We present Dyve, a dynamic process verifier that enhances reasoning error detection in large language models by integrating fast and slow thinking, inspired by Kahneman's Systems Theory. Dyve adaptively applies immediate token-level confirmation System 1 for straightforward steps and comprehensive analysis System 2 for complex ones. Leveraging a novel step-wise consensus-filtered process supervision technique, combining Monte Carlo estimation with LLM based evaluation, Dyve curates high-quality supervision signals from noisy data. Experimental results on ProcessBench and the MATH dataset confirm that Dyve significantly outperforms existing process-based verifiers and boosts performance in Best-of-N settings.

Visual-Language Models (VLMs) have shown remarkable performance across various tasks, particularly in recognizing geographic information from images. However, significant challenges remain, including biases and privacy concerns. To systematically address these issues in the context of geographic information recognition, we introduce a benchmark dataset consisting of 1,200 images paired with detailed geographic metadata. Evaluating four VLMs, we find that while these models demonstrate the ability to recognize geographic information from images, achieving up to $53.8\%$ accuracy in city prediction, they exhibit significant regional biases. Specifically, performance is substantially higher for economically developed and densely populated regions compared to less developed ($-12.5\%$) and sparsely populated ($-17.0\%$) areas. Moreover, the models exhibit regional biases, frequently overpredicting certain locations; for instance, they consistently predict Sydney for images taken in Australia. The strong performance of VLMs also raises privacy concerns, particularly for users who share images online without the intent of being identified. Our code and dataset are publicly available at https://github.com/uscnlp-lime/FairLocator.

DeepSeek-R1, known for its low training cost and exceptional reasoning capabilities, has achieved state-of-the-art performance on various benchmarks. However, detailed evaluations from the perspective of real-world applications are lacking, making it challenging for users to select the most suitable DeepSeek models for their specific needs. To address this gap, we evaluate the DeepSeek-V3, DeepSeek-R1, DeepSeek-R1-Distill-Qwen series, and DeepSeek-R1-Distill-Llama series on A-Eval, an application-driven benchmark. By comparing original instruction-tuned models with their distilled counterparts, we analyze how reasoning enhancements impact performance across diverse practical tasks. Our results show that reasoning-enhanced models, while generally powerful, do not universally outperform across all tasks, with performance gains varying significantly across tasks and models. To further assist users in model selection, we quantify the capability boundary of DeepSeek models through performance tier classifications and intuitive line charts. Specific examples provide actionable insights to help users select and deploy the most cost-effective DeepSeek models, ensuring optimal performance and resource efficiency in real-world applications.

Large language models (LLMs) have demonstrated remarkable capabilities in code-related tasks, such as code understanding and code generation. However, an equally important yet underexplored question is whether LLMs can serve as general-purpose surrogate code executors, to predict the output and behavior of a program without actually running it. To systematically investigate this capability, we introduce SURGE, a comprehensive benchmark covering eight key aspects: multi-language programming tasks, competition-level programming problems, repository-level code analysis, high-cost scientific computing, time-complexity-intensive algorithms, buggy code analysis, programs dependent on specific compilers or execution environments, and formal mathematical proof verification. We evaluate multiple open-source and proprietary LLMs on SURGE and conduct a scaling study to analyze the impact of model size and training data scale on surrogate execution accuracy. Additionally, we categorize model prediction errors and explore potential areas for improvement. Our findings indicate that while LLMs can predict code execution results in certain cases, they exhibit limitations in general-purpose surrogate execution. This study provides empirical insights into the feasibility of using LLMs as surrogate code executors. Code and dataset are released at https://github.com/Imbernoulli/SURGE.

Transformer has attracted increasing interest in STVG, owing to its end-to-end pipeline and promising result. Existing Transformer-based STVG approaches often leverage a set of object queries, which are initialized simply using zeros and then gradually learn target position information via iterative interactions with multimodal features, for spatial and temporal localization. Despite simplicity, these zero object queries, due to lacking target-specific cues, are hard to learn discriminative target information from interactions with multimodal features in complicated scenarios (\e.g., with distractors or occlusion), resulting in degradation. Addressing this, we introduce a novel Target-Aware Transformer for STVG (TA-STVG), which seeks to adaptively generate object queries via exploring target-specific cues from the given video-text pair, for improving STVG. The key lies in two simple yet effective modules, comprising text-guided temporal sampling (TTS) and attribute-aware spatial activation (ASA), working in a cascade. The former focuses on selecting target-relevant temporal cues from a video utilizing holistic text information, while the latter aims at further exploiting the fine-grained visual attribute information of the object from previous target-aware temporal cues, which is applied for object query initialization. Compared to existing methods leveraging zero-initialized queries, object queries in our TA-STVG, directly generated from a given video-text pair, naturally carry target-specific cues, making them adaptive and better interact with multimodal features for learning more discriminative information to improve STVG. In our experiments on three benchmarks, TA-STVG achieves state-of-the-art performance and significantly outperforms the baseline, validating its efficacy.

Recently, Long Chain-of-Thoughts (CoTs) have gained widespread attention for improving the reasoning capabilities of Large Language Models (LLMs). This necessitates that existing LLMs, which lack the ability to generate Long CoTs, to acquire such capability through post-training methods. Without additional training, LLMs typically enhance their mathematical reasoning abilities through inference scaling methods such as MCTS. However, they are hindered by the large action space and inefficient search strategies, making it challenging to generate Long CoTs effectively. To tackle this issue, we propose constraining the action space and guiding the emergence of Long CoTs through a refined search strategy. In our proposed Constrained Monte Carlo Tree Search (C-MCTS) framework, we limit the actions selected from a constrained action space, which is divided into five disjoint subsets: \emph{understanding}, \emph{planning}, \emph{reflection}, \emph{coding}, and \emph{summary}. Each subset is further constrained to a small number of predefined prompts, rather than allowing LLMs to generate actions arbitrarily. Additionally, we refine the search strategy by incorporating prior knowledge about the action sets, such as a human-like partial order of the action subsets and the pretrained process reward models. These strategies work together to significantly reduce the vast search space of Long CoTs. Extensive evaluations on mathematical reasoning benchmarks show that, under zero-shot settings, our method enables the 7B model to achieve reasoning capabilities that surpass those of the 72B model.

Multilingual Retrieval-Augmented Generation (mRAG) systems enhance language models by integrating external multilingual information to produce context-aware responses. However, mRAG systems struggle with retrieving relevant information due to linguistic variations between queries and documents, generating inconsistent responses when multilingual sources conflict. In this work, we systematically investigate language preferences in both retrieval and generation of mRAG through a series of experiments. Our analysis indicates that retrievers tend to prefer high-resource and query languages, yet this preference does not consistently improve generation performance. Moreover, we observe that generators prefer the query language or Latin scripts, leading to inconsistent outputs. To overcome these issues, we propose Dual Knowledge Multilingual RAG (DKM-RAG), a simple yet effective framework that fuses translated multilingual passages with complementary model knowledge. Empirical results demonstrate that DKM-RAG mitigates language preference in generation and enhances performance across diverse linguistic settings.

Modern large language models (LLMs) employ various forms of logical inference, both implicitly and explicitly, when addressing reasoning tasks. Understanding how to optimally leverage these inference paradigms is critical for advancing LLMs' reasoning capabilities. This paper adopts an exploratory approach by introducing a controlled evaluation environment for analogical reasoning -- a fundamental cognitive task -- that is systematically parameterized across three dimensions: modality (textual, visual, symbolic), difficulty (easy, medium, hard), and task format (multiple-choice or free-text generation). We analyze the comparative dynamics of inductive, abductive, and deductive inference pipelines across these dimensions, and demonstrate that our findings generalize to broader in-context learning tasks. Additionally, we investigate advanced paradigms such as hypothesis selection, verification, and refinement, revealing their potential to scale up logical inference in LLM reasoning. This exploratory study provides a foundation for future research in enhancing LLM reasoning through systematic logical inference strategies.

In specialized fields like the scientific domain, constructing large-scale human-annotated datasets poses a significant challenge due to the need for domain expertise. Recent methods have employed large language models to generate synthetic queries, which serve as proxies for actual user queries. However, they lack control over the content generated, often resulting in incomplete coverage of academic concepts in documents. We introduce Concept Coverage-based Query set Generation (CCQGen) framework, designed to generate a set of queries with comprehensive coverage of the document's concepts. A key distinction of CCQGen is that it adaptively adjusts the generation process based on the previously generated queries. We identify concepts not sufficiently covered by previous queries, and leverage them as conditions for subsequent query generation. This approach guides each new query to complement the previous ones, aiding in a thorough understanding of the document. Extensive experiments demonstrate that CCQGen significantly enhances query quality and retrieval performance.

Recent advancements in tree search algorithms guided by verifiers have significantly enhanced the reasoning capabilities of large language models (LLMs), but at the cost of increased computational resources. In this work, we identify two key challenges contributing to this inefficiency: $\textit{over-exploration}$ due to redundant states with semantically equivalent content, and $\textit{under-exploration}$ caused by high variance in verifier scoring leading to frequent trajectory switching. To address these issues, we propose FETCH, an e$\textbf{f}$fici$\textbf{e}$nt $\textbf{t}$ree sear$\textbf{ch}$ framework, which is a flexible, plug-and-play system compatible with various tree search algorithms. Our framework mitigates over-exploration by merging semantically similar states using agglomerative clustering of text embeddings obtained from a fine-tuned SimCSE model. To tackle under-exploration, we enhance verifiers by incorporating temporal difference learning with adjusted $\lambda$-returns during training to reduce variance, and employing a verifier ensemble to aggregate scores during inference. Experiments on GSM8K, GSM-Plus, and MATH datasets demonstrate that our methods significantly improve reasoning accuracy and computational efficiency across four different tree search algorithms, paving the way for more practical applications of LLM-based reasoning. The code will be released upon acceptance.