Enhancing the mathematical reasoning capabilities of Large Language Models (LLMs) is of great scientific and practical significance. Researchers typically employ process-supervised reward models (PRMs) to guide the reasoning process, effectively improving the models’ reasoning abilities. However, existing methods for constructing process supervision training data, such as manual annotation and per-step Monte Carlo estimation, are often costly or suffer from poor quality. To address these challenges, this paper introduces a framework called EpicPRM (Efficient, Precise, Cheap), which annotates each intermediate reasoning step based on its quantified contribution and uses an adaptive binary search algorithm to enhance both annotation precision and efficiency. Using this approach, we efficiently construct a high-quality process supervision training dataset named Epic50k, consisting of 50k annotated intermediate steps. Compared to other publicly available datasets, the PRM trained on Epic50k demonstrates significantly superior performance.

Recent advances in text-only “slow-thinking” reasoning have prompted efforts to transfer this capability to vision-language models (VLMs), for training visual reasoning models (VRMs). However, such transfer faces critical challenges: Effective “slow thinking” in VRMs requires visual reflection, the ability to check the reasoning process based on visual information. Through quantitative analysis, we observe that current VRMs exhibit limited visual reflection, as their attention to visual information diminishes rapidly with longer generated responses. To address this challenge, we propose a new VRM Reflection-V, which enhances visual reflection based on reasoning data construction for cold-start and reward design for reinforcement learning (RL). Firstly, we construct vision-centered reasoning data by leveraging an agent that interacts between VLMs and reasoning LLMs, enabling cold-start learning of visual reflection patterns. Secondly, a visual attention based reward model is employed during RL to encourage reasoning based on visual information. Therefore, Reflection-V demonstrates significant improvements across multiple visual reasoning benchmarks. Furthermore, Reflection-V maintains a stronger and more consistent reliance on visual information during visual reasoning, indicating effective enhancement in visual reflection capabilities.

Efficiently updating multilingual knowledge in large language models (LLMs) without disrupting coherent factual representations across languages remains a significant challenge. While deploying separate editing systems for each language might seem viable, this approach incurs substantial costs due to the need to manage multiple models. A more efficient solution involves integrating knowledge updates across all languages into a unified model. However, sequential edits across languages often lead to destructive parameter interference, significantly degrading multilingual generalization and the accuracy of injected knowledge. To address this issue, we propose LangEdit, a novel null-space constrained framework designed to precisely isolate language-specific knowledge updates. The core innovation of LangEdit lies in its ability to project parameter updates for each language onto the orthogonal complement of other languages’ subspaces. This approach mathematically guarantees update independence while preserving multilingual generalization capabilities. We conduct a comprehensive evaluation across three model architectures, six languages, and four downstream tasks, demonstrating that LangEdit effectively mitigates parameter interference and outperforms existing state-of-the-art editing methods. Our results highlight its potential for enabling efficient and accurate multilingual knowledge updates in LLMs.

Extreme Multi-label classification (XML) is an important yet challenging machine learning task, that assigns to each instance its most relevant candidate labels from an extremely large label collection, where the numbers of labels, features and instances could be thousands or millions. XML is more and more on demand in the Internet industries, accompanied with the increasing business scale / scope and data accumulation. The extremely large label collections yield challenges such as computational complexity, inter-label dependency and noisy labeling. Many methods have been proposed to tackle these challenges, based on different mathematical formulations. In this paper, we propose a deep learning XML method, with a word-vector-based self-attention, followed by a ranking-based AutoEncoder architecture. The proposed method has three major advantages: 1) the autoencoder simultaneously considers the inter-label dependencies and the feature-label dependencies, by projecting labels and features onto a common embedding space; 2) the ranking loss not only improves the training efficiency and accuracy but also can be extended to handle noisy labeled data; 3) the efficient attention mechanism improves feature representation by highlighting feature importance. Experimental results on benchmark datasets show the proposed method is competitive to state-of-the-art methods.