arXiv subjects

James Gong

Publications and source records attributed to James Gong.

4 recordsLinked to original sources

arXiv2026

Yuan3.0 Ultra: A Trillion-Parameter Enterprise-Oriented MoE LLM

We introduce Yuan3.0 Ultra, an open-source Mixture-of-Experts (MoE) large language model featuring 68.8B activated parameters and 1010B total parameters, specially designed to enhance performance on enterprise scenarios tasks while maintaining competitive capabilities on general purpose tasks. We propose Layer-Adaptive Expert Pruning (LAEP) algorithm designed for the pre-training stage of MoE LLMs. In contrast to previous expert pruning approaches that operate primarily in the post-training phase, the proposed algorithm enhances training efficiency by selectively pruning underutilized experts and reorganizing experts across computing devices according to token distribution statistics. Comprehensive experiments demonstrate that LAEP effectively reduces model size and substantially improves pre-training efficiency. When pre-training Yuan3.0 Ultra from scratch original with 1515B parameters, this algorithm delivers a 49\% boost in pre-training efficiency and a 33.3\% reduction in total parameters, while preserving the model's outstanding multi-domain performance. On enterprise scenario benchmarks including Docmatix, ChatRAG, SummEval and MMTab, Yuan3.0 Ultra achieves leading accuracy. The model and codes are publicly available at https://github.com/Yuan-lab-LLM/Yuan3.0-Ultra.

cs.LG↗

arXiv2026

Yuan3.0 Flash: An Open Multimodal Large Language Model for Enterprise Applications

We introduce Yuan3.0 Flash, an open-source Mixture-of-Experts (MoE) MultiModal Large Language Model featuring 3.7B activated parameters and 40B total parameters, specifically designed to enhance performance on enterprise-oriented tasks while maintaining competitive capabilities on general-purpose tasks. To address the overthinking phenomenon commonly observed in Large Reasoning Models (LRMs), we propose Reflection-aware Adaptive Policy Optimization (RAPO), a novel RL training algorithm that effectively regulates overthinking behaviors. In enterprise-oriented tasks such as retrieval-augmented generation (RAG), complex table understanding, and summarization, Yuan3.0 Flash consistently achieves superior performance. Moreover, it also demonstrates strong reasoning capabilities in domains such as mathematics, science, etc., attaining accuracy comparable to frontier model while requiring only approximately 1/4 to 1/2 of the average tokens. Yuan3.0 Flash has been fully open-sourced to facilitate further research and real-world deployment: https://github.com/Yuan-lab-LLM/Yuan3.0.

cs.AI↗

arXiv2025

Reshaping the Forward-Forward Algorithm with a Similarity-Based Objective

Backpropagation is the pivotal algorithm underpinning the success of artificial neural networks, yet it has critical limitations such as biologically implausible backward locking and global error propagation. To circumvent these constraints, the Forward-Forward algorithm was proposed as a more biologically plausible method that replaces the backward pass with an additional forward pass. Despite this advantage, the Forward-Forward algorithm significantly trails backpropagation in accuracy, and its optimal form exhibits low inference efficiency due to multiple forward passes required. In this work, the Forward-Forward algorithm is reshaped through its integration with similarity learning frameworks, eliminating the need for multiple forward passes during inference. This proposed algorithm is named Forward-Forward Algorithm Unified with Similarity-based Tuplet loss (FAUST). Empirical evaluations on MNIST, Fashion-MNIST, and CIFAR-10 datasets indicate that FAUST substantially improves accuracy, narrowing the gap with backpropagation. On CIFAR-10, FAUST achieves 56.22\% accuracy with a simple multi-layer perceptron architecture, approaching the backpropagation benchmark of 57.63\% accuracy.

cs.LG↗

arXiv2025

Mono-Forward: Revisiting Forward-Forward through Objective-Locality Decomposition

Backpropagation remains the dominant algorithm for training deep neural networks, but it incurs substantial memory overhead and relies on global error propagation, which is often regarded as biologically implausible. The Forward-Forward (FF) algorithm is an appealing local-learning alternative to backpropagation, yet it still lags behind backpropagation in accuracy. A central unresolved question is whether this gap arises from FF's locality or from the positive-negative double-pass goodness objective used to train each layer. In this work, we revisit FF under the supervised setting through a decomposition that separates these two design choices. Our analysis suggests that FF's performance limitations are not explained by locality alone, but are also likely influenced by its goodness objective. Motivated by this view, we introduce Mono-Forward (MF), a simplification of FF that preserves its locality while replacing the contrastive goodness objective with a standard multi-class cross-entropy objective applied locally at each layer, serving as a controlled baseline for evaluating local learning under a standard classification objective. Across MLPs and convolutional networks, MF outperforms vanilla FF and remains competitive in multiple FF variants. On MLP-Mixers, MF achieves stronger results on PathMNIST than backpropagation while requiring only 31% of backpropagation's memory.

cs.LG↗