Recent sign language translation methods have advanced visual-language alignment, yet autoregressive generation remains vulnerable to unstable policy updates under ambiguous visual evidence. Such instability may cause locally plausible continuations to accumulate into sentence-level semantic deviations, especially when reinforcement learning is used with noisy reward and advantage estimates. Existing PPO-style policy optimization applies confidence-agnostic clipping bounds to all generated tokens, which limits its ability to distinguish between over-amplified confident predictions and low-probability tokens that may still carry useful corrective signals. To address this issue, we propose CAPO-SLT, a reinforcement learning based framework for stable sign language translation generation. CAPO introduces a confidence-aware policy optimization strategy that adapts the upper clipping bound according to old-policy token probability. It applies more conservative updates to high-confidence predictions while preserving update capacity for low-probability positive-advantage tokens. In addition, CAPO caps negative-advantage updates to improve robustness under noisy supervision. This design provides a simple drop-in replacement for standard clipping and stabilizes generation-stage optimization without changing the underlying visual encoder or reward function. Experiments on Chinese and American sign language benchmarks show consistent improvements over pose-based baselines under the same input setting, suggesting that CAPO-SLT provides an effective optimization strategy for stable sign language translation generation.

As Multimodal Large Language Models (MLLMs) mature, GUI agents are evolving from static interactions to complex navigation. While Reinforcement Learning (RL) has emerged as a promising paradigm for training MLLM agents on dynamic GUI tasks, its effective application faces a dilemma.Standard Offline RL often relies on static step-level data, neglecting global trajectory semantics such as task completion and execution quality. Conversely, Online RL captures the long-term dynamics but suffers from high interaction costs and potential environmental instability. To bridge this gap, we propose SOLAR-RL (Semi Online Long-horizon RL). Instead of relying solely on expensive online interactions, our framework integrates global trajectory insights directly into the offline learning process. Specifically, we reconstruct diverse rollout candidates from static data, detect the first failure point using per-step validity signals, and retroactively assign dense step-level rewards with target-aligned shaping to reflect trajectory-level execution quality—effectively simulating online feedback without interaction costs.Extensive experiments demonstrate that SOLAR-RL significantly improves long-horizon task completion rates and robustness compared to strong baselines, offering a sample-efficient solution for autonomous GUI navigation.

Practical video editing is not only pixel generation: an editor must turn a brief, a clip pool, music metadata, and hard constraints into an executable timeline. We study this decision layer as executable video-editing planning and introduce RefineCut, which, unlike workflow systems that wrap a prompted frontier model, trains a compact open-weight planner for it. The planner edits a typed timeline through structured patches covering clip selection, trimming, ordering, transitions, and duration and music alignment; a deterministic verifier applies each patch and checks it against an explicit constraint ledger. Because editing has no single ground-truth repair, we do not imitate teachers directly: RefineCut replays every multi-teacher branch through the verifier and keeps verifier-best repairs as supervision. A second stage, RefineCut-Evo, lets the student score its own repairs with the verifier and a task rubric and trains on high-margin preference pairs, so the final 8B planner runs in a closed verifier loop with no teacher calls at inference. On RefineCut-Bench (3,578 tasks, 7,971 captioned clips, 499 music tracks, explicit ledgers), verifier-replayed distillation lifts the planner from 0.620 to 0.858 on the protocol-specific Video-Editing Score and RefineCut-Evo reaches 0.924; the gain transfers to Llama-3.1-8B and GLM-4-9B, and in the same closed loop the 8B planner matches or exceeds its frontier teachers. Code and RefineCut-Bench are publicly released; see the Data Availability statement.

High-quality tool-use data is critical for training language models to interact effectively with external tools. However, existing synthetic approaches typically follow a generate-then-filter paradigm with static post-hoc verification, often yielding inefficient data with imbalanced feature distributions. We propose ToolLoop, a closed-loop framework that decomposes synthesis into three progressive stages: (1) sampling function name combinations as ground truth; (2) backward derivation of user queries; and (3) forward derivation of tool calls. At each stage, dynamic self-feedback iteratively guides the model toward high-quality generation, realizing a transition from generate-then-filter to generate-verify-refine. On the Berkeley Function Calling Leaderboard (BFCL), a 4B parameter model trained with our 11K synthetic examples achieves 86.40% accuracy in non-reasoning mode, while an Isolate variant that removes BFCL-overlapping candidate functions still reaches 86.07%. Cross-benchmark evaluation on ACEBench further demonstrates strong generalization, with 72.1% overall accuracy using only 18.3% of baseline training data.

In natural language processing (NLP) tasks, pure reinforcement learning fine-tuning methods often suffer from inefficient exploration and slow convergence; while supervised fine-tuning (SFT) methods, although efficient in training, have limited performance ceiling and less solid theoretical foundation compared to reinforcement learning. To address efficiency-capability trade-off, we propose the Guess-Think-Answer (GTA) framework that combines the efficiency of SFT with the capability gains of RL in a unified training paradigm. GTA works by having the model first produce a provisional guess (optimized via cross-entropy loss), then reflect on this guess before generating the final answer, with RL rewards shaping both the final output and the format of the entire GTA structure. This hybrid approach achieves both faster convergence than pure RL and higher performance ceiling than pure SFT. To mitigate gradient conflicts between the two training signals, we employ loss masking and gradient constraints. Empirical results on three text classification benchmarks demonstrate that GTA substantially accelerates convergence while outperforming both standalone SFT and RL baselines.

Large Language Models (LLMs) have become integral to daily life, especially advancing as intelligent assistants through on-device deployment on smartphones. However, existing LLM evaluation benchmarks predominantly focus on objective tasks like mathematics and coding in English, which do not necessarily reflect the practical use cases of on-device LLMs in real-world mobile scenarios, especially for Chinese users. To address these gaps, we introduce SmartBench, the first benchmark designed to evaluate the capabilities of on-device LLMs in Chinese mobile contexts. We analyze functionalities provided by representative smartphone manufacturers and divide them into five categories: text summarization, text Q&A, information extraction, content creation, and notification management, further detailed into 20 specific tasks. For each task, we construct high-quality datasets comprising 50 to 200 question-answer pairs that reflect everyday mobile interactions, and we develop automated evaluation criteria tailored for these tasks. We conduct comprehensive evaluations of on-device LLMs and MLLMs using SmartBench and also assess their performance after quantized deployment on real smartphone NPUs. Our contributions provide a standardized framework for evaluating on-device LLMs in Chinese, promoting further development and optimization in this critical area. Code and data will be available at https://github.com/vivo-ai-lab/SmartBench.

In recent years, the use of large language models (LLMs) for text classification has attracted widespread attention. Despite this, the classification accuracy of LLMs has not yet universally surpassed that of smaller models. LLMs can enhance their performance in text classification through fine-tuning. However, existing data quality research based on LLMs is challenging to apply directly to solve text classification problems. To further improve the performance of LLMs in classification tasks, this paper proposes a data quality enhancement (DQE) method for text classification based on LLMs. This method starts by using a greedy algorithm to select data, dividing the dataset into sampled and unsampled subsets, and then performing fine-tuning of the LLMs using the sampled data. Subsequently, this model is used to predict the outcomes for the unsampled data, categorizing incorrectly predicted data into uncovered, difficult, and noisy data. Experimental results demonstrate that our method effectively enhances the performance of LLMs in text classification tasks and significantly improves training efficiency, saving nearly half of the training time. Our method has achieved state-of-the-art performance in several open-source classification tasks.

Transformer-based large language models (LLMs) encounter challenges in processing long sequences on edge devices due to the quadratic complexity of attention mechanisms and growing memory demands from Key-Value (KV) cache. Existing KV cache optimizations struggle with irreversible token eviction in long-output tasks, while alternative sequence modeling architectures prove costly to adopt within established Transformer infrastructure. We present EdgeInfinite, a memory-efficient solution for infinite contexts that integrates compressed memory into Transformer-based LLMs through a trainable memory-gating module. This approach maintains full compatibility with standard Transformer architectures, requiring fine-tuning only a small part of parameters, and enables selective activation of the memory-gating module for long and short context task routing. The experimental result shows that EdgeInfinite achieves comparable performance to baseline Transformer-based LLM on long context benchmarks while optimizing memory consumption and time to first token.

Due to the continuous emergence of new data, version updates have become an indispensable requirement for Large Language Models (LLMs). The training paradigms for version updates of LLMs include pre-training from scratch (PTFS) and continual pre-training (CPT). Preliminary experiments demonstrate that PTFS achieves better pre-training performance, while CPT has lower training cost. Moreover, their performance and training cost gaps widen progressively with version updates. To investigate the underlying reasons for this phenomenon, we analyze the effect of learning rate adjustments during the two stages of CPT: preparing an initialization checkpoint and continual pre-training based on this checkpoint. We find that a large learning rate in the first stage and a complete learning rate decay process in the second stage are crucial for version updates of LLMs. Hence, we propose a learning rate path switching training paradigm. Our paradigm comprises one main path, where we pre-train a LLM with the maximal learning rate, and multiple branching paths, each of which corresponds to an update of the LLM with newly-added training data. Extensive experiments demonstrate the effectiveness and generalization of our paradigm. Particularly, when training four versions of LLMs, our paradigm reduces the total training cost to 58% compared to PTFS, while maintaining comparable pre-training performance.

“本文介绍了我们在第二十四届中国计算语言学大会手语数字人翻译质量评测中提交的参赛系统。本次评测任务旨在评测手语数字人将汉语翻译成中国手语方面的自然性和准确性。本文介绍的手语数字人翻译系统首先通过手语翻译算法将汉语文本翻译成手语文本,然后将手语文本对应的手语动作单元运用动作融合算法合成为自然、完整的手语数字人动作,同时借助面部驱动算法将口型、表情等非语言元素自然地融入手语合成中,实现带微表情的和唇形同步的手语数字人。最终,我们在官方手语数字人翻译质量的人工评测集上取得了3.513的综合评分,获得了该任务第一名的成绩。”