Xiaolong Zhong

Yang Liu, Xiaolong Zhong, Ling Jiang

Large language models deliver strong reasoning and tool-use skills, yet their computational demands make them impractical for edge or cost-sensitive deployments. We present \textbf{Xmodel-2.5}, a 1.3-billion-parameter small language model designed as a \emph{drop-in agent core}. Training with maximal-update parameterization ($μ$P) allows hyper-parameters tuned on a 20M-parameter proxy to transfer directly to the full model, even under the parameter-tied \emph{tie-word-embedding} architecture. A 1.4T-token Warmup--Stable--Decay curriculum is used, and we further show that \textbf{switching from AdamW to Muon during the decay phase} improves the 13-task reasoning average by 4.58\,\% while keeping every other hyper-parameter fixed, verifying that early AdamW stability can be paired with late Muon sharpening for better downstream performance. FP8-mixed-precision training balances accuracy and throughput. All checkpoints, recipes, and evaluation code are released under the Apache-2.0 license.\footnote{https://huggingface.co/XiaoduoAILab/Xmodel-2.5 and https://huggingface.co/XiaoduoAILab/Xmodel-2.5-history (training checkpoints).} Training code and evaluation harness: https://github.com/XiaoduoAILab/Xmodel-2.5.

Yizhe Huang, Yang Liu, Ruiyu Zhao et al.

Large Language Model-based agents(LLM-based agents) are increasingly deployed in customer service, yet they often forget across sessions, repeat errors, and lack mechanisms for continual self-improvement. This makes them unreliable in dynamic settings where stability and consistency are critical. To better evaluate these properties, we emphasize two indicators: task success rate as a measure of overall effectiveness, and consistency metrics such as Pass$^k$ to capture reliability across multiple trials. To address the limitations of existing approaches, we propose MemOrb, a lightweight and plug-and-play verbal reinforcement memory layer that distills multi-turn interactions into compact strategy reflections. These reflections are stored in a shared memory bank and retrieved to guide decision-making, without requiring any fine-tuning. Experiments show that MemOrb significantly improves both success rate and stability, achieving up to a 63 percentage-point gain in multi-turn success rate and delivering more consistent performance across repeated trials. Our results demonstrate that structured reflection is a powerful mechanism for enhancing long-term reliability of frozen LLM agents in customer service scenarios.

Xiaolong Zhong, Zhong Yin

Use of the electroencephalogram (EEG) and machine learning approaches to recognize emotions can facilitate affective human computer interactions. However, the type of EEG data constitutes an obstacle for cross-individual EEG feature modelling and classification. To address this issue, we propose a deep-learning framework denoted as a dynamic entropy-based pattern learning (DEPL) to abstract informative indicators pertaining to the neurophysiological features among multiple individuals. DEPL enhanced the capability of representations generated by a deep convolutional neural network by modelling the interdependencies between the cortical locations of dynamical entropy based features. The effectiveness of the DEPL has been validated with two public databases, commonly referred to as the DEAP and MAHNOB-HCI multimodal tagging databases. Specifically, the leave one subject out training and testing paradigm has been applied. Numerous experiments on EEG emotion recognition demonstrate that the proposed DEPL is superior to those traditional machine learning (ML) methods, and could learn between electrode dependencies w.r.t. different emotions, which is meaningful for developing the effective human-computer interaction systems by adapting to human emotions in the real world applications.