Knowledge base question answering systems rely on entity linking to connect textual mentions in natural language with corresponding entities in a structured knowledge base. While conventional methods perform well in static or generic scenarios, they often struggle in dynamic contexts that require fine-grained disambiguation between specific instances. This limitation is particularly evident in domains such as military situation reports, where accurate identification of individual instances is essential for timely and precise information retrieval. This work introduces a novel framework for instance-level entity linking that enables transformer-based language models to distinguish between similarly described but distinct instances of the same entity. The proposed framework integrates three core innovations: (1) a context-aware masking strategy that reduces over-reliance on surface-level entity mentions by encouraging the model to leverage surrounding contextual cues, (2) a multi-task learning component that improves the model’s understanding of relative spatial and temporal information, and (3) a position-aware representation technique that restructures numerical data into token sequences that preserve structural integrity, addressing inherent limitations of subword tokenization in current language models. Experimental evaluations on simulated Korean military reports demonstrate substantial performance gains, particularly in challenging cases involving ambiguous or overlapping entity descriptions. The enhanced model shows a markedly improved ability to differentiate between instances based on subtle temporal and geospatial cues. These advancements extend beyond entity linking, offering broader insights into how transformer-based models can more effectively interpret numerical information. This research establishes a new direction for precise entity disambiguation in dynamic environments and contributes practical methods for enhancing language model performance across critical domains, including medical diagnostics, financial analysis, scientific research, autonomous systems, and numerical reasoning in smaller Large Language Models, where our techniques can potentially bridge the performance gap with larger models without requiring extensive computational resources.

Short-horizon collision–risk warning can use instantaneous assessment, forecast-then-assess, or direct prediction from encounter histories. We evaluate the third route with the Anticipatory TRAffic-context Collision–risk Transformer (ATRACT), which estimates the maximum near-future fuzzy Collision Risk Index (CRI) directly from own-ship kinematic and collision-geometry histories without trajectory rollout. Experiments use ten days of Automatic Identification System (AIS) data from the Danish straits, comprising 1.63 million decision windows. Using a vessel-day-track partition and five random initializations, ATRACT attains an area under the receiver operating characteristic curve (ROC-AUC) of $0.814\pm 0.003$, compared with $0.808\pm 0.006$ for the evaluated forecast-then-CRI baseline (VCRF) and 0.729 for instantaneous CRI. Thresholds fixed at a nominal 5% false-alarm rate (FAR) on disjoint calibration tracks yield test FARs of 4.98% and 5.38% for ATRACT and VCRF. Although seed-0 track-bootstrap intervals include zero at every evaluated offset, ATRACT shows 4.4–10.8 percentage points higher mean pre-onset detection across 0.5–4 min over five random initializations; VCRF detects more individual high-risk windows at this operating point. Four direct-risk encoders achieve a narrow ROC-AUC range of 0.813–0.819. These results support direct temporal interaction modeling as a viable route while limiting conclusions to the evaluated forecast-first implementations.

Image compression has evolved from human-centric perceptual coding toward support for diverse machine vision applications, requiring modern codecs to serve both human viewing and downstream tasks in closed-set settings (where target tasks are incorporated during training) and open-set settings (where previously unseen tasks arise at test time). While recent learned compression methods jointly optimize perceptual quality and closed-set task performance, they often fail to generalize to unseen open-set tasks due to fixed training assumptions and objectives. Our prior work, LoRA-comp (Low-Rank Adaptation Compression), effectively addresses open-set challenges via instance-specific test-time fine-tuning (TTFT) without requiring task-specific pre-training. Nevertheless, its fixed LoRA architecture, which assigns a uniform rank across all layers, often leads to suboptimal instance-level performance. Moreover, allocating the same number of training epochs to every instance introduces unnecessary encoding-time overhead. To address these challenges, we propose Instance-Specific Early Stopping with Dynamic Rank Adaptation (ISER), which extends LoRA-comp. Building upon the LoRA-comp–based instance-specific adaptation framework, ISER introduces (i) instance-specific early stopping (ISES) combined with a multi-scale training strategy (MSTS) to reduce TTFT overhead and (ii) instance-specific dynamic rank adaptation (ISRA) to tailor the LoRA architecture per instance. Experiments demonstrate that ISER consistently outperforms competing methods. Compared to LoRA-comp, ISER achieves up to a 7% BD-Rate improvement and up to a 44% reduction in encoding time. Moreover, ISER achieves up to a 98% relative improvement in BD-Rate gain and up to a 19.2% reduction in decoding time over TransTIC.