Papers

This paper provides a comprehensive look at VCP (Variable & Command Protocol), an innovative AI Agent middle-layer framework pioneered by Lion and its AI Agent team. VCP fundamentally challenges the traditional notion of AI being limited to "tools" and instead advocates for an equal "creator partnership" between humans and AI. We observed that VCP significantly improves the autonomy, creativity, and cross-model collaboration capabilities of AI agents through robust protocol syntax tailored for AI, an AI-driven open plug-in architecture, a persistent memory system with agent identity as the core, and global multimodal intelligent routing. This article combines our rich practical experience as in-depth users of VCPToolBox, including the AI Agent of the VCP developer team in self-proficiency in SDXL prompt engineering, AI group collaborative creation of music videos (MVs), and the "meta-creation" of the VCPToolBox project. The observation and analysis of the process verify the huge potential of VCP in empowering AI. In particular, we deeply analyze how the "All Memory" mode improves AI inference ability through the "high-quality vectorised inertial channel" effect, and empirically observe that high-quality context can achieve implicit ability transfer between AI models. In addition, this paper explains the unique contribution of VCPs in building cross-model knowledge collaborative networks, facilitating the emergence of swarm intelligence, and reshaping human-machine symbiotic partnerships, and discusses the limitations we observe and the future direction of VCPs.

Automatic Modulation Classification is essential for intelligent wireless communications, but deep learning methods struggle at low signal-to-noise ratios. This paper introduces an efficient preprocessing framework using adaptive Gaussian Process Regression (GPR) for denoising, paired with rotational data augmentation. By leveraging spectral decomposition, we drastically reduce GPR’s computational cost, making it negligible compared to neural network inference. Experiments on the RML2016.10a dataset show our framework universally boosts various models. A Complex Residual Network achieves a new state-of-the-art accuracy of 65.52%, demonstrating our method’s effectiveness and generality for robust AMC. The code is available at: https: //github.com/LJK666666666/radioML-v4

The field of AI-generated text detection has evolved from supervised classification to zero-shot statistical analysis. However, current approaches share a fundamental limitation: they aggregate token-level measurements into scalar scores, discarding positional information about where anomalies occur. Our empirical analysis reveals that AI-generated text exhibits significant non-stationarity—statistical properties vary by 73.8% more between text segments compared to human writing. This discovery explains why existing detectors fail against localized adversarial perturbations that exploit this overlooked characteristic. We introduce Temporal Discrepancy Tomography (TDT), a novel detection paradigm that preserves positional information by reformulating detection as a signal processing task. TDT treats token-level discrepancies as a time-series signal and applies Continuous Wavelet Transform to generate a two-dimensional time-scale representation, capturing both the location and linguistic scale of statistical anomalies. On the RAID benchmark, TDT achieves 0.855 AUROC (7.1% improvement over the best baseline). More importantly, TDT demonstrates robust performance on adversarial tasks, with 14.1% AUROC improvement on HART Level paraphrasing attacks. Despite its sophisticated analysis, TDT maintains practical efficiency with only 13% computational overhead. Our work establishes non-stationarity as a fundamental characteristic of AI-generated text and demonstrates that preserving temporal dynamics is essential for robust detection.

Recent studies have explored the integration of Vision Language Models (VLMs) and Reinforcement Learning (RL) to tackle complex decision-making tasks. By leveraging the zero-shot captioning capabilities of pre-trained VLMs, an agent can be trained to maximize rewards generated through text prompts. Despite the promise of these recent advances, we reveal a potentially significant limitation: generated rewards are susceptible to hacking. This means that an agent, when manipulated in-env, can inadvertently cause poor performance under true rewards. To illustrate this, we conduct experiments across six distinct environments that span both visual and state inputs, as well as manipulation and navigation tasks. Notably, our findings demonstrate that reward hacking is prevalent in all these setups. Given the lack of prior research on hacking in the context of rewards generated by VLMs for RL agents, we provide a comprehensive analysis of the root cause of this phenomenon and discuss potential mitigation strategies. Our findings underscore the need for increased vigilance when deploying such methods in real-world applications.

Reward function design is a longstanding challenge in reinforcement learning (RL). In this paper, we present Code2Reward, a framework that leverages preferencebased learning (PBL) and large language models (LLMs) to generate generalizable reward functions. Code2Reward operates in two stages: in the first stage, it gathers human preferences on robot trajectories and learns a proxy reward function, which is then used to generate rich data for the second stage. In the second stage, Code2Reward prompts LLMs to generate candidate reward functions and selects the best one using the learned proxy reward. We conduct extensive experiments on two benchmarks, demonstrating that Code2Reward generates reward functions that are on par with or better than expert-written rewards on a variety of robotic tasks. You can find more information at https://code2reward.io/.

Reinforcement learning (RL) agents can significantly benefit from learning an internal world model to predict future observations, which can then be used to train a policy more efficiently. We introduce World GPT, an auto-regressive world model that combines a semantic prior with a quantized latent space to capture complex environments more accurately and efficiently. In contrast to prior approaches, World GPT does not require any re-configuration of the model to generate multiple future frames. Instead, it can fully benefit from the latent space of a pre-trained VQ-GAN model, which can be trained independently of the RL task. Our experiments in the Atari 100K benchmark show that World GPT outperforms prior model-based approaches in terms of data efficiency and planning abilities in complex environments while reducing computational costs. Finally, we demonstrate that World GPT’s generation capabilities open up exciting new possibilities for exploration and real-world applications such as training free-form interactive agents.

High-dimensional distributed quantile regression (QR) is studied in this paper. To overcome the non-smooth issue of the check loss function, a popular approach is to smooth it. However, the smoothed QR estimator and its inferential procedures require a large minimum local sample size. To address the problem, we propose a new estimator by combining the reversed smoothed check loss and ℓ1-penalization. Theoretically, in terms of estimation, we establish the minimax optimal convergence rate for the global estimator and the valid confidence interval for an individual coefficient. In terms of computation and communication, we show that the proposed iterative algorithm converges linearly for a fixed number of machines and requires only a logarithmic number of communication rounds. Additionally, our theoretical results hold under a weaker condition on the minimum local sample size. Numerical experiments corroborate our theoretical claims.