positional discrimination across both heads and dimensions to sharpen token distinctions. To counteract underload, it incorporates Elastic-Softmax, a modified normalization function that relaxes the standard softmax constraint to suppress attention on irrelevant tokens. Experiments on the FineWeb-Edu corpus, evaluated across nine diverse benchmarks, demonstrate that Lazy Attention successfully mitigates attention sink and achieves competitive performance compared to both standard attention and modern architectures, while reaching up to 59.58% attention sparsity.

The Best of the Two Worlds: Harmonizing Semantic and Hash IDs for Sequential Recommendation

Conventional Sequential Recommender Systems (SRS) typically assign unique Hash IDs (HID) to construct item embeddings. These HID embeddings effectively learn collaborative information from historical user-item interactions, making them vulnerable to situations where most items are rarely consumed (the long-tail problem). Recent methods that incorporate auxiliary information often suffer from noisy collaborative sharing caused by co-occurrence signals or semantic homogeneity caused by flat dense embeddings. Semantic IDs (SIDs), with their capability of code sharing and multi-granular semantic modeling, provide a promising alternative. However, the collaborative overwhelming phenomenon hinders the further development of SID-based methods. The quantization mechanisms commonly compromise the uniqueness of identifiers required for modeling head items, creating a performance seesaw between head and tail items. To address this dilemma, we propose \textbf{\name}, a novel framework that harmonizes the SID and HID. Specifically, we devise a dual-branch modeling architecture that enables the model to capture both the multi-granular semantics within SID while preserving the unique collaborative identity of HID. Furthermore, we introduce a dual-level alignment strategy that bridges the two representations, facilitating knowledge transfer and supporting robust preference modeling. Extensive experiments on three real-world datasets show that \name~ effectively balances recommendation quality for both head and tail items while surpassing the existing baselines. The implementation code can be found online\footnote{https://github.com/ziwliu8/H2Rec}.

A Multi-Expert Structural-Semantic Hybrid Framework for Unveiling Historical Patterns in Temporal Knowledge Graphs

Temporal knowledge graph reasoning aims to predict future events with knowledge of existing facts and plays a key role in various downstream tasks. Previous methods focused on either graph structure learning or semantic reasoning, failing to integrate dual reasoning perspectives to handle different prediction scenarios. Moreover, they lack the capability to capture the inherent differences between historical and non-historical events, which limits their generalization across different temporal contexts. To this end, we propose a Multi-Expert Structural-Semantic Hybrid (MESH) framework that employs three kinds of expert modules to integrate both structural and semantic information, guiding the reasoning process for different events. Extensive experiments on three datasets demonstrate the effectiveness of our approach.

Model Merging for Knowledge Editing

Large Language Models (LLMs) require continuous updates to maintain accurate and current knowledge as the world evolves. While existing knowledge editing approaches offer various solutions for knowledge updating, they often struggle with sequential editing scenarios and harm the general capabilities of the model, thereby significantly hampering their practical applicability. This paper proposes a two-stage framework combining robust supervised fine-tuning (R-SFT) with model merging for knowledge editing. Our method first fine-tunes the LLM to internalize new knowledge fully, then merges the fine-tuned model with the original foundation model to preserve newly acquired knowledge and general capabilities. Experimental results demonstrate that our approach significantly outperforms existing methods in sequential editing while better preserving the original performance of the model, all without requiring any architectural changes. Code is available at: https://github.com/Applied-Machine-Learning-Lab/MM4KE.

Training-free LLM Merging for Multi-task Learning

Large Language Models (LLMs) have demonstrated exceptional capabilities across diverse natural language processing (NLP) tasks. The release of open-source LLMs like LLaMA and Qwen has triggered the development of numerous fine-tuned models tailored for various tasks and languages. In this paper, we explore an important question: is it possible to combine these specialized models to create a unified model with multi-task capabilities. We introduces Hierarchical Iterative Merging (Hi-Merging), a training-free method for unifying different specialized LLMs into a single model. Specifically, Hi-Merging employs model-wise and layer-wise pruning and scaling, guided by contribution analysis, to mitigate parameter conflicts. Extensive experiments on multiple-choice and question-answering tasks in both Chinese and English validate Hi-Merging's ability for multi-task learning. The results demonstrate that Hi-Merging consistently outperforms existing merging techniques and surpasses the performance of models fine-tuned on combined datasets in most scenarios. Code is available at: https://github.com/Applied-Machine-Learning-Lab/Hi-Merging.

Navigate the Unknown: Enhancing LLM Reasoning with Intrinsic Motivation Guided Exploration

Reinforcement learning (RL) has emerged as a pivotal method for improving the reasoning capabilities of Large Language Models (LLMs). However, prevalent RL approaches such as Proximal Policy Optimization (PPO) and Group-Regularized Policy Optimization (GRPO) face critical limitations due to their reliance on sparse outcome-based rewards and inadequate mechanisms for incentivizing exploration. These limitations result in inefficient guidance for multi-step reasoning processes. Specifically, sparse reward signals fail to deliver effective or sufficient feedback, particularly for challenging problems. Furthermore, such reward structures induce systematic biases that prioritize exploitation of familiar trajectories over novel solution discovery. These shortcomings critically hinder performance in complex reasoning tasks, which inherently demand iterative refinement across ipntermediate steps. To address these challenges, we propose an Intrinsic Motivation guidEd exploratioN meThOd foR LLM Reasoning (i-MENTOR), a novel method designed to both deliver dense rewards and amplify explorations in the RL-based training paradigm. i-MENTOR introduces three key innovations: trajectory-aware exploration rewards that mitigate bias in token-level strategies while maintaining computational efficiency; dynamic reward scaling to stabilize exploration and exploitation in large action spaces; and advantage-preserving reward implementation that maintains advantage distribution integrity while incorporating exploratory guidance. Experiments across three public datasets demonstrate i-MENTOR's effectiveness with a 22.39% improvement on the difficult dataset Countdown-4.

Bridge the Domains: Large Language Models Enhanced Cross-domain Sequential Recommendation

Cross-domain Sequential Recommendation (CDSR) aims to extract the preference from the user's historical interactions across various domains. Despite some progress in CDSR, two problems set the barrier for further advancements, i.e., overlap dilemma and transition complexity. The former means existing CDSR methods severely rely on users who own interactions on all domains to learn cross-domain item relationships, compromising the practicability. The latter refers to the difficulties in learning the complex transition patterns from the mixed behavior sequences. With powerful representation and reasoning abilities, Large Language Models (LLMs) are promising to address these two problems by bridging the items and capturing the user's preferences from a semantic view. Therefore, we propose an LLMs Enhanced Cross-domain Sequential Recommendation model (LLM4CDSR). To obtain the semantic item relationships, we first propose an LLM-based unified representation module to represent items. Then, a trainable adapter with contrastive regularization is designed to adapt the CDSR task. Besides, a hierarchical LLMs profiling module is designed to summarize user cross-domain preferences. Finally, these two modules are integrated into the proposed tri-thread framework to derive recommendations. We have conducted extensive experiments on three public cross-domain datasets, validating the effectiveness of LLM4CDSR. We have released the code online.

Generative Auto-Bidding with Value-Guided Explorations

Auto-bidding, with its strong capability to optimize bidding decisions within dynamic and competitive online environments, has become a pivotal strategy for advertising platforms. Existing approaches typically employ rule-based strategies or Reinforcement Learning (RL) techniques. However, rule-based strategies lack the flexibility to adapt to time-varying market conditions, and RL-based methods struggle to capture essential historical dependencies and observations within Markov Decision Process (MDP) frameworks. Furthermore, these approaches often face challenges in ensuring strategy adaptability across diverse advertising objectives. Additionally, as offline training methods are increasingly adopted to facilitate the deployment and maintenance of stable online strategies, the issues of documented behavioral patterns and behavioral collapse resulting from training on fixed offline datasets become increasingly significant. To address these limitations, this paper introduces a novel offline Generative Auto-bidding framework with Value-Guided Explorations (GAVE). GAVE accommodates various advertising objectives through a score-based Return-To-Go (RTG) module. Moreover, GAVE integrates an action exploration mechanism with an RTG-based evaluation method to explore novel actions while ensuring stability-preserving updates. A learnable value function is also designed to guide the direction of action exploration and mitigate Out-of-Distribution (OOD) problems. Experimental results on two offline datasets and real-world deployments demonstrate that GAVE outperforms state-of-the-art baselines in both offline evaluations and online A/B tests. The implementation code is publicly available to facilitate reproducibility and further research.