Large language models (LLMs) have demonstrated remarkable performance across diverse tasks by encoding vast amounts of factual knowledge. However, they are still prone to hallucinations, generating incorrect or misleading information, often accompanied by high uncertainty. Existing methods for hallucination detection primarily focus on quantifying internal uncertainty, which arises from missing or conflicting knowledge within the model. However, hallucinations can also stem from external uncertainty, where ambiguous user queries lead to multiple possible interpretations. In this work, we introduce **Semantic Volume**, a novel mathematical measure for quantifying both external and internal uncertainty in LLMs. Our approach perturbs queries and responses, embeds them in a semantic space, and computes the Gram matrix determinant of the embedding vectors, capturing their dispersion as a measure of uncertainty. Our framework provides a generalizable and unsupervised uncertainty detection method without requiring internal access to LLMs. We conduct extensive experiments on both external and internal uncertainty detections, demonstrating that our Semantic Volume method consistently outperforms existing baselines in both tasks. Additionally, we provide theoretical insights linking our measure to differential entropy, unifying and extending previous sampling-based uncertainty measures such as the semantic entropy. Semantic Volume is shown to be a robust and interpretable approach to improving the reliability of LLMs by systematically detecting uncertainty in both user queries and model responses.

Large Language Models (LLMs) increasingly serve diverse global audiences, making it critical for responsible AI deployment across cultures.While recent works have proposed various approaches to enhance cultural alignment in LLMs, a systematic analysis of their evaluation benchmarks remains needed.We propose a novel framework that conceptualizes alignment along three dimensions: Cultural Group (who to align with), Cultural Elements (what to align), and Awareness Scope (how to align: majority-focused vs. diversity-aware).Through this framework, we analyze 105 cultural alignment evaluation benchmarks, revealing significant imbalances: Region (37.9%) and Language (28.9%) dominate Cultural Group representation; Social and Political Relations (25.1%) and Speech and Language (20.9%) concentrate Cultural Elements coverage; and an overwhelming majority (97.1%) of datasets adopt majority-focused Awareness Scope approaches.In a case study examining AI safety evaluation across nine Asian countries (Section 5), we demonstrate how our framework reveals critical gaps between existing benchmarks and real-world cultural biases identified in the study, providing actionable guidance for developing more comprehensive evaluation resources tailored to specific deployment contexts.1 We extend the demographic proxies from Adilazuarda et al. (2024) to include appearance, sexual orientation, health, socioeconomic status, online communities, and time periods (e.g., dynasties) to better accommodate our dataset analysis.2 From Thompson et al. (2020), semantic domains include: food and drink, clothing and grooming, speech and language, modern world, physical world, the house, basic actions and technology, agriculture and vegetation, animals, social and political relations, emotions and values, kinship, and cognition, with names from Adilazuarda et al. (2024).We extend this list to include arts, mental perception, and current and historical events based on our dataset findings.3 These can be understood as prototypical representations of cultural knowledge (cf.Rosch and Mervis 1975; see also Shore 1996).We also recognize that some papers we classified as majority-focused do consider multiple annotator perspec-

Reasoning-enhanced large language models (RLLMs), whether explicitly trained for reasoning or prompted via chain-of-thought (CoT), have achieved state-of-the-art performance on many complex reasoning tasks. However, we uncover a surprising and previously overlooked phenomenon: explicit CoT reasoning can significantly degrade instruction-following accuracy. Evaluating 20+ models on two benchmarks: IFEval (with simple, rule-verifiable constraints) and ComplexBench (with complex, compositional constraints), we consistently observe performance drops when CoT prompting is applied. Through large-scale case studies and an attention-based analysis, we identify common patterns where reasoning either helps (e.g., with formatting or lexical precision) or hurts (e.g., by neglecting simple constraints or introducing unnecessary content). We propose a metric, constraint attention, to quantify model focus during generation and show that CoT reasoning often diverts attention away from instruction-relevant tokens. To mitigate these effects, we introduce and evaluate four strategies: in-context learning, self-reflection, self-selective reasoning, and classifier-selective reasoning. Our results demonstrate that selective reasoning strategies, particularly classifier-selective reasoning, can substantially recover lost performance. To our knowledge, this is the first work to systematically expose reasoning-induced failures in instruction-following and offer practical mitigation strategies.

Targeting interest to match a user with services (e.g. news, products, games, advertisements) and predicting friendship to build connections among users are two fundamental tasks for social network systems. In this paper, we show that the information contained in interest networks (i.e. user-service interactions) and friendship networks (i.e. user-user connections) is highly correlated and mutually helpful. We propose a framework that exploits homophily to establish an integrated network linking a user to interested services and connecting different users with common interests, upon which both friendship and interests could be efficiently propagated. The proposed friendship-interest propagation (FIP) framework devises a factor-based random walk model to explain friendship connections, and simultaneously it uses a coupled latent factor model to uncover interest interactions. We discuss the flexibility of the framework in the choices of loss objectives and regularization penalties and benchmark different variants on the Yahoo! Pulse social networking system. Experiments demonstrate that by coupling friendship with interest, FIP achieves much higher performance on both interest targeting and friendship prediction than systems using only one source of information.

Recency ranking refers to the ranking of web results by accounting for both relevance and freshness. This is particularly important for "recency sensitive" queries such as breaking news queries. In this study, we propose a set of novel click features to improve machine learned recency ranking. Rather than computing simple aggregate click through rates, we derive these features using the temporal click through data and query reformulation chains. One of the features that we use is click buzz that captures the spiking interest of a url for a query. We also propose time weighted click through rates which treat recent observations as being exponentially more important. The promotion of fresh content is typically determined by the query intent which can change dynamically over time. Quite often users query reformulations convey clues about the query's intent. Hence we enrich our click features by following query reformulations which typically benefit the first query in the chain of reformulations. Our experiments show these novel features can improve the NDCG5 of a major online search engine's ranking for "recency sensitive" queries by up to 1.57%. This is one of the very few studies that exploits temporal click through data and query reformulations for recency ranking.

Millions of users retrieve information from the Internet using search engines. Mining these user sessions can provide valuable information about the quality of user experience and the perceived quality of search results. Often search engines rely on accurate estimates of Click Through Rate (CTR) to evaluate the quality of user experience. The vast heterogeneity in the user population and presence of automated software programs (bots) can result in high variance in the estimates of CTR. To improve the estimation accuracy of user experience metrics like CTR, we argue that it is important to identify typical and atypical user sessions in clickstreams. Our approach to identify these sessions is based on detecting outliers using Mahalanobis distance in the user session space. Our user session model incorporates several key clickstream characteristics including a novel conformance score obtained by Markov Chain analysis. Editorial results show that our approach of identifying typical and atypical sessions has a precision of about 89%. Filtering out these atypical sessions reduces the uncertainty (95% confidence interval) of the mean CTR by about 40%. These results demonstrate that our approach of identifying typical and atypical user sessions is extremely valuable for cleaning "noisy" user session data for increased accuracy in evaluating user experience.

Medium access techniques for wireless sensor networks raise the important question of providing periodic energy-efficient radio sleep cycles while minimizing the end-to-end communication delays. This study aims to minimize the communication latency given that each sensor has a duty cycling requirement of being awake for only 1/k time slots on an average. As a first step we consider the single wake-up schedule case, where each sensor can choose exactly one of the k slots to wake up. We formulate a novel graph-theoretical abstraction of this problem in the general setting of a low-traffic wireless sensor network with arbitrary communication flows and prove that minimizing the end-to-end communication delays is in general NP-hard. However, we are able to derive and analyze optimal solutions for two special cases: tree topologies and ring topologies. Several heuristics for arbitrary topologies are proposed and evaluated by simulations. Our simulations suggest that distributed heuristics may perform poorly because of the global nature of the constraints involved. We also show that by carefully choosing multiple wake-up slots for each sensor significant delay savings can be obtained over the single wake-up schedule case while maintaining the same duty cycling. Using this technique, we propose algorithms that offer a desirable bound of d+O(k) on the delay for specialized topologies like the tree and grid and a weaker guarantee of O((d+k)log n) for arbitrary graphs, where d is the shortest path between 2 nodes in the underlying topology and n is the total number of nodes.

One of the main challenges in ad hoc networks research is understanding the effect of mobility on the performance of routing protocols. In a previous study, we have shown why mobility affects the performance of routing protocols. In this study, we further extend our approach to analyze the interplay between mobility and the protocol mechanistic building blocks. Through this approach we hope to explain how performance varies with mobility by decomposing the protocol into parameterized mechanistic building blocks based on their functionalities. Then, we apply this approach to reactive MANET routing protocols like AODV and DSR, which enables us to build a common building block architecture that encompasses these reactive protocols. The effect of mobility on each building block is evaluated. We are specifically interested in understanding the contribution of each building block to the overall protocol performance. Through simulations, several lessons on protocol design are learnt. For example, in both AODV and DSR, flooding and caching seem to have a great effect on performance, while salvaging in DSR barely seems to have an effect on the protocol performance.

We examine the problem of maximizing data collection from an energy-limited store-and-extract wireless sensor network, which is analogous to the maximum lifetime problem of interest in continuous data-gathering sensor networks. One significant difference is that this problem requires attention to "data-awareness" in addition to "energy-awareness." We formulate the maximum data extraction problem as a linear program and present a 1+omega iterative approximation algorithm for it. As a practical distributed implementation we develop a faster greedy heuristic for this problem that uses an exponential metric based on the approximation algorithm. We then show through simulation results that the greedy heuristic incorporating this exponential metric performs near-optimally (within 1 to 20% of optimal, with low overhead) and significantly better than other shortest-path routing approaches, particularly when nodes are heterogeneous in their energy and data availability

We develop a detailed approach to study how mobility impacts the performance of reactive mobile ad hoc network routing protocols. In particular, we examine how the statistics of path durations including probability density functions vary with the parameters such as the mobility model, relative speed, number of hops, and radio range. We find that at low speeds, certain mobility models may induce multimodal distributions that reflect the characteristics of the spatial map, mobility constraints and the communicating traffic pattern. However, this paper suggests that at moderate and high velocities the exponential distribution with appropriate parameterizations is a good approximation of the path duration distribution for a range of mobility models. Analytically, we show that the reciprocal of the average path duration has a strong linear relationship with the throughput and overhead of dynamic source routing (DSR), which is also confirmed by simulation results. In addition, we show how the mathematical expression obtained for the path duration distribution can also be used to prove that the nonpropagating cache hit ratio in DSR is independent of velocity for the freeway mobility model. These two case studies illustrate how various aspects of protocol performance can be analyzed with respect to a number of significant parameters including the statistics of link and path durations.

A mobile ad hoc network (MANET) is a collection of wireless mobile nodes forming a temporary network without using any existing infrastructure. Since not many MANETs are currently deployed, research in this area is mostly simulation based. Random waypoint is the commonly used mobility model in these simulations. Random waypoint is a simple model that may be applicable to some scenarios. However, we believe that it is not sufficient to capture some important mobility characteristics of scenarios in which MANETs may be deployed. Our framework aims to evaluate the impact of different mobility models on the performance of MANET routing protocols. We propose various protocol independent metrics to capture interesting mobility characteristics, including spatial and temporal dependence and geographic restrictions. In addition, a rich set of parameterized mobility models is introduced including random waypoint, group mobility, freeway and Manhattan models. Based on these models several 'test-suite' scenarios are chosen carefully to span the metric space. We demonstrate the utility of our test-suite by evaluating various MANET routing protocols, including DSR, AODV and DSDV. Our results show that the protocol performance may vary drastically across mobility models and performance rankings of protocols may vary with the mobility models used. This effect can be explained by the interaction of the mobility characteristics with the connectivity graph properties. Finally, we attempt to decompose the routing protocols into mechanistic "building blocks" to gain a deeper insight into the performance variations across protocols in the face of mobility.

We develop a detailed approach to study how mobility impacts the performance of reactive MANET routing protocols. In particular we examine how the statistics of path durations including PDFs vary with the parameters such as the mobility model, relative speed, number of hops, and radio range. We find that at low speeds, certain mobility models may induce multi-modal distributions that reflect the characteristics of the spatial map, mobility constraints and the communicating traffic pattern. However, our study suggests that at moderate and high velocities the exponential distribution with appropriate parameterizations is a good approximation of the path duration distribution for a range of mobility models. The reciprocal of the average path duration is analytically shown to have a strong linear relationship with the throughput and overhead that is confirmed by the simulation results for DSR.