LLMEval-3: A Large-Scale Longitudinal Study on Robust and Fair Evaluation of Large Language Models

Existing evaluation of Large Language Models (LLMs) on static benchmarks is vulnerable to data contamination and leaderboard overfitting, critical issues that obscure true model capabilities. To address this, we introduce LLMEval-3, a framework for dynamic evaluation of LLMs. LLMEval-3 is built on a proprietary bank of 220k graduate-level questions, from which it dynamically samples unseen test sets for each evaluation run. Its automated pipeline ensures integrity via contamination-resistant data curation, a novel anti-cheating architecture, and a calibrated LLM-as-a-judge process achieving 90% agreement with human experts, complemented by a relative ranking system for fair comparison. An 20-month longitudinal study of nearly 50 leading models reveals a performance ceiling on knowledge memorization and exposes data contamination vulnerabilities undetectable by static benchmarks. The framework demonstrates exceptional robustness in ranking stability and consistency, providing strong empirical validation for the dynamic evaluation paradigm. LLMEval-3 offers a robust and credible methodology for assessing the true capabilities of LLMs beyond leaderboard scores, promoting the development of more trustworthy evaluation standards.

Regret Minimization in Population Network Games: Vanishing Heterogeneity and Convergence to Equilibria

Understanding and predicting the behavior of large-scale multi-agents in games remains a fundamental challenge in multi-agent systems. This paper examines the role of heterogeneity in equilibrium formation by analyzing how smooth regret-matching drives a large number of heterogeneous agents with diverse initial policies toward unified behavior. By modeling the system state as a probability distribution of regrets and analyzing its evolution through the continuity equation, we uncover a key phenomenon in diverse multi-agent settings: the variance of the regret distribution diminishes over time, leading to the disappearance of heterogeneity and the emergence of consensus among agents. This universal result enables us to prove convergence to quantal response equilibria in both competitive and cooperative multi-agent settings. Our work advances the theoretical understanding of multi-agent learning and offers a novel perspective on equilibrium selection in diverse game-theoretic scenarios.

Quantize More, Lose Less: Autoregressive Generation from Residually Quantized Speech Representations

Text-to-speech (TTS) synthesis has seen renewed progress under the discrete modeling paradigm. Existing autoregressive approaches often rely on single-codebook representations, which suffer from significant information loss. Even with post-hoc refinement techniques such as flow matching, these methods fail to recover fine-grained details (e.g., prosodic nuances, speaker-specific timbres), especially in challenging scenarios like singing voice or music synthesis. We propose QTTS, a novel TTS framework built upon our new audio codec, QDAC. The core innovation of QDAC lies in its end-to-end training of an ASR-based auto-regressive network with a GAN, which achieves superior semantic feature disentanglement for scalable, near-lossless compression. QTTS models these discrete codes using two innovative strategies: the Hierarchical Parallel architecture, which uses a dual-AR structure to model inter-codebook dependencies for higher-quality synthesis, and the Delay Multihead approach, which employs parallelized prediction with a fixed delay to accelerate inference speed. Our experiments demonstrate that the proposed framework achieves higher synthesis quality and better preserves expressive content compared to baseline. This suggests that scaling up compression via multi-codebook modeling is a promising direction for high-fidelity, general-purpose speech and audio generation.

Zero-Shot Solving of Imaging Inverse Problems via Noise-Refined Likelihood Guided Diffusion Models

Diffusion models have achieved remarkable success in imaging inverse problems owing to their powerful generative capabilities. However, existing approaches typically rely on models trained for specific degradation types, limiting their generalizability to various degradation scenarios. To address this limitation, we propose a zero-shot framework capable of handling various imaging inverse problems without model retraining. We introduce a likelihood-guided noise refinement mechanism that derives a closed-form approximation of the likelihood score, simplifying score estimation and avoiding expensive gradient computations. This estimated score is subsequently utilized to refine the model-predicted noise, thereby better aligning the restoration process with the generative framework of diffusion models. In addition, we integrate the Denoising Diffusion Implicit Models (DDIM) sampling strategy to further improve inference efficiency. The proposed mechanism can be applied to both optimization-based and sampling-based schemes, providing an effective and flexible zero-shot solution for imaging inverse problems. Extensive experiments demonstrate that our method achieves superior performance across multiple inverse problems, particularly in compressive sensing, delivering high-quality reconstructions even at an extremely low sampling rate (5%).

Single-Node Trigger Backdoor Attacks in Graph-Based Recommendation Systems

Graph recommendation systems have been widely studied due to their ability to effectively capture the complex interactions between users and items. However, these systems also exhibit certain vulnerabilities when faced with attacks. The prevailing shilling attack methods typically manipulate recommendation results by injecting a large number of fake nodes and edges. However, such attack strategies face two primary challenges: low stealth and high destructiveness. To address these challenges, this paper proposes a novel graph backdoor attack method that aims to enhance the exposure of target items to the target user in a covert manner, without affecting other unrelated nodes. Specifically, we design a single-node trigger generator, which can effectively expose multiple target items to the target user by inserting only one fake user node. Additionally, we introduce constraint conditions between the target nodes and irrelevant nodes to mitigate the impact of fake nodes on the recommendation system's performance. Experimental results show that the exposure of the target items reaches no less than 50% in 99% of the target users, while the impact on the recommendation system's performance is controlled within approximately 5%.

Towards a Generalizable Bimanual Foundation Policy via Flow-based Video Prediction

Learning a generalizable bimanual manipulation policy is extremely challenging for embodied agents due to the large action space and the need for coordinated arm movements. Existing approaches rely on Vision-Language-Action (VLA) models to acquire bimanual policies. However, transferring knowledge from single-arm datasets or pre-trained VLA models often fails to generalize effectively, primarily due to the scarcity of bimanual data and the fundamental differences between single-arm and bimanual manipulation. In this paper, we propose a novel bimanual foundation policy by fine-tuning the leading text-to-video models to predict robot trajectories and training a lightweight diffusion policy for action generation. Given the lack of embodied knowledge in text-to-video models, we introduce a two-stage paradigm that fine-tunes independent text-to-flow and flow-to-video models derived from a pre-trained text-to-video model. Specifically, optical flow serves as an intermediate variable, providing a concise representation of subtle movements between images. The text-to-flow model predicts optical flow to concretize the intent of language instructions, and the flow-to-video model leverages this flow for fine-grained video prediction. Our method mitigates the ambiguity of language in single-stage text-to-video prediction and significantly reduces the robot-data requirement by avoiding direct use of low-level actions. In experiments, we collect high-quality manipulation data for real dual-arm robot, and the results of simulation and real-world experiments demonstrate the effectiveness of our method.

The Avengers: A Simple Recipe for Uniting Smaller Language Models to Challenge Proprietary Giants

As proprietary giants increasingly dominate the race for ever-larger language models, a pressing question arises for the open-source community: can smaller models remain competitive across a broad range of tasks? In this paper, we present the Avengers--a simple recipe that effectively leverages the collective intelligence of open-source, smaller language models. Our framework is built upon four lightweight operations: (i) embedding: encode queries using a text embedding model; (ii) clustering: group queries based on their semantic similarity; (iii) scoring: scores each model's performance within each cluster; and (iv) voting: improve outputs via repeated sampling and voting. At inference time, each query is embedded and assigned to its nearest cluster. The top-performing model(s) within that cluster are selected to generate the response using the Self-Consistency or its multi-model variant. Remarkably, with 10 open-source models (~7B parameters each), the Avengers collectively outperforms GPT-4.1 on 10 out of 15 datasets (spanning mathematics, code, logic, knowledge, and affective tasks). In particular, it surpasses GPT-4.1 on mathematics tasks by 18.21% and on code tasks by 7.46%. Furthermore, the Avengers delivers superior out-of-distribution generalization, and remains robust across various embedding models, clustering algorithms, ensemble strategies, and values of its sole parameter--the number of clusters. We have open-sourced the code on GitHub: https://github.com/ZhangYiqun018/Avengers

Code2Logic: Game-Code-Driven Data Synthesis for Enhancing VLMs General Reasoning

Visual-language Chain-of-Thought (CoT) data resources are relatively scarce compared to text-only counterparts, limiting the improvement of reasoning capabilities in Vision Language Models (VLMs). However, high-quality vision-language reasoning data is expensive and labor-intensive to annotate. To address this issue, we leverage a promising resource: game code, which naturally contains logical structures and state transition processes. Therefore, we propose Code2Logic, a novel game-code-driven approach for multimodal reasoning data synthesis. Our approach leverages Large Language Models (LLMs) to adapt game code, enabling automatic acquisition of reasoning processes and results through code execution. Using the Code2Logic approach, we developed the GameQA dataset to train and evaluate VLMs. GameQA is cost-effective and scalable to produce, challenging for state-of-the-art models, and diverse with 30 games and 158 tasks. Surprisingly, despite training solely on game data, VLMs demonstrated out of domain generalization, specifically Qwen2.5-VL-7B improving performance by 2.33\% across 7 diverse vision-language benchmarks. Our code and dataset are available at https://github.com/tongjingqi/Code2Logic.

Taming Recommendation Bias with Causal Intervention on Evolving Personal Popularity

Popularity bias occurs when popular items are recommended far more frequently than they should be, negatively impacting both user experience and recommendation accuracy. Existing debiasing methods mitigate popularity bias often uniformly across all users and only partially consider the time evolution of users or items. However, users have different levels of preference for item popularity, and this preference is evolving over time. To address these issues, we propose a novel method called CausalEPP (Causal Intervention on Evolving Personal Popularity) for taming recommendation bias, which accounts for the evolving personal popularity of users. Specifically, we first introduce a metric called {Evolving Personal Popularity} to quantify each user's preference for popular items. Then, we design a causal graph that integrates evolving personal popularity into the conformity effect, and apply deconfounded training to mitigate the popularity bias of the causal graph. During inference, we consider the evolution consistency between users and items to achieve a better recommendation. Empirical studies demonstrate that CausalEPP outperforms baseline methods in reducing popularity bias while improving recommendation accuracy.

HyperDet: Source Detection in Hypergraphs via Interactive Relationship Construction and Feature-rich Attention Fusion

Hypergraphs offer superior modeling capabilities for social networks, particularly in capturing group phenomena that extend beyond pairwise interactions in rumor propagation. Existing approaches in rumor source detection predominantly focus on dyadic interactions, which inadequately address the complexity of more intricate relational structures. In this study, we present a novel approach for Source Detection in Hypergraphs (HyperDet) via Interactive Relationship Construction and Feature-rich Attention Fusion. Specifically, our methodology employs an Interactive Relationship Construction module to accurately model both the static topology and dynamic interactions among users, followed by the Feature-rich Attention Fusion module, which autonomously learns node features and discriminates between nodes using a self-attention mechanism, thereby effectively learning node representations under the framework of accurately modeled higher-order relationships. Extensive experimental validation confirms the efficacy of our HyperDet approach, showcasing its superiority relative to current state-of-the-art methods.