Archive, newest first. Filter by topic, impact and company →
Problem — The paper addresses the gap in existing literature on hateful video detection, which predominantly focuses on binary classification without providing contextual rationales for decisions. This lack of explainability…
Problem This work addresses the gap in real-time monitoring and adaptation of deployed safety classifiers in the presence of distributional shifts, a critical issue for maintaining classifier reliability in dynamic…
Problem The paper addresses the gap in the integration of neural computation with relational databases, specifically the limitations of existing methods that either rely on graph representations or treat embeddings…
Problem This work addresses the challenge of multi-turn retrieval and question answering in conversational systems, particularly in the context of the SemEval-2026 Task 8. The task evaluates systems across diverse…
Problem This work addresses the gap in educational NLP for the Bangla language, which is underrepresented in automated assessment research. The authors highlight the challenges faced in low-resource settings, where…
Problem This work addresses the challenge of predicting psychological traits from asynchronous video interviews (AVIs), a task complicated by the limited availability of labeled datasets and the high-dimensional nature of…
Problem This work addresses the gap in autonomous research capabilities, specifically the lack of frameworks that can effectively manage long-term research processes without human intervention. The authors propose a novel…
Problem The paper addresses the scarcity of high-quality parallel sign video-text pairs for fine-tuning sign language translation (SLT) models, which hampers generalization on long-tail vocabulary and unseen constructions. Despite advancements…
Problem Recent advancements in respiratory sound classification (RSC) have predominantly utilized CLS-token driven self-attention architectures, such as the Audio Spectrogram Transformer (AST). However, these models exhibit a low-pass filtering behavior…
Problem Current Large Reasoning Models (LRMs) demonstrate significant deficiencies in spatial reasoning tasks, which are often attributed to a lack of knowledge. Existing methodologies primarily rely on supervised fine-tuning (SFT)…
Problem This work addresses a gap in the literature regarding software aging in GPU-based large language model (LLM) serving systems. Traditional studies have predominantly focused on CPU-centric software, which typically…
Problem This work addresses the challenge of respiratory sound classification, particularly focusing on the limitations of existing methods in handling varying audio quality and class imbalance. The authors highlight a…
Problem The paper addresses the inefficiencies in CSI-based localization within vehicular IoT networks, particularly the challenge of transmitting channel state information (CSI) from spatially distributed antenna arrays (RAAs) to a…
Problem The paper addresses the gap in the integration of statistical tools for hypothesis testing with topological data analysis, specifically focusing on persistence diagrams (PDs). While PDs are widely used…
Problem The paper addresses a significant gap in the capability of existing retrieval-augmented generation methods for traffic law liability determination, particularly the inability to effectively manage multi-dimensional legal queries. Current…
Problem Existing benchmarks for embodied spatial intelligence are static, labor-intensive to create, and often fail to adapt as models evolve, leading to saturation in distinguishing new capabilities. This paper addresses…
Problem This work addresses the gap in understanding the robustness of Vision-Language-Action (VLA) models to linguistic variation, particularly in non-English contexts. Despite their strong performance in language-conditioned robotic manipulation, the…
Problem This paper addresses the limitations of existing methods that integrate Large Language Models (LLMs) into Text-Attributed Graphs (TAGs), particularly their inability to generalize effectively across diverse graphs and tasks.…
Problem This work addresses the underexplored area of informal laboratory notes in AI-driven scientific discovery workflows. Prior research predominantly focuses on structured data sources like publications and protocols, neglecting the…
Problem The paper addresses the insufficient characterization of the alignment between large language models (LLMs) and the neural mechanisms of human higher-order cognition, particularly in the context of deductive reasoning.…
Problem The paper addresses a gap in the literature regarding the effectiveness of different critic architectures in multi-objective reinforcement learning (RL) for humanoid robots. Specifically, it explores whether a unified…
Problem The paper addresses the lack of computational approaches to emotional validation in dialogue systems, a critical aspect of providing deeper emotional support. Despite its therapeutic value, emotional validation has…
Problem This work addresses the limitations of existing Parameter-Efficient Fine-Tuning (PEFT) techniques for Large Language Models (LLMs), specifically Low-Rank Adaptation (LoRA) and Soft Prompting. Both methods necessitate modifications to the…
Problem This work addresses a critical gap in the security of Large Language Models (LLMs) used for code generation, particularly the unintended consequences of employing Grammar-Constrained Decoding (GCD). While GCD…
Problem — The paper addresses the lack of robust evaluation frameworks for assessing language model agents' ability to forecast real-world events with valid reasoning. Existing methods often rely solely on…
Problem This work addresses the gap in understanding how external operational experience can be effectively utilized in production LLM systems, particularly focusing on the trade-offs between quality and cost during…
Problem Video Large Multimodal Models (VLMMs) have shown significant advancements in video understanding tasks; however, they are still susceptible to hallucinations—instances where the generated outputs do not accurately reflect the…
Problem This work addresses the performance degradation of Large Language Models (LLMs) when applied to low-resource languages, specifically Kupang Malay. The authors highlight a gap in the literature regarding effective…
Problem The paper addresses the gap in training efficiency during the distillation of large speech foundation models (SFMs) into smaller, more efficient student models. While distillation is beneficial for low-resource…
Problem The paper addresses the lack of systematic and scalable methods for evaluating the creativity of large language models (LLMs) across various open-ended tasks. Existing creativity metrics are often tied…
