Aligning large language models (LLMs) with human preferences is critical for real-world deployment, yet existing methods like RLHF face computational and stability challenges. While DPO establishes an offline paradigm with single hyperparameter $\beta$, subsequent methods like SimPO reintroduce complexity through dual parameters ($\beta$, $\gamma$). We propose {ReLU-based Preference Optimization (RePO)}, a streamlined algorithm that eliminates $\beta$ via two advances: (1) retaining SimPO's reference-free margins but removing $\beta$ through gradient analysis, and (2) adopting a ReLU-based max-margin loss that naturally filters trivial pairs. Theoretically, RePO is characterized as SimPO's limiting case ($\beta \to \infty$), where the logistic weighting collapses to binary thresholding, forming a convex envelope of the 0-1 loss. Empirical results on AlpacaEval 2 and Arena-Hard show that RePO outperforms DPO and SimPO across multiple base models, requiring only one hyperparameter to tune.

Large language models (LLMs), due to their advanced natural language capabilities, have seen significant success in applications where the user interface is usually a conversational artificial intelligence (AI) agent and engages the user through multi-round conversations. However, many scenarios require the agents to exhibit stronger social and conversational intelligence and demonstrate more human-like (anthropomorphic) reactions. This is an aspect that foundational LLMs have yet to fully address such that a single call of foundational models might be insufficient. To bridge this gap, we propose a two-stage solution. In this work, we focus on the first stage, introducing a multi-module framework designed to replicate the key aspects of human intelligence involved in conversations. This framework comprises thinking modules for reasoning, resource modules for managing knowledge and external information, and response modules for generating contextually appropriate interactions. With all the modules cooperating, the framework would empower the agents to provide a better human-like conversation experience. In the second stage of our approach, these conversational data, after filtering and labeling, can serve as training and testing data for reinforcement learning, enabling AI to better capture human preferences. This stage is left for future work. In our experiments, volunteers engaged in over 3000 rounds of conversation with the same AI character powered by a standalone LLM and our framework which integrates the same LLM. A separate group of evaluators rated the conversation samples, revealing that our framework significantly enhanced the social and conversational intelligence, even without fine-tuning the LLM.

Preference learning is critical for aligning large language models (LLMs) with human values, with the quality of preference datasets playing a crucial role in this process. While existing metrics primarily assess data quality based on either explicit or implicit reward margins, they often provide contradictory evaluations for the same data. To address this issue, we introduce the alignment potential metric, which quantifies the gap from the model's current implicit reward margin to the target explicit reward margin, thereby estimating the model's potential to align with the preference data. Empirical results demonstrate that training on data selected by this metric consistently enhances alignment performance, surpassing existing metrics across different base models and optimization objectives. Furthermore, our method extends to self-play data generation frameworks, where the metric is used to identify high-quality data within the self-generated content by LLMs. Under this data generation scenario, our method surpasses current state-of-the-art (SOTA) results across various training settings and demonstrates continuous improvements in alignment performance as dataset size and training iterations increase.

Tool learning has emerged as a crucial capability for large language models (LLMs) to solve complex real-world tasks through interaction with external tools. Existing approaches face significant challenges, including reliance on hand-crafted prompts, difficulty in multi-step planning, and lack of precise error diagnosis and reflection mechanisms. We propose ToolCoder, a novel framework that reformulates tool learning as a code generation task. Inspired by software engineering principles, ToolCoder transforms natural language queries into structured Python function scaffold and systematically breaks down tasks with descriptive comments, enabling LLMs to leverage coding paradigms for complex reasoning and planning. It then generates and executes function implementations to obtain final responses. Additionally, ToolCoder stores successfully executed functions in a repository to promote code reuse, while leveraging error traceback mechanisms for systematic debugging, optimizing both execution efficiency and robustness. Experiments demonstrate that ToolCoder achieves superior performance in task completion accuracy and execution reliability compared to existing approaches, establishing the effectiveness of code-centric approaches in tool learning.

Large language models (LLMs) have had a profound impact on various aspects of people's lives, particularly as the foundational technology behind conversational applications such as chatbots. These models have become indispensable as virtual assistants, offering powerful capabilities for various tasks, including addressing common-sense queries, generating summaries for academic papers [16], and solving programming challenges and tasks [11]. Despite their impressive functionality, LLMs are of some limitations. Issues such as hallucinations and the inability to provide the most up-to-date information or private knowledge hinder their reliability in directly serving for knowledge-intensive applications. These shortcomings can be mitigated by integrating LLMs with external information in the input context [20, 28]. One notable approach is retrieval-augmented generation (RAG) techniques [1, 10], which enhances LLMs by equipping them with retrieval capabilities, allows LLMs to address questions that exceed the scope of their pre-trained internal knowledge. RAG has proven highly effective in improving performance on question-answering (QA) tasks emphasizing faithfulness to truths, showcasing its potential to bridge the gap between static pre-trained knowledge and dynamic, context-specific information. While many real-world applications have adopted RAG techniques [13, 22], open-source frameworks have also emerged to facilitate the adaptation of RAG to a wide range of tasks [14, 18] for the public to hold RAG application services themselves with local data. While these open-source RAG frameworks provide convenient starting points for building RAG-based applications, there remain significant opportunities for improvement, especially in more practical and complicated scenarios, e.g., efficient multi-source knowledge retrieval, which provides primary motivations for this paper.

Leveraging large language models (LLMs), an agent can utilize retrieval-augmented generation (RAG) techniques to integrate external knowledge and increase the reliability of its responses. Current RAG-based agents integrate single, domain-specific knowledge sources, limiting their ability and leading to hallucinated or inaccurate responses when addressing cross-domain queries. Integrating multiple knowledge bases into a unified RAG-based agent raises significant challenges, including increased retrieval overhead and data sovereignty when sensitive data is involved. In this work, we propose RopMura, a novel multi-agent system that addresses these limitations by incorporating highly efficient routing and planning mechanisms. RopMura features two key components: a router that intelligently selects the most relevant agents based on knowledge boundaries and a planner that decomposes complex multi-hop queries into manageable steps, allowing for coordinating cross-domain responses. Experimental results demonstrate that RopMura effectively handles both single-hop and multi-hop queries, with the routing mechanism enabling precise answers for single-hop queries and the combined routing and planning mechanisms achieving accurate, multi-step resolutions for complex queries.

The structural properties of naturally arising social graphs are extensively studied to understand their evolution. Prior approaches for modeling network dynamics typically rely on rule-based models, which lack realism and generalizability, or deep learning-based models, which require large-scale training datasets. Social graphs, as abstract graph representations of entity-wise interactions, present an opportunity to explore network evolution mechanisms through realistic simulations of human-item interactions. Leveraging the pre-trained social consensus knowledge embedded in large language models (LLMs), we present GraphAgent-Generator (GAG), a novel simulation-based framework for dynamic, text-attributed social graph generation. GAG simulates the temporal node and edge generation processes for zero-shot social graph generation. The resulting graphs exhibit adherence to seven key macroscopic network properties, achieving an 11% improvement in microscopic graph structure metrics. Through the node classification benchmarking task, we validate GAG effectively captures the intricate text-structure correlations in graph generation. Furthermore, GAG supports generating graphs with up to nearly 100,000 nodes or 10 million edges through large-scale LLM-based agent simulation with parallel acceleration, achieving a minimum speed-up of 90.4%. The source code is available at https://github.com/Ji-Cather/GraphAgent.

In the domain of Multimodal Large Language Models (MLLMs), achieving human-centric video understanding remains a formidable challenge. Existing benchmarks primarily emphasize object and action recognition, often neglecting the intricate nuances of human emotions, behaviors, and speech visual alignment within video content. We present HumanVBench, an innovative benchmark meticulously crafted to bridge these gaps in the evaluation of video MLLMs. HumanVBench comprises 17 carefully designed tasks that explore two primary dimensions: inner emotion and outer manifestations, spanning static and dynamic, basic and complex, as well as single-modal and cross-modal aspects. With two advanced automated pipelines for video annotation and distractor-included QA generation, HumanVBench utilizes diverse state-of-the-art (SOTA) techniques to streamline benchmark data synthesis and quality assessment, minimizing human annotation dependency tailored to human-centric multimodal attributes. A comprehensive evaluation across 16 SOTA video MLLMs reveals notable limitations in current performance, especially in cross-modal and temporal alignment, underscoring the necessity for further refinement toward achieving more human-like understanding. HumanVBench is open-sourced to facilitate future advancements and real-world applications in video MLLMs.

The burgeoning field of foundation models necessitates advanced data processing mechanisms capable of harnessing vast valuable data with varied types utilized by these models. Nevertheless, the current landscape presents unique challenges that traditional data processing frameworks cannot handle effectively, especially with multimodal intricacies. In response, we present Data-Juicer 2.0, a new system offering fruitful data processing capabilities backed by over a hundred operators spanning various modalities like text, image, audio, and video. With seamless compatibility and dedicated optimization to popular dataset hubs like Hugging Face and computing engines like Ray, Data-Juicer 2.0 enhances its predecessor in both usability, efficiency, and programmability. It features an easily accessible user interface layer that supports decoupled Python interactions, RESTful APIs, and conversational commands. Alongside this, it contains a core runtime layer optimized for adaptive execution and management across different dataset scales, processing demands, and computational environments, while shielding unnecessary system details. Extensive empirical evaluations demonstrate Data-Juicer 2.0's remarkable performance and scalability, highlighting its capability to efficiently process tens of billions of data samples with tens of thousands of CPU cores. The system is publicly available, actively maintained, and broadly adopted in diverse research endeavors, practical applications, and real-world products such as Alibaba Cloud PAI.

We propose a general two-stage algorithm that enjoys a provable scaling law for the test-time compute of large language models (LLMs). Given an input problem, the proposed algorithm first generates $N$ candidate solutions, and then chooses the best one via a multiple-round knockout tournament where each pair of candidates are compared for $K$ times and only the winners move on to the next round. In a minimalistic implementation, both stages can be executed with a black-box LLM alone and nothing else (e.g., no external verifier or reward model), and a total of $N \times (K + 1)$ highly parallelizable LLM calls are needed for solving an input problem. Assuming that a generated candidate solution is correct with probability $p_{\text{gen}} > 0$ and a comparison between a pair of correct and incorrect solutions identifies the right winner with probability $p_{\text{comp}} > 0.5$ (i.e., better than a random guess), we prove theoretically that the failure probability of the proposed algorithm decays to zero exponentially with respect to $N$ and $K$: $$\mathbb{P}(\text{final output is incorrect}) \le (1 - p_{\text{gen}})^N + \lceil \log_2 N \rceil e^{-2 K (p_{\text{comp}} - 0.5)^2}.$$ Our empirical results with the challenging MMLU-Pro benchmark validate the technical assumptions, as well as the efficacy of the proposed algorithm and the gains from scaling up its test-time compute.