We propose Semantic F1 Scores, novel evaluation metrics for subjective or fuzzy multi-label classification that quantify semantic relatedness between predicted and gold labels. Unlike the conventional F1 metrics that treat semantically related predictions as complete failures, Semantic F1 incorporates a label similarity matrix to compute soft precision-like and recall-like scores, from which the Semantic F1 scores are derived. Unlike existing similarity-based metrics, our novel two-step precision-recall formulation enables the comparison of label sets of arbitrary sizes without discarding labels or forcing matches between dissimilar labels. By granting partial credit for semantically related but nonidentical labels, Semantic F1 better reflects the realities of domains marked by human disagreement or fuzzy category boundaries. In this way, it provides fairer evaluations: it recognizes that categories overlap, that annotators disagree, and that downstream decisions based on similar predictions lead to similar outcomes. Through theoretical justification and extensive empirical validation on synthetic and real data, we show that Semantic F1 demonstrates greater interpretability and ecological validity. Because it requires only a domain-appropriate similarity matrix, which is robust to misspecification, and not a rigid ontology, it is applicable across tasks and modalities.

Language Models (LMs) have significantly advanced natural language processing and enabled remarkable progress across diverse domains, yet their black-box nature raises critical concerns about the interpretability of their internal mechanisms and decision-making processes. This lack of transparency is particularly problematic for adoption in high-stakes domains, where stakeholders need to understand the rationale behind model outputs to ensure accountability. On the other hand, while explainable artificial intelligence (XAI) methods have been well studied for non-LMs, they face many limitations when applied to LMs due to their complex architectures, considerable training corpora, and broad generalization abilities. Although various surveys have examined XAI in the context of LMs, they often fail to capture the distinct challenges arising from the architectural diversity and evolving capabilities of these models. To bridge this gap, this survey presents a comprehensive review of XAI techniques with a particular emphasis on LMs, organizing them according to their underlying transformer architectures: encoder-only, decoder-only, and encoder-decoder, and analyzing how methods are adapted to each while assessing their respective strengths and limitations. Furthermore, we evaluate these techniques through the dual lenses of plausibility and faithfulness, offering a structured perspective on their effectiveness. Finally, we identify open research challenges and outline promising future directions, aiming to guide ongoing efforts toward the development of robust, transparent, and interpretable XAI methods for LMs.

Program verification relies on loop invariants, yet automatically discovering strong invariants remains a long-standing challenge. We introduce a principled framework for evaluating LLMs on invariant synthesis. Our approach uses a verifier-based decision procedure with a formal soundness guarantee and assesses not only correctness but also the speedup that invariants provide in verification. We evaluate 7 state-of-the-art LLMs, and existing LLM-based verifiers against the traditional solver UAutomizer. While LLM-based verifiers represent a promising direction, they do not yet offer a significant advantage over UAutomizer. Model capability also proves critical, as shown by sharp differences in speedups across models, and our benchmark remains an open challenge for current LLMs. Finally, we show that supervised fine-tuning and Best-of-N sampling can improve performance: fine-tuning on 3589 instances raises the percentage of speedup cases for Qwen3-Coder-480B from 8% to 29.2%, and Best-of-N sampling with N=16 improves Claude-sonnet-4 from 8.8% to 22.1%.

Audio editing plays a central role in VR/AR immersion, virtual conferencing, sound design, and other interactive media. However, recent generative audio editing models depend on template-like instruction formats and are restricted to mono-channel audio. These models fail to deal with declarative audio editing, where the user declares what the desired outcome should be, while leaving the details of editing operations to the system. We introduce SmartDJ, a novel framework for stereo audio editing that combines the reasoning capability of audio language models with the generative power of latent diffusion. Given a high-level instruction, SmartDJ decomposes it into a sequence of atomic edit operations, such as adding, removing, or spatially relocating events. These operations are then executed by a diffusion model trained to manipulate stereo audio. To support this, we design a data synthesis pipeline that produces paired examples of high-level instructions, atomic edit operations, and audios before and after each edit operation. Experiments demonstrate that SmartDJ achieves superior perceptual quality, spatial realism, and semantic alignment compared to prior audio editing methods. Demos are available at https://zitonglan.github.io/project/smartdj/smartdj.html.

The expanding long-context capabilities of large language models are constrained by a significant memory bottleneck: the key-value (KV) cache required for autoregressive generation. This bottleneck is substantial; for instance, a Llama-3.1-8B model processing a 32K-token prompt at a batch size of 4 requires approximately 16GB for its KV cache, a size exceeding the model's weights. While KV-cache compression via low-rank projection is a promising direction, existing methods rely on a static, offline-learned subspace that performs poorly under data distribution shifts. To overcome these limitations, we introduce OjaKV, a novel framework that integrates a strategic hybrid storage policy with online subspace adaptation. First, OjaKV recognizes that not all tokens are equally important for compression; it preserves the crucial first and most recent tokens in full-rank, maintaining high-fidelity anchors for attention. Second, for the vast majority of intermediate tokens, it applies low-rank compression by incrementally adapting the projection basis using Oja's algorithm for online principal component analysis. This adaptation involves a comprehensive update during prompt prefilling and lightweight periodic updates during decoding, ensuring the subspace remains aligned with the evolving context. Crucially, our framework is fully compatible with modern attention modules like FlashAttention. Experiments demonstrate that OjaKV maintains or even improves zero-shot accuracy at high compression ratios. In particular, OjaKV achieves its strongest gains on very long-context benchmarks that require complex reasoning, highlighting the importance of online subspace adaptation in dynamically tracking context shifts. These results establish our hybrid framework as a practical, plug-and-play solution for memory-efficient long-context inference without requiring model fine-tuning.

Multi-Objective Reinforcement Learning (MORL) presents significant challenges and opportunities for optimizing multiple objectives in Large Language Models (LLMs). We introduce a MORL taxonomy and examine the advantages and limitations of various MORL methods when applied to LLM optimization, identifying the need for efficient and flexible approaches that accommodate personalization functionality and inherent complexities in LLMs and RL. We propose a vision for a MORL benchmarking framework that addresses the effects of different methods on diverse objective relationships. As future research directions, we focus on meta-policy MORL development that can improve efficiency and flexibility through its bi-level learning paradigm, highlighting key research questions and potential solutions for improving LLM performance.

Abstract-- Object-level Simultaneous Localization and Mapping (SLAM), which incorporates semantic information for high-level scene understanding, faces challenges of under-constrained optimization due to sparse observations. Prior work has introduced additional constraints using commonsense knowledge, but obtaining such priors has traditionally been labor-intensive and lacks generalizability across diverse object categories. We address this limitation by leveraging large language models (LLMs) to provide commonsense knowledge of object geometric attributes, specifically size and orientation, as prior factors in a graph-based SLAM framework. These priors are particularly beneficial during the initial phase when object observations are limited. We implement a complete pipeline integrating these priors, achieving robust data association on sparse object-level features and enabling real-time object SLAM. Our system, evaluated on the TUM RGB-D and 3RScan datasets, improves mapping accuracy by 36.8% over the latest baseline. Object Simultaneous Localization and Mapping (SLAM) builds environment maps by identifying and localizing objects, and using this information to infer the robot's position. Unlike traditional feature-based SLAM, object-level representations are sparse, focusing on semantic object data. Comparing to semantic segmentation on dense representations, such sparsity improves computational efficiency and reduces storage requirements.

Personalized multi-document summarization (MDS) is essential for meeting individual user preferences of writing style and content focus for summaries. In this paper, we propose that for effective personalization, it is important to identify fine-grained differences between users' preferences by comparing the given user's preferences with other users' preferences.Motivated by this, we propose ComPSum, a personalized MDS framework. It first generates a structured analysis of a user by comparing their preferences with other users' preferences. The generated structured analysis is then used to guide the generation of personalized summaries. To evaluate the performance of ComPSum, we propose AuthorMap, a fine-grained reference-free evaluation framework for personalized MDS. It evaluates the personalization of a system based on the authorship attribution between two personalized summaries generated for different users. For robust evaluation of personalized MDS, we construct PerMSum, a personalized MDS dataset in the review and news domain. We evaluate the performance of ComPSum on PerMSum using AuthorMap, showing that it outperforms strong baselines.

Graphical User Interface (GUI) grounding is commonly framed as a coordinate prediction task -- given a natural language instruction, generate on-screen coordinates for actions such as clicks and keystrokes. However, recent Vision Language Models (VLMs) often fail to predict accurate numeric coordinates when processing high-resolution GUI images with complex layouts. To address this issue, we reframe GUI grounding as an \emph{interactive search task}, where the VLM generates actions to move a cursor in the GUI to locate UI elements. At each step, the model determines the target object, evaluates the spatial relations between the cursor and the target, and moves the cursor closer to the target conditioned on the movement history. In this interactive process, the rendered cursor provides visual feedback to help the model align its predictions with the corresponding on-screen locations. We train our GUI grounding model, GUI-Cursor, using multi-step online reinforcement learning with a dense trajectory-based reward function. Our experimental results show that GUI-Cursor, based on Qwen2.5-VL-7B, improves the GUI grounding accuracy and achieves state-of-the-art results on ScreenSpot-v2 ($88.8\% \rightarrow 93.9\%$) and ScreenSpot-Pro ($26.8\% \rightarrow 56.5\%$). Moreover, we observe that GUI-Cursor learns to solve the problem within two steps for 95\% of instances and can adaptively conduct more steps on more difficult examples.

Questions in political interviews and hearings serve strategic purposes beyond information gathering including advancing partisan narratives and shaping public perceptions. However, these strategic aspects remain understudied due to the lack of large-scale datasets for studying such discourse. Congressional hearings provide an especially rich and tractable site for studying political questioning: Interactions are structured by formal rules, witnesses are obliged to respond, and members with different political affiliations are guaranteed opportunities to ask questions, enabling comparisons of behaviors across the political spectrum. We develop a pipeline to extract question-answer pairs from unstructured hearing transcripts and construct a novel dataset of committee hearings from the 108th--117th Congress. Our analysis reveals systematic differences in questioning strategies across parties, by showing the party affiliation of questioners can be predicted from their questions alone. Our dataset and methods not only advance the study of congressional politics, but also provide a general framework for analyzing question-answering across interview-like settings.