This chapter explores advancements in decoding strategies for large language models (LLMs), focusing on enhancing the Locally Typical Sampling (LTS) algorithm. Traditional decoding methods, such as top-k and nucleus sampling, often struggle to balance fluency, diversity, and coherence in text generation. To address these challenges, Adaptive Semantic-Aware Typicality Sampling (ASTS) is proposed as an improved version of LTS, incorporating dynamic entropy thresholding, multi-objective scoring, and reward-penalty adjustments. ASTS ensures contextually coherent and diverse text generation while maintaining computational efficiency. Its performance is evaluated across multiple benchmarks, including story generation and abstractive summarization, using metrics such as perplexity, MAUVE, and diversity scores. Experimental results demonstrate that ASTS outperforms existing sampling techniques by reducing repetition, enhancing semantic alignment, and improving fluency.

Natural-language-initiated querying is usually framed as translation into a predetermined backend language such as SQL, Cypher, or SPARQL. That framing is appropriate when the semantic target is known in advance, but it does not cover the full space of natural-language query workloads. In document-centric, mixed, and heterogeneous environments, the first semantic problem may be to determine what target should be constructed before backend-specific execution can begin. This paper proposes the NLIQ lens for this broader space. It introduces target adequacy as the criterion for distinguishing settings in which the target is given, only partially specified, or must itself be constructed, and argues that intermediate representations in the latter regimes are not merely implementation devices but first-class semantic objects. The paper develops a compact framework of NLIQ regimes, illustrates the distinction through representative examples, and identifies a new research terrain around semantic target formation, intermediate representation design, heterogeneous compilation, and answer formation in complex data environments.

Natural language (NL) interfaces to databases broaden access to heterogeneous data but often yield many ambiguous intermediate logical plans (ILPs) due to uncertain operator scope and predicate semantics. Many candidates are infeasible because of type mismatches, missing bindings, or engine-specific constraints. We address this challenge with feasibility constraints for detecting local inconsistencies and introduce the Packed Plan Forest (PPF) a polynomially bounded structure that compactly encodes all feasible ILPs while pruning infeasible ones early. Extending packed parse forest ideas to multi-model settings, PPF supports efficient feasibility analysis through annotated operators. Formal results show polynomial size under bounded arity and annotation vocabularies, and experiments confirm that PPFs capture exponentially many ILPs with minimal overhead, establishing a scalable foundation for NL-to-DB query planning across heterogeneous systems

Physiological time series signals reflect complex, multi-scale dynamical processes of the human body. Existing modeling studies focus on static tasks such as classification, event forecasting, or short-horizon next step prediction, while long-horizon signal-level forecasting and predictive nature of physiological signals remain underexplored. We introduce NormWear-2, a world model that encodes both multivariate physiological signals and clinical intervention variables into a shared latent space and models their joint temporal evolution as a dynamical system. Our approach combines inference from prior pre-trained knowledge (intuition) with instant non-parametric latent state transition adaptation (insight), enabling coherent forecasting across multiple temporal scales, conditioned on heterogeneous clinical interventions. During the pretraining phase, we find that chaos-theoretic balancing of dynamical regime diversity yields more robust representations, with a smaller balanced corpus outperforming one twice its size and capturing bifurcation regimes. We evaluate the world model performance across diverse real-world physiological datasets spanning heterogeneous temporal resolutions and intervention regimes, covering daily life, point-of-care, and clinical settings, including fitness planning, hemodialysis, diabetes management, and surgical monitoring. These evaluation datasets comprise records from 8,026 subjects, spanning study durations from 3.2 hours for high-resolution signal data to 2.3 years for longitudinal clinical biomarker tracking. NormWear-2 achieves the best overall forecasting performance across time, frequency, and latent representation domains, with significant improvements over state-of-the-art time series foundation models, while maintaining competitive downstream representation quality, providing a step toward general-purpose world models for physiological signals.

INTRODUCTION:Anemia management in hemodialysis (HD) depends on individualized erythropoiesis-stimulating agent (ESA) dosing to achieve and maintain target hemoglobin (Hgb) levels. However, predicting individual Hgb responses remains a significant challenge due to nonlinear erythropoietic dynamics and patient heterogeneity. We developed a machine learning-based approach to predict individual Hgb response to ESA therapy in HD patients. METHODS:We included adult patients undergoing HD at Sanderling clinics who received intravenous (IV) ESA (Mircera®) and IV elemental iron and had ≥5 qualifying dosing events with associated post-dose Hgb measurements. Clinical, dialysis, and laboratory data were structured into an event-based dataset, where each event included pre-dose Hgb, contemporaneous Mircera® and iron doses, and follow-up Hgb 7-21 days later. Pharmacokinetically informed features, including cumulative and time-weighted Mircera® dosing, were engineered to capture delayed erythropoietic effects. A transformer-based neural network was trained using sequences of five consecutive events. Model performance was assessed using a patient-level train-test split, with mean absolute percentage error (MAPE) as the primary metric. The MAPE was calculated as the average absolute percentage error between the predicted future Hgb and the observed future Hgb. To illustrate clinical utility, we conducted simulation-based forecasting of 30-day Hgb trajectories under varying hypothetical Mircera® doses (0-200 µg) in 4 representative patients. In addition, simulation error was summarized at the cohort level across all eligible patients to assess overall forecasting performance under hypothetical dosing scenarios. RESULTS:Among 728 HD patients, the mean age was 63 ± 14 years, with dialysis vintage 2.1 ± 1.8 years. Of these, 43% were female and 40% had diabetes. Patients received 114 ± 56 mg/week of IV iron and 149 ± 92 µg/month of IV Mircera®. Mean Hgb was 10.1 ± 0.9 g/dL, transferrin saturation (TSAT) 31 ± 10%, and ferritin level 703 ± 452 ng/mL. The transformer model predicted future Hgb with a MAPE of ∼6.0%. Simulation-based forecasting in representative patients, together with cohort-level summary of simulation error (MAPE 6.29%), demonstrated the model's ability to project individualized Hgb trajectories under varying Mircera® dosing scenarios. CONCLUSION:Our results showed the promising performance of a transformer-based machine learning model in predicting future Hgb levels within 6% of observed levels in HD patients being treated with Mircera® and IV iron. These findings suggest the potential for personalized ESA dosing strategies informed by sequential clinical data. Future studies should focus on refining these models and evaluating their real-world clinical utility with the goal of personalized ESA dosing to maximize on-target Hgb levels.

Speculative decoding has emerged as an effective approach for accelerating autoregressive inference by parallelizing token generation through a draft-then-verify paradigm. However, existing methods rely on static drafting lengths and rigid verification criteria, limiting their adaptability across varying model uncertainties and input complexities. This paper proposes an information-theoretic framework for speculative decoding based on confidence-modulated drafting. By leveraging entropy and margin-based uncertainty measures over the drafter's output distribution, the proposed method dynamically adjusts the number of speculatively generated tokens at each iteration. This adaptive mechanism reduces rollback frequency, improves resource utilization, and maintains output fidelity. Additionally, the verification process is modulated using the same confidence signals, enabling more flexible acceptance of drafted tokens without sacrificing generation quality. Experiments on machine translation and summarization tasks demonstrate significant speedups over standard speculative decoding while preserving or improving BLEU and ROUGE scores. The proposed approach offers a principled, plug-in method for efficient and robust decoding in large language models under varying conditions of uncertainty.

Modeling domain intent within an evolving domain structure presents a significant challenge for domain-specific conversational recommendation systems (CRS). The conventional approach involves training an intent model using utterance-intent pairs. However, as new intents and patterns emerge, the model must be continuously updated while preserving existing relationships and maintaining efficient retrieval. This process leads to substantial growth in utterance-intent pairs, making manual labeling increasingly costly and impractical. In this paper, we propose an efficient solution for constructing a dynamic hierarchical structure that minimizes the number of user utterances required to achieve adequate domain knowledge coverage. To this end, we introduce a neural network-based attention-driven hierarchical clustering algorithm designed to optimize intent grouping using minimal data. The proposed method builds upon and integrates concepts from two existing flat clustering algorithms DEC and NAM, both of which utilize neural attention mechanisms. We apply our approach to a curated subset of 44,000 questions from the business food domain. Experimental results demonstrate that constructing the hierarchy using a stratified sampling strategy significantly reduces the number of questions needed to represent the evolving intent structure. Our findings indicate that this approach enables efficient coverage of dynamic domain knowledge without frequent retraining, thereby enhancing scalability and adaptability in domain-specific CSRs.

BACKGROUND:AKI is a global health concern associated with high morbidity and mortality. Early diagnosis and treatment of subclinical AKI are critical for mitigating adverse outcomes. Here, we evaluated whether the urine microscopy score (UMS), a simple and cost-effective method for detecting structural kidney injury, could serve as a substitute biomarker within the AKI Risk Assessment Model (ARA-F4) to identify subclinical AKI and predict clinical AKI development. METHODS:A prospective cohort study was conducted, enrolling hospitalized adult patients (non-ICU) at moderate to high risk of AKI according to ARA-F4 model. At admission, urine microscopy was performed, and patients with UMS ≥2 without concurrent serum creatinine elevation were classified as subclinical AKI (AKI-1S); those with UMS = 1 were classified as non-AKI. The primary outcomes was development of clinical AKI within 48 hours, the need for kidney replacement therapy (KRT), and mortality. The discriminative ability of the UMS for predicting AKI was assessed using the area under the receiver operating characteristic curve (AUC). RESULTS:A total of 103 patients were included in the study, with 39 (37.9%) classified as AKI-1S and 64 (62.1%) as non-AKI at admission. Among the AKI-1S group, 89.7% developed clinical AKI within 48 hours compared to 10.9% of non-AKI patients (p<0.05). The AKI-1S group had a significant higher requirement for KRT (10.3% vs. 1.6%, p<0.05) and increased mortality rate (43.6% vs. 14.1%, p<0.05). The UMS demonstrated good predictive performance for AKI development, with an AUC of 0.84 (95% CI: 0.75-0.92). The sensitivity and specificity of the UMS were 74.5% and 92.9%, respectively. CONCLUSIONS:The UMS can be used in the ARA-F4 model to identify patients with subclinical AKI and predict the subsequent development of clinical AKI. Early recognition of subclinical AKI using the UMS can facilitate timely interventions and may reduce the burden of AKI in low- and middle-income countries.

BackgroundA substantially lower proportion of female individuals participate in sufficient daily activity compared to male individuals despite the known health benefits of exercise. Investment in female sports and exercise medicine research may help close this gap; however, female individuals are underrepresented in this research. Hesitancy to include female participants is partly due to assumptions that biological rhythms driven by menstrual cycles and occurring on the timescale of approximately 28 days increase intraindividual biological variability and weaken statistical power. An analysis of continuous skin temperature data measured using a commercial wearable device found that temperature cycles indicative of menstrual cycles did not substantially increase variability in female individuals’ skin temperature. In this study, we explore physical activity (PA) data as a variable more related to behavior, whereas temperature is more reflective of physiological changes. ObjectiveWe aimed to determine whether intraindividual variability of PA is affected by biological sex, and if so, whether having menstrual cycles (as indicated by temperature rhythms) contributes to increased female intraindividual PA variability. We then sought to compare the effect of sex and menstrual cycles on PA variability to the effect of PA rhythms on the timescales of days and weeks and to the effect of nonrhythmic temporal structure in PA on the timescale of decades of life (age). MethodsWe used minute-level metabolic equivalent of task data collected using a wearable device across a 206-day study period for each of 596 individuals as an index of PA to assess the magnitudes of variability in PA accounted for by biological sex and temporal structure on different timescales. Intraindividual variability in PA was represented by the consecutive disparity index. ResultsFemale individuals (regardless of whether they had menstrual cycles) demonstrated lower intraindividual variability in PA than male individuals (Kruskal-Wallis H=29.51; P<.001). Furthermore, individuals with menstrual cycles did not have greater intraindividual variability than those without menstrual cycles (Kruskal-Wallis H=0.54; P=.46). PA rhythms differed at the weekly timescale: individuals with increased or decreased PA on weekends had larger intraindividual variability (Kruskal-Wallis H=10.13; P=.001). In addition, intraindividual variability differed by decade of life, with older age groups tending to have less variability in PA (Kruskal-Wallis H=40.55; P<.001; Bonferroni-corrected significance threshold for 15 comparisons: P=.003). A generalized additive model predicting the consecutive disparity index of 24-hour metabolic equivalent of task sums (intraindividual variability of PA) showed that sex, age, and weekly rhythm accounted for only 11% of the population variability in intraindividual PA variability. ConclusionsThe exclusion of people from PA research based on their biological sex, age, the presence of menstrual cycles, or the presence of weekly rhythms in PA is not supported by our analysis.

The world is constantly moving towards AI based systems and autonomous vehicles are now reality in different parts of the world. These vehicles require sensors and cameras to detect objects and maneuver according to that. It becomes important to for such vehicles to also predict from a distant if a person is about to cross a road or not. The current study focused on predicting the intent of crossing the road by pedestrians in an experimental setup. The study involved working with deep learning models to predict poses and sequence modelling for temporal predictions. The study analysed three different sequence modelling to understand the prediction behaviour and it was found out that GRU was better in predicting the intent compared to LSTM model but 1D CNN was the best model in terms of speed. The study involved video analysis, and the output of pose detection model was integrated later on to sequence modelling techniques for an end-to-end deep learning framework for predicting road crossing intents.

Contrastive Decoding (CD) has emerged as an effective inference-time strategy for enhancing open-ended text generation by exploiting the divergence in output probabilities between a large expert language model and a smaller amateur model. Although CD improves coherence and fluency, its dependence on a single amateur restricts its capacity to capture the diverse and multifaceted failure modes of language generation, such as repetition, hallucination, and stylistic drift. This paper proposes Multi-Amateur Contrastive Decoding (MACD), a generalization of the CD framework that employs an ensemble of amateur models to more comprehensively characterize undesirable generation patterns. MACD integrates contrastive signals through both averaging and consensus penalization mechanisms and extends the plausibility constraint to operate effectively in the multi-amateur setting. Furthermore, the framework enables controllable generation by incorporating amateurs with targeted stylistic or content biases. Experimental results across multiple domains, such as news, encyclopedic, and narrative, demonstrate that MACD consistently surpasses conventional decoding methods and the original CD approach in terms of fluency, coherence, diversity, and adaptability, all without requiring additional training or fine-tuning.

This work-in-progress report describes MISCON, a conversational consultant being developed for a public mission project called NOURISH. With MISCON, aspiring small business owners in a food-insecure region and their advisors in Community-based organizations would be able to get information, recommendation and analysis regarding setting up food businesses. MISCON conversations are modeled as state machine that uses a heterogeneous knowledge graph as well as several analytical tools and services including a variety of LLMs. In this short report, we present the functional architecture and some design considerations behind MISCON.

We present OLG++, a semantic extension of the Obligation Logic Graph (OLG) for modeling regulatory and legal rules in municipal and interjurisdictional contexts. OLG++ introduces richer node and edge types, including spatial, temporal, party group, defeasibility, and logical grouping constructs, enabling nuanced representations of legal obligations, exceptions, and hierarchies. The model supports structured reasoning over rules with contextual conditions, precedence, and complex triggers. We demonstrate its expressiveness through examples from food business regulations, showing how OLG++ supports legal question answering using property graph queries. OLG++ also improves over LegalRuleML by providing native support for subClassOf, spatial constraints, and reified exception structures. Our examples show that OLG++ is more expressive than prior graph-based models for legal knowledge representation.

In the internet era, almost every business entity is trying to have its digital footprint in digital media and other social media platforms. For these entities, word of mouse is also very important. Particularly, this is quite crucial for the hospitality sector dealing with hotels, restaurants etc. Consumers do read other consumers reviews before making final decisions. This is where it becomes very important to understand which aspects are affecting most in the minds of the consumers while giving their ratings. The current study focuses on the consumer reviews of Indian hotels to extract aspects important for final ratings. The study involves gathering data using web scraping methods, analyzing the texts using Latent Dirichlet Allocation for topic extraction and sentiment analysis for aspect-specific sentiment mapping. Finally, it incorporates Random Forest to understand the importance of the aspects in predicting the final rating of a user.

Road accidents are quite common in almost every part of the world, and, in majority, fatal accidents are attributed to over speeding of vehicles. The tendency to over speeding is usually tried to be controlled using check points at various parts of the road but not all traffic police have the device to check speed with existing speed estimating devices such as LIDAR based, or Radar based guns. The current project tries to address the issue of vehicle speed estimation with handheld devices such as mobile phones or wearable cameras with network connection to estimate the speed using deep learning frameworks.

Dynamic Vision Sensor (DVS)-based solutions have recently garnered significant interest across various computer vision tasks, offering notable benefits in terms of dynamic range, temporal resolution, and inference speed. However, as a relatively nascent vision sensor compared to Active Pixel Sensor (APS) devices such as RGB cameras, DVS suffers from a dearth of ample labeled datasets. Prior efforts to convert APS data into events often grapple with issues such as a considerable domain shift from real events, the absence of quantified validation, and layering problems within the time axis. In this paper, we present a novel method for video-to-events stream conversion from multiple perspectives, considering the specific characteristics of DVS. A series of carefully designed losses helps enhance the quality of generated event voxels significantly. We also propose a novel local dynamic-aware timestamp inference strategy to accurately recover event timestamps from event voxels in a continuous fashion and eliminate the temporal layering problem. Results from rigorous validation through quantified metrics at all stages of the pipeline establish our method unquestionably as the current state-of-the-art (SOTA).