JoyAI-LLM Flash: Advancing Mid-Scale LLMs with Token Efficiency

• URL: http://arxiv.org/abs/2604.03044v2
• Date: Wed, 08 Apr 2026 07:22:54 GMT
• Title: JoyAI-LLM Flash: Advancing Mid-Scale LLMs with Token Efficiency
• Abstract summary: JoyAI-LLM Flash is an efficient Mixture-of-Experts (MoE) language model designed to redefine the trade-off between strong performance and token efficiency.<n>JoyAI-LLM Flash is pretrained on a massive corpus of 20 trillion tokens and further optimized through a rigorous post-training pipeline.
• Score: 56.46698214215968
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We introduce JoyAI-LLM Flash, an efficient Mixture-of-Experts (MoE) language model designed to redefine the trade-off between strong performance and token efficiency in the sub-50B parameter regime. JoyAI-LLM Flash is pretrained on a massive corpus of 20 trillion tokens and further optimized through a rigorous post-training pipeline, including supervised fine-tuning (SFT), Direct Preference Optimization (DPO), and large-scale reinforcement learning (RL) across diverse environments. To improve token efficiency, JoyAI-LLM Flash strategically balances \emph{thinking} and \emph{non-thinking} cognitive modes and introduces FiberPO, a novel RL algorithm inspired by fibration theory that decomposes trust-region maintenance into global and local components, providing unified multi-scale stability control for LLM policy optimization. To enhance architectural sparsity, the model comprises 48B total parameters while activating only 2.7B parameters per forward pass, achieving a substantially higher sparsity ratio than contemporary industry leading models of comparable scale. To further improve inference throughput, we adopt a joint training-inference co-design that incorporates dense Multi-Token Prediction (MTP) and Quantization-Aware Training (QAT). We release the checkpoints for both JoyAI-LLM-48B-A3B Base and its post-trained variants on Hugging Face to support the open-source community.

Related papers

• DiRL: An Efficient Post-Training Framework for Diffusion Language Models [54.405206032785706]
  Diffusion Language Models (dLLMs) have emerged as promising alternatives to Auto-Regressive (AR) models.<n>Existing methods suffer from computational inefficiency and objective mismatches between training and inference.<n>We introduce DiRL, an efficient post-training framework that tightly integrates FlexAttention-accelerated blockwise training with LMDeploy-optimized inference.
  arXiv  Detail & Related papers  (2025-12-23T08:33:19Z)
• Enhancing Reasoning for Diffusion LLMs via Distribution Matching Policy Optimization [44.14678335188207]
  Diffusion large language models (dLLMs) are promising alternatives to autoregressive large language models (AR-LLMs)<n>Reinforcement learning (RL) is a crucial component for dLLMs to achieve comparable performance with AR-LLMs on important tasks, such as reasoning.<n>This paper proposes Distribution Matching Policy Optimization (DMPO), a principled and theoretically grounded RL fine-tuning method.
  arXiv  Detail & Related papers  (2025-10-09T13:59:50Z)
• DiFFPO: Training Diffusion LLMs to Reason Fast and Furious via Reinforcement Learning [37.20873499361773]
  We propose a unified framework for training masked diffusion large language models (dLLMs) to reason better (furious)<n>We first unify the existing baseline approach by proposing to train surrogate policies via off-policy RL, whose likelihood is much more tractable as an approximation to the true dLLM policy.<n>We also propose a new direction of joint training efficient samplers/controllers of dLLMs policy. Via RL, we incentivize dLLMs' natural multi-token prediction capabilities by letting the model learn to adaptively allocate an inference threshold for each prompt.
  arXiv  Detail & Related papers  (2025-10-02T16:57:24Z)
• Ring-lite: Scalable Reasoning via C3PO-Stabilized Reinforcement Learning for LLMs [51.21041884010009]
  Ring-lite is a Mixture-of-Experts (MoE)-based large language model optimized via reinforcement learning (RL)<n>Our approach matches the performance of state-of-the-art (SOTA) small-scale reasoning models on challenging benchmarks.
  arXiv  Detail & Related papers  (2025-06-17T17:12:34Z)
• MobiLlama: Towards Accurate and Lightweight Fully Transparent GPT [87.4910758026772]
  "Bigger the better" has been the predominant trend in recent Large Language Models (LLMs) development. This paper explores the "less is more" paradigm by addressing the challenge of designing accurate yet efficient Small Language Models (SLMs) for resource constrained devices.
  arXiv  Detail & Related papers  (2024-02-26T18:59:03Z)
• Semi-Federated Learning: Convergence Analysis and Optimization of A Hybrid Learning Framework [70.83511997272457]
  We propose a semi-federated learning (SemiFL) paradigm to leverage both the base station (BS) and devices for a hybrid implementation of centralized learning (CL) and FL. We propose a two-stage algorithm to solve this intractable problem, in which we provide the closed-form solutions to the beamformers.
  arXiv  Detail & Related papers  (2023-10-04T03:32:39Z)
• BiERL: A Meta Evolutionary Reinforcement Learning Framework via Bilevel Optimization [34.24884427152513]
  We propose a general meta ERL framework via bilevel optimization (BiERL) We design an elegant meta-level architecture that embeds the inner-level's evolving experience into an informative population representation. We perform extensive experiments in MuJoCo and Box2D tasks to verify that as a general framework, BiERL outperforms various baselines and consistently improves the learning performance for a diversity of ERL algorithms.
  arXiv  Detail & Related papers  (2023-08-01T09:31:51Z)
• FedGPO: Heterogeneity-Aware Global Parameter Optimization for Efficient Federated Learning [11.093360539563657]
  Federated learning (FL) has emerged as a solution to deal with the risk of privacy leaks in machine learning training. We propose FedGPO to optimize the energy-efficiency of FL use cases while guaranteeing model convergence. In our experiments, FedGPO improves the model convergence time by 2.4 times, and achieves 3.6 times higher energy efficiency over the baseline settings.
  arXiv  Detail & Related papers  (2022-11-30T01:22:57Z)
• Low-Latency Federated Learning over Wireless Channels with Differential Privacy [142.5983499872664]
  In federated learning (FL), model training is distributed over clients and local models are aggregated by a central server. In this paper, we aim to minimize FL training delay over wireless channels, constrained by overall training performance as well as each client's differential privacy (DP) requirement.
  arXiv  Detail & Related papers  (2021-06-20T13:51:18Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.