This study positions large language models (LLMs) as "subjective literary critics" to explore aesthetic preferences and evaluation patterns in literary assessment. Ten Japanese science fiction short stories were translated into English and evaluated by six state-of-the-art LLMs across seven independent sessions. Principal component analysis and clustering techniques revealed significant variations in evaluation consistency (α ranging from 1.00 to 0.35) and five distinct evaluation patterns. Additionally, evaluation variance across stories differed by up to 4.5-fold, with TF-IDF analysis confirming distinctive evaluation vocabularies for each model. Our seven-session within-day protocol using an original Science Fiction corpus strategically minimizes external biases, allowing us to observe implicit value systems shaped by RLHF and their influence on literary judgment. These findings suggest that LLMs may possess individual evaluation characteristics similar to human critical schools, rather than functioning as neutral benchmarkers.

Recent studies suggest that large language models (LLMs) possess the capability to solve graph reasoning tasks. Notably, even when graph structures are embedded within textual descriptions, LLMs can still effectively answer related questions. This raises a fundamental question: How can a decoder-only Transformer architecture understand underlying graph structures? To address this, we start with the substructure extraction task, interpreting the inner mechanisms inside the transformers and analyzing the impact of the input queries. Specifically, through both empirical results and theoretical analysis, we present Induced Substructure Filtration (ISF), a perspective that captures the substructure identification in the multi-layer transformers. We further validate the ISF process in LLMs, revealing consistent internal dynamics across layers. Building on these insights, we explore the broader capabilities of Transformers in handling diverse graph types. Specifically, we introduce the concept of thinking in substructures to efficiently extract complex composite patterns, and demonstrate that decoder-only Transformers can successfully extract substructures from attributed graphs, such as molecular graphs. Together, our findings offer a new insight on how sequence-based Transformers perform the substructure extraction task over graph data.

We address the task of zero-shot video classification for extremely fine-grained actions (e.g., Windmill Dunk in basketball), where no video examples or temporal annotations are available for unseen classes. While image-language models (e.g., CLIP, SigLIP) show strong open-set recognition, they lack temporal modeling needed for video understanding. We propose ActAlign, a truly zero-shot, training-free method that formulates video classification as a sequence alignment problem, preserving the generalization strength of pretrained image-language models. For each class, a large language model (LLM) generates an ordered sequence of sub-actions, which we align with video frames using Dynamic Time Warping (DTW) in a shared embedding space. Without any video-text supervision or fine-tuning, ActAlign achieves 30.5% accuracy on ActionAtlas--the most diverse benchmark of fine-grained actions across multiple sports--where human performance is only 61.6%. ActAlign outperforms billion-parameter video-language models while using 8x fewer parameters. Our approach is model-agnostic and domain-general, demonstrating that structured language priors combined with classical alignment methods can unlock the open-set recognition potential of image-language models for fine-grained video understanding.

Recent advancements in large language models (LLMs) have significantly advanced complex reasoning capabilities, particularly through extended chain-of-thought (CoT) reasoning that incorporates mechanisms such as backtracking, self-reflection, and self-correction. Despite these developments, the self-correction abilities of LLMs during long CoT reasoning remain underexplored. And recent findings on overthinking suggest that such models often engage in unnecessarily redundant reasoning. In this work, we empirically show that the first reasoning step exerts a disproportionately large influence on the final prediction. I.e., errors introduced at this stage can substantially degrade subsequent reasoning quality. This phenomenon is consistently observed across various state-of-the-art open- and closed-source reasoning models. Leveraging this insight, we propose an efficient sampling strategy that leverages a reward model to identify and retain high-quality first reasoning steps while discarding suboptimal ones, achieving up to a 70% reduction in inference cost without sacrificing any accuracy. Our work highlights the central role of the first reasoning step in generating a high-quality reasoning trajectory, and thus enabling significantly efficient sampling.

Quantization has emerged as an effective and lightweight solution to reduce the memory footprint of the KV cache in Large Language Models. Nevertheless, minimizing the accuracy degradation caused by ultra-low-bit KV cache quantization remains a significant challenge. While scalar quantization is constrained by 1-bit bound, vector quantization exploits intra-vector correlations and enables sub-bit regimes, making it more suitable for ultra-low-bit quantization. To further mitigate quantization-induced degradation, we reveal that the degradation is highly uneven across tokens in attention quality. To investigate this unevenness, we introduce anchor score to measure each token's sensitivity to quantization. Our analysis and experiments show that preserving a small subset (1\%) of tokens with the highest Anchor Score significantly mitigates accuracy loss under aggressive quantization. We propose AnTKV, a dual-stage framework that leverages anchor token-aware vector quantization to compress the KV cache. It combines offline token-aware centroids learning and online anchor token selection to balance compression and accuracy. To enable efficient deployment, we design an online anchor token selection kernel compatible with FlashAttention. It allows LLaMA3-8B to scale to 840K tokens on a single 80GB A100, while delivering up to $3.5\times$ higher decoding throughput over the FP16 baseline. Experiments demonstrate that AnTKV matches or surpasses prior methods at 4-bit, and significantly reduce perplexity under ultra-low-bit quantization, achieving 6.32 at 1-bit on Mistral-7B, compared to 7.25 for CQ and 15.36 for KVQuant.

The advent of artificial intelligence has propelled AI-Generated Game Commentary (AI-GGC) into a rapidly expanding field, offering benefits such as unlimited availability and personalized narration. However, current researches in this area remain fragmented, and a comprehensive survey that systematically unifies existing efforts is still missing. To bridge this gap, our survey introduces a unified framework that systematically organizes the AI-GGC landscape. We present a novel taxonomy focused on three core commentator capabilities: Live Observation, Strategic Analysis, and Historical Recall. Commentary is further categorized into three functional types: Descriptive, Analytical, and Background. Building on this structure, we provide an in-depth review of state-of-the-art methods, datasets, and evaluation metrics across various game genres. Finally, we highlight key challenges such as real-time reasoning, multimodal integration, and evaluation bottlenecks, and outline promising directions for future research and system development in AI-GGC.

The integration of vision-language models (VLMs) is driving a new generation of embodied agents capable of operating in human-centered environments. However, as deployment expands, these systems face growing safety risks, particularly when executing hazardous instructions. Current safety evaluation benchmarks remain limited: they cover only narrow scopes of hazards and focus primarily on final outcomes, neglecting the agent's full perception-planning-execution process and thereby obscuring critical failure modes. Therefore, we present SAFE, a benchmark for systematically assessing the safety of embodied VLM agents on hazardous instructions. SAFE comprises three components: SAFE-THOR, an extensible adversarial simulation sandbox with a universal adapter that maps high-level VLM outputs to low-level embodied controls, supporting diverse agent workflow integration; SAFE-VERSE, a risk-aware task suite inspired by Asimov's Three Laws of Robotics, comprising 45 adversarial scenarios, 1,350 hazardous tasks, and 9,900 instructions that span risks to humans, environments, and agents; and SAFE-DIAGNOSE, a multi-level and fine-grained evaluation protocol measuring agent performance across perception, planning, and execution. Applying SAFE to nine state-of-the-art VLMs and two embodied agent workflows, we uncover systematic failures in translating hazard recognition into safe planning and execution. Our findings reveal fundamental limitations in current safety alignment and demonstrate the necessity of a comprehensive, multi-stage evaluation for developing safer embodied intelligence.

Training large language models (LLMs) with chain-of-thought (CoT) supervision has proven effective for enhancing their reasoning abilities. However, obtaining reliable and accurate reasoning supervision remains a significant challenge. We propose a scalable method for generating a high-quality CoT supervision dataset by leveraging the determinism of program execution. Unlike existing reasoning dataset generation methods that rely on costly human annotations or error-prone LLM-generated CoT, our approach extracts verifiable, step-by-step reasoning traces from code execution and transforms them into a natural language CoT reasoning. Experiments on reasoning benchmarks across various domains show that our method effectively equips LLMs with transferable reasoning abilities across diverse tasks. Furthermore, the ablation studies validate that our method produces highly accurate reasoning data and reduces overall token length during inference by reducing meaningless repetition and overthinking.

Retrieval-Augmented Generation (RAG) improves factual accuracy by grounding responses in external knowledge. However, existing RAG methods either rely solely on text corpora and neglect structural knowledge, or build ad-hoc knowledge graphs (KGs) at high cost and low reliability. To address these issues, we propose KG-Infused RAG, a framework that incorporates pre-existing large-scale KGs into RAG and applies spreading activation to enhance both retrieval and generation. KG-Infused RAG directly performs spreading activation over external KGs to retrieve relevant structured knowledge, which is then used to expand queries and integrated with corpus passages, enabling interpretable and semantically grounded multi-source retrieval. We further improve KG-Infused RAG through preference learning on sampled key stages of the pipeline. Experiments on five QA benchmarks show that KG-Infused RAG consistently outperforms vanilla RAG (by 3.9% to 17.8%). Compared with KG-based approaches such as GraphRAG and LightRAG, our method obtains structured knowledge at lower cost while achieving superior performance. Additionally, integrating KG-Infused RAG with Self-RAG and DeepNote yields further gains, demonstrating its effectiveness and versatility as a plug-and-play enhancement module for corpus-based RAG methods.

Graphical User Interface (GUI) agents show promising capabilities for automating computer-use tasks and facilitating accessibility, but existing interactive benchmarks are mostly English-only, covering web-use or Windows, Linux, and Android environments, but not macOS. macOS is a major OS with distinctive GUI patterns and exclusive applications. To bridge the gaps, we present macOSWorld, the first comprehensive benchmark for evaluating GUI agents on macOS. macOSWorld features 202 multilingual interactive tasks across 30 applications (28 macOS-exclusive), with task instructions and OS interfaces offered in 5 languages (English, Chinese, Arabic, Japanese, and Russian). As GUI agents are shown to be vulnerable to deception attacks, macOSWorld also includes a dedicated safety benchmarking subset. Our evaluation on six GUI agents reveals a dramatic gap: proprietary computer-use agents lead at above 30% success rate, while open-source lightweight research models lag at below 5\%, highlighting the need for macOS domain adaptation. Multilingual benchmarks also expose common weaknesses, especially in Arabic, with a 28.8% average degradation compared to English. Results from safety benchmarking also highlight that deception attacks are more general and demand immediate attention. Project page: https://macos-world.github.io.