Code editing, including modifying, refactoring, and maintaining existing code, is the most frequent task in software development and has garnered significant attention from AI-powered tools. However, existing solutions that translate explicit natural language instructions into code edits face critical limitations, such as heavy reliance on human instruction input and high latency, which hinder their effective integration into a developer's workflow. We observe that developers' habitual behaviors and coding objectives are often reflected in their historical editing patterns, making this data key to addressing existing limitations. To leverage these insights, we propose NES (Next Edit Suggestion), an LLM-driven code editing framework that delivers an instruction-free and low-latency experience. Built on a dual-model architecture and trained with our high-quality SFT and DAPO datasets, NES enhances productivity by understanding developer intent while optimizing inference to minimize latency. NES is a scalable, industry-ready solution with a continuous Tab key interaction workflow, seamlessly adopted by a FinTech company with over 20,000 developers. Evaluations on real-world datasets show NES achieves 75.6% and 81.6% accuracy in two tasks of predicting next edit locations, alongside 91.36% ES and 27.7% EMR for intent-aligned edits, outperforming SOTA models. Our open-sourced SFT and DAPO datasets have been demonstrated to enhance the performance of open-source CodeLLMs. The demonstration of NES is available at https://youtu.be/yGoyYOe6fbY.

Token-level code completion is one of the most critical features in modern Integrated Development Environments (IDEs). It assists developers by suggesting relevant identifiers and APIs during coding. While completions are typically derived from static analysis, their usefulness depends heavily on how they are ranked, as correct predictions buried deep in the list are rarely seen by users. Most current systems rely on hand-crafted heuristics or lightweight machine learning models trained on user logs, which can be further improved to capture context information and generalize across projects and coding styles. In this work, we propose a new scoring approach to ranking static completions using language models in a lightweight and model-agnostic way. Our method organizes all valid completions into a prefix tree and performs a single greedy decoding pass to collect token-level scores across the tree. This enables a precise token-aware ranking without needing beam search, prompt engineering, or model adaptations. The approach is fast, architecture-agnostic, and compatible with already deployed models for code completion. These findings highlight a practical and effective pathway for integrating language models into already existing tools within IDEs, and ultimately providing smarter and more responsive developer assistance.

Recent advances have shown that optimizing prompts for Large Language Models (LLMs) can significantly improve task performance, yet many optimization techniques rely on heuristics or manual exploration. We present LatentPrompt, a model-agnostic framework for prompt optimization that leverages latent semantic space to automatically generate, evaluate, and refine candidate prompts without requiring hand-crafted rules. Beginning with a set of seed prompts, our method embeds them in a continuous latent space and systematically explores this space to identify prompts that maximize task-specific performance. In a proof-of-concept study on the Financial PhraseBank sentiment classification benchmark, LatentPrompt increased classification accuracy by approximately 3 percent after a single optimization cycle. The framework is broadly applicable, requiring only black-box access to an LLM and an automatic evaluation metric, making it suitable for diverse domains and tasks.

This study investigates the reliability and validity of five advanced Large Language Models (LLMs)--Claude 3.5, DeepSeek v2, Gemini 2.5, GPT 4, and Mistral 24B--for automated essay scoring in a real-world higher education context. A total of 67 Italian-language student essays, written as part of a university psychology course, were evaluated using a four-criterion rubric (Pertinence, Coherence, Originality, Feasibility). Each model scored all essays across three prompt replications to assess intra-model stability. Human-LLM agreement was consistently low and non-significant (Quadratic Weighted Kappa), and within-model reliability across replications was similarly weak (median Kendall's W < .30). Systematic scoring divergences emerged, including a tendency to inflate Coherence and inconsistent handling of context-dependent dimensions. Inter-model agreement analysis revealed moderate convergence for Coherence and Originality, but negligible concordance for Pertinence and Feasibility. Although limited in scope, these findings suggest that current LLMs may struggle to replicate human judgment in tasks requiring disciplinary insight and contextual sensitivity. Human oversight remains critical when evaluating open-ended academic work, particularly in interpretive domains.

Measuring innovation often relies on context-specific proxies and on expert evaluation. Hence, empirical innovation research is often limited to settings where such data is available. We investigate how large language models (LLMs) can be leveraged to overcome the constraints of manual expert evaluations and assist researchers in measuring innovation. We design an LLM framework that reliably approximates domain experts' assessment of innovation from unstructured text data. We demonstrate the performance and broad applicability of this framework through two studies in different contexts: (1) the innovativeness of software application updates and (2) the originality of user-generated feedback and improvement ideas in product reviews. We compared the performance (F1-score) and reliability (consistency rate) of our LLM framework against alternative measures used in prior innovation studies, and to state-of-the-art machine learning- and deep learning-based models. The LLM framework achieved higher F1-scores than the other approaches, and its results are highly consistent (i.e., results do not change across runs). This article equips R&D personnel in firms, as well as researchers, reviewers, and editors, with the knowledge and tools to effectively use LLMs for measuring innovation and evaluating the performance of LLM-based innovation measures. In doing so, we discuss, the impact of important design decisions-including model selection, prompt engineering, training data size, training data distribution, and parameter settings-on performance and reliability. Given the challenges inherent in using human expert evaluation and existing text-based measures, our framework has important implications for harnessing LLMs as reliable, increasingly accessible, and broadly applicable research tools for measuring innovation.

--Multimodal Affective Computing (MAC) aims to recognize and interpret human emotions by integrating information from diverse modalities such as text, video, and audio. Recent advancements in Multimodal Large Language Models (MLLMs) have significantly reshaped the landscape of MAC by offering a unified framework for processing and aligning cross-modal information. However, practical challenges remain, including performance variability across complex MAC tasks and insufficient understanding of how architectural designs and data characteristics impact affective analysis. T o address these gaps, we conduct a systematic benchmark evaluation of state-of-the-art open-source MLLMs capable of concurrently processing audio, visual, and textual modalities across multiple established MAC datasets. Our evaluation not only compares the performance of these MLLMs but also provides actionable insights into model optimization by analyzing the influence of model architectures and dataset properties. Furthermore, we propose a novel hybrid strategy that combines generative knowledge prompting with supervised fine-tuning to enhance MLLMs' affective computing capabilities. Experimental results demonstrate that this integrated approach significantly improves performance across various MAC tasks, offering a promising avenue for future research and development in this field. Multimodal Affective Computing (MAC) aims to recognize, perceive, infer, and interpret human emotions through the integration of information from multiple modalities, including text, video, and audio [1]. Human emotional expressions are inherently complex and multimodal in nature [2], a characteristic that makes unimodal approaches particularly vulnerable to ambiguity, noise interference, and information loss [3].

Large vision-language models (LVLMs) have demonstrated remarkable multimodal comprehension and reasoning capabilities, but they still suffer from severe object hallucination. Previous studies primarily attribute the flaw to linguistic prior caused by the scale mismatch between visual encoders and large language models (LLMs) in LVLMs. Specifically, as current LVLMs are built upon LLMs, they tend to over-rely on textual prompts and internal knowledge of LLMs, generating descriptions inconsistent with visual cues. However, through an in-depth investigation of the hallucinated mechanisms, we empirically reveal a previously overlooked phenomenon: LVLMs may ignore not only visual information but also textual modality during hallucination, a behavior termed as modality bias, which indicates that LVLMs struggle to simultaneously attend to both visual and textual modalities, leading to fragmented understanding of user-provided instructions. Based on this observation, we propose a simple yet effective training-free method to mitigate object hallucination. Concretely, we intervene and adjust the attention weights of textual and visual tokens, balancing cross-modal compatibility for better alignment with user intentions. Furthermore, we adopt a contrastive decoding strategy to reduce the LVLM's overreliance on its parametric knowledge, synergistically enhancing our attention manipulation. Extensive experiments confirm the widespread presence of modality bias in LVLMs. Notably, our method effectively mitigates hallucination across multiple open-source LVLMs and benchmarks, highlighting its generalizability and efficacy.

Recent advances in large language models (LLMs) have significantly boosted long-context processing. However, the increasing key-value (KV) cache size poses critical challenges to memory and execution efficiency. Most KV cache compression methods rely on heuristic token eviction using all attention heads in Grouped Query Attention (GQA)-based LLMs. This method ignores the different functionalities of attention heads, leading to the eviction of critical tokens and thus degrades the performance of LLMs. To address the issue above, instead of using all the attention heads in GQA-based LLMs to determine important tokens as in the previous work, we first identify the attention heads in each layer that are not only capable of retrieving the initial and final tokens of a prompt, but also capable of retrieving important tokens within the text and attending to their surrounding semantic context. Afterwards, we exploit such heads to determine the important tokens and retain their corresponding KV cache pairs. Furthermore, we analyze the cache eviction error of each layer individually and introduce a layer-adaptive KV cache allocation strategy. Experimental results demonstrate the proposed CompressKV consistently outperforms state-of-the-art approaches under various memory budgets on LongBench and Needle-in-a-Haystack benchmarks. Our code is publicly available at: https://github.com/TUDa-HWAI/CompressKV.git.

Quantization significantly accelerates inference in large language models (LLMs) by replacing original high-precision matrices with low-precision counterparts. Recent advances in weight-activation quantization have primarily focused on mapping both weights and activations to the INT4 format. Although the new FP4 Tensor Cores in NVIDIA's Blackwell architecture offer up to 4x speedup over FP16, existing INT4-based kernels fail to fully exploit this capability due to mismatched data formats. To bridge this gap, we propose MicroMix, a co-designed mixed-precision quantization algorithm and matrix multiplication kernel based on Microscaling (MX) data formats. Tailored for the Blackwell architecture, the MicroMix kernel supports arbitrary combinations of MXFP4, MXFP6, and MXFP8 channels, and produces BFloat16 outputs. To achieve a favorable trade-off between accuracy and efficiency for each linear layer, we introduce quantization thresholds that identify activation elements where lower-precision formats (MXFP4 or MXFP6) incur excessive quantization error. Our algorithm selectively allocates higher-precision channels to preserve accuracy while maintaining compute efficiency. MicroMix achieves competitive or superior performance across diverse downstream tasks, including zero-shot and few-shot learning, language modeling, code generation, and mathematical reasoning. On both consumer-grade (RTX 5070Ti laptop) and server-grade (RTX 5090) GPUs, our kernel delivers at least 20% faster execution than TensorRT-FP8. Furthermore, when applied to various Llama and Qwen models, MicroMix consistently improves prefill latency and memory efficiency across a range of batch sizes compared to TensorRT baselines. Our code is available at https://github.com/lwy2020/MicroMix.

Autonomous Mobile Robots (AMRs) are deployed in diverse environments (e.g., warehouses, retail spaces, and offices), where they work alongside humans. Given that human behavior can be unpredictable and that AMRs may not have been trained to handle all possible unknown and uncertain behaviors, it is important to test AMRs under a wide range of human interactions to ensure their safe behavior. Moreover, testing in real environments with actual AMRs and humans is often costly, impractical, and potentially hazardous (e.g., it could result in human injury). To this end, we propose a Vision Language Model (VLM)-based testing approach (RVSG) for industrial AMRs developed by PAL Robotics in Spain. Based on the functional and safety requirements, RVSG uses the VLM to generate diverse human behaviors that violate these requirements. We evaluated RVSG with several requirements and navigation routes in a simulator using the latest AMR from PAL Robotics. Our results show that, compared with the baseline, RVSG can effectively generate requirement-violating scenarios. Moreover, RVSG-generated scenarios increase variability in robot behavior, thereby helping reveal their uncertain behaviors.