Large language models (LLMs) have made impressive strides in mathematical reasoning, often fine-tuned using rejection sampling, which retains only correct reasoning trajectories. While effective, this paradigm treats supervision as a binary filter that systematically excludes teacher-generated errors, leaving a gap in how reasoning failures are modeled during training. In this paper, we propose TrajFusion, a fine-tuning strategy that reframes rejection sampling as a structured supervision construction process. Specifically, TrajFusion forms fused trajectories that explicitly model trial-and-error reasoning by interleaving selected incorrect trajectories with reflection prompts and correct trajectories. The length of the fused sample is adaptively controlled based on the frequency and diversity of teacher errors, providing richer supervision for challenging problems while safely reducing to vanilla rejection sampling fine-tuning (RFT) when error signals are uninformative. TrajFusion requires no changes to the architecture or training objective. Extensive experiments across multiple math benchmarks demonstrate that TrajFusion consistently outperforms RFT, particularly on challenging and long-form reasoning problems.

Vision-Language Models (VLMs) encode images and videos into abundant tokens, which contain substantial redundancy and computation cost. While visual token pruning mitigates the issue, most existing methods lack insight into the intrinsic property of the vision encoder itself. In this work, we dive into the vision encoder and prove that the middle layers pay more attention to the main objects of the image qualitatively and quantitatively, while the deep layers to tokens with rich global information. Utilizing this Hierarchical attention pattern, we propose HiPrune, a training-free and model-agnostic token Pruning method. HiPrune identifies three types of visual tokens according to their attention in different phases of the vision encoder, which preserves different levels of information. By coupling with the similarity of text tokens, we propose a prompt-aware variance, HiPrune++, which further improves instruction following performance under a very low token budget. Extensive experiments across four representative VLMs show that HiPrune achieves up to 99.3% of task accuracy with only 1/3 of the tokens, while reducing inference FLOPs by 58.7%. HiPrune++ maintains up to 99.9% accuracy with 2/9 tokens, highlighting robustness under high-resolution.

Automatically generated radiology reports often receive high scores from existing evaluation metrics but fail to earn clinicians’ trust. This gap reveals fundamental flaws in how current metrics assess the quality of generated reports. We rethink the design and evaluation of these metrics and propose a clinically grounded Meta-Evaluation framework. We define clinically grounded criteria spanning clinical alignment and key metric capabilities, including discrimination, robustness, and monotonicity. Using a fine-grained dataset of ground truth and rewritten report pairs annotated with error types, clinical significance labels, and explanations, we systematically evaluate existing metrics and reveal their limitations in interpreting clinical semantics, such as failing to distinguish clinically significant errors, over-penalizing harmless variations, and lacking consistency across error severity levels. Our framework offers guidance for building more clinically reliable evaluation methods.

“尽管大语言模型(LLM)在自然语言处理领域取得巨大成功,但是伴随其千亿级参数 规 模 的 训 练 也 产 生 了 巨 大 的 计 算 成 本 。 小 规 模 大 语 言 模 型(SLLM)作 为 低 资 源场景下实现LLM部署的可替代方案,任务处理能力与LLM尚存在明显差距。尽管上下文学习(ICL)等提示方法在一定程度上提升了SLLM的问题处理能力,但基于人工构建的提示往往需要参与者具备特定的专业领域知识,这给LLM的普适推广带来了挑战。针对以上问题,本文提出了一个基于SLLM的问题推理框架,通过在推理路径生成和答案生成两个阶段之间引入基于逐步语义验证器(SSVRP)的推理路径排序选择机制,在无人干预情况下实现SLLM推理能力提升。实验结果表明,SSVRP有效地增强了SLLM的推理性能,在4个推理任务中的平均准确率分别达到了54.3%,90.6%,64.3%和63.7%,并在其中3个推理任务中都取得了最新的SOTA结果。”

Previous Sign Language Translation (SLT) methods achieve superior performance by relying on gloss annotations. However, labeling high-quality glosses is a labor-intensive task, which limits the further development of SLT. Although some approaches work towards gloss-free SLT through jointly training the visual encoder and translation network, these efforts still suffer from poor performance and inefficient use of the powerful Large Language Model (LLM). Most seriously, we find that directly introducing LLM into SLT will lead to insufficient learning of visual representations as LLM dominates the learning curve. To address these problems, we propose Factorized Learning assisted with Large Language Model (FLa-LLM) for gloss-free SLT. Concretely, we factorize the training process into two stages. In the visual initialing stage, we employ a lightweight translation model after the visual encoder to pre-train the visual encoder. In the LLM fine-tuning stage, we freeze the acquired knowledge in the visual encoder and integrate it with a pre-trained LLM to inspire the LLM’s translation potential. This factorized training strategy proves to be highly effective as evidenced by significant improvements achieved across three SLT datasets which are all conducted under the gloss-free setting.

Unsupervised constituency parsing aims to learn a constituency parser from a training corpus without parse tree annotations. While many methods have been proposed to tackle the problem, including statistical and neural methods, their experimental results are often not directly comparable due to discrepancies in datasets, data preprocessing, lexicalization, and evaluation metrics. In this paper, we first examine experimental settings used in previous work and propose to standardize the settings for better comparability between methods. We then empirically compare several existing methods, including decade-old and newly proposed ones, under the standardized settings on English and Japanese, two languages with different branching tendencies. We find that recent models do not show a clear advantage over decade-old models in our experiments. We hope our work can provide new insights into existing methods and facilitate future empirical evaluation of unsupervised constituency parsing.