Abstract

Masonry structures, both contemporary and historic, often require strengthening to meet modern safety demands and resist seismic actions. Conventional retrofitting techniques are frequently associated with drawbacks, including increased mass, compatibility issues with substrates, and intrusive application procedures. Composite-based solutions have therefore gained prominence as effective alternatives. Among these, fibre-reinforced polymers (FRPs) were initially adopted due to their superior mechanical properties and ease of handling; however, their dependence on epoxy resins introduces limitations related to durability, cost, and performance under adverse environmental conditions. These limitations have driven the development of textile-reinforced mortar (TRM) systems as compatible and sustainable strengthening solutions. This study presents a critical synthesis of TRM-based strengthening for masonry structures, integrating experimental evidence, field applications, and existing design provisions. The mechanical behaviour, bond-controlled interaction at the TRM-masonry interface, and durability characteristics of TRM systems are examined. Existing design provisions from international guidelines are systematically compared and critically reinterpreted within a limit-state framework to evaluate the in-plane and out-of-plane behaviour of TRM-strengthened walls. The framework explicitly accounts for failure mechanisms, bond behaviour, and effective strain limits governing reinforcement efficiency. Key findings indicate that the effectiveness of TRM strengthening is governed by interface behaviour, reinforcement configuration, and anchorage conditions, rather than textile tensile capacity alone. Field applications are critically analyzed to extract design-relevant insights, demonstrating improved strength, deformation capacity, and crack control while maintaining compatibility with masonry substrates. Emerging natural fibre-based TRM systems are also identified as promising alternatives for sustainable strengthening. The study provides design-oriented insights for improving the reliability and applicability of TRM retrofitting strategies.

Abstract

utomatic pronunciation assessment (APA) of L2 learners is
critical in computer-assisted language learning. Existing ap-
proaches typically rely on canonical phonemes, its aligned
boundaries and large labeled data, which is costly and cumber-
some. To address these, we propose ALFreeD, a segmentation-
free APA framework guided by teacher's speech without the
need of canonical phonemes using minimal labeled data. In
this, pronunciation quality is assessed by quantifying deviation
between learners' and teacher's utterances. First, frame-level
embeddings are extracted using HuBERT, and deviation scores
are computed between them after obtaining their temporal cor-
respondence with DTW. Next, utterance level pronunciation de-
viation vector is obtained by suppressing variabilities other than
pronunciation quality with i-vector framework and is passed to
a MLP for predicting quality. Experiments on SpeechOcean762
demonstrate competitive performance with recent E2E methods
using minimal labeled data

Abstract

Text-to-image diffusion models can generate violent, harmful, or
weapon-related content, creating a difficult trade-off between safety
alignment and visual quality. Existing preference-based approaches
typically optimize a weighted combination of safety and quality
objectives, which can become unstable when the two objectives conflict.

We present BiDPO, a bilevel direct preference optimization
framework for safety-first diffusion alignment with two disjoint LoRA
adapters. A lower-level adapter is trained first on safety-ranked image
pairs, and an upper-level adapter is then trained on quality-ranked
pairs constructed from safety-gated candidates. This separation reduces
direct interference between safety and quality updates while keeping the
base model frozen.

We also study a failure mode that we call upper-level safety
drift}, where quality-focused updates reduce the safety gains obtained
at the lower level. To mitigate this effect, we evaluate several
controls, including chained references, gap-weighted pair mining, and
adaptive detector weighting. On SD1.5, the best BiDPO variant achieves
a lower unsafe-prompt safety-risk score than our joint baseline
(0.235 vs. 0.248) while slightly improving quality under our
evaluation pipeline, and it also outperforms the strongest comparison
methods we tested. Across controlled experiments, BiDPO shows a
consistent safety advantage over matched joint training. These results
suggest that a bilevel training structure can provide a stronger
safety-quality trade-off than a single weighted objective.

Abstract

Memory bandwidth contention in multi-core systems severely impacts application performance and quality-of-service (QoS) guarantees. Regulating the shared memory bandwidth mitigates the memory performance uncertainty thereby making it a manageable resource and improving trustworthiness of multi-core systems. In this work we propose a memory bandwidth regulation mechanism LMS-AR, i.e., LMS Prediction-based Adaptive Regulator within a Linux kernel module to distribute the memory bandwidth as a resource among the CPU cores. We describe a design in which both monitoring and regulation is enforced from outside by a master core - which is not a dedicated controller for regulation. This allows for plugging in computationally heavy prediction and regulation algorithms without interfering with the regulated core. An adaptive filtering technique was employed for prediction of per-core bandwidth requirement. We conducted several experiments with SPEC CPU 2017 benchmarks distributed across multiple cores. Our proposed approach demonstrated significant improvement over Memguard with respect to slowdown ratios caused due to memory contention

Abstract

We study the problem of selecting large language models (LLMs) for user queries in settings where multiple LLM providers submit the cost of solving a query. From the user's perspective, choosing the optimal model is a sequential, query-dependent decision problem: high-capacity models offer more reliable outputs but are costlier, while lightweight models are faster and cheaper. We formalize this interaction as a reverse auction mechanism design problem with contextual online learning, where the user adaptively discovers which model performs best while eliciting costs from competing LLM providers. Existing multi-armed bandit (MAB) mechanisms focus on forward auctions and social welfare, leaving open the challenges of reverse auctions, provider-optimal outcomes, and contextual adaptation. We address these gaps by designing a resampling-based procedure that generalizes truthful forward MAB mechanisms to reverse auctions and prove that any monotone allocation rule with this procedure is truthful. Using this, we propose a contextual MAB algorithm that learns query-dependent model quality with sublinear regret. Our framework unifies mechanism design and adaptive learning, enabling efficient, truthful, and query-aware LLM selection.

Abstract

Standard depth sensors systematically fail on transparent surfaces, creating corrupted 3D maps and severe navigation hazards. While specialized hardware sensors can detect glass, they lack modularity and have extensive hardware dependencies. Consequently, learning-based monocular depth estimation has emerged as a compelling alternative. However, domain-specific glass-aware monocular depth estimators struggle with unfamiliar indoor layouts; restricted by the severe scarcity of real-world glass depth annotations, they fail to generalize zero-shot to new settings. This motivates us to explore whether the extensive priors of text-to-image diffusion models can enable generalizable perception of transparent surfaces. We introduce SILICA, a unified pipeline leveraging these priors to jointly predict glass segmentation and glass-aware depth. This mutual information exchange establishes a robust spatial hierarchy, entirely eliminating the need for paired real-world glass depth annotations. Subsequently, we use the predicted segmentation mask to explicitly filter incorrect glass depth points from standard sensors, recovering accurate metric glass depth for downstream 3D mapping and autonomous collision avoidance. Supported by our novel Mirage 18k dataset, extensive experiments demonstrate that SILICA achieves remarkable zero-shot transfer across diverse, unseen environments, outperforming state-of-the-art models by almost 20% and setting a new benchmark for transparent surface perception.

Abstract

How do humans allocate scarce cognitive resources across a stream of related decisions? Russek et al. (2025) predict reaction time on each chess move from the within‑move Benefit of Computation, treating moves as if they were drawn independently. We ask whether those costs also depend on computation carried forward from the previous move. In chess, a player who has calculated a likely opponent reply can often reuse that work; a surprising reply should force revision. On 135,608 live‑game tactical sequences (extracted from the Lichess puzzle database and re‑timed against the original 2019 Lichess game database) played by 74,609 unique players, the within‑sequence log‑RT spike is +0.309 (Cohen's 𝑑 = 0.235; paired 𝑡 = 86, N = 135,608; BF10 > 10100). Hierarchical regression climbs from 𝑅² = 0.011 (Russek‑static) to 0.282; the amortization interaction is large when predictability is operationalized by the human‑style Maia‑2 (𝛽 = −0.054, 𝑝 < 10⁻²⁷) and not significant under Stockfish best‑move predictability. A three‑parameter Bayesian Sampling Model with Cache recovers cache strength 𝑠 = 0.085 [0.071, 0.099], with 100% of bootstrap draws positive. The H1 spike is roughly twice as large in solved sequences as in failed ones (𝑡 = 26.2, 𝑝 < 10⁻¹⁵⁰); this pattern suggests successful tactical recognition followed by decisive recalculation rather than passive surprise. The findings replicate on a 2024 Lichess cohort and on FIDE World Rapid + Blitz 2024 over‑the‑board play.

Abstract

We investigate whether scaling model parameters improves humor generation through a controlled ablation study. Using five Qwen3 variants (8B-235B, dense and MoE), we generate jokes across 50 themes. Beyond evaluating humor scaling, this work serves as an empirical study into the nature of LLM versus human evaluations on highly subjective creative tasks. While an automated judge yields a perfect monotonic ranking between parameter count and win rate, human annotators find no significant aggregate difference in humor quality. Restricting to themes where annotators agree reveals a significant preference for the largest model (p = 0.039), suggesting scaling effects exist but are masked by a "quality floor." Crucially, our analysis of bias characteristics shows that the automated judge exhibits severe positional and length biases compared to human evaluators, further suggesting that LLMs may systematically distort quality differences on subjective tasks.

Abstract

We present a neurosymbolic pipeline for computational humor generation grounded in the General Theory of Verbal Humor. The system constructs the joke in five sequential stages: context analysis, humor architecture (identifying core incongruity), delivery strategy, content writing, and pairwise judging, orchestrated through the DSPy framework. The system generates four candidate jokes per input with independent humor strategies, then selects the best through knockout tournament-style evaluation. Despite using Gemma 3 27B, a model with roughly 20× fewer total parameters than frontier systems, our approach achieves competitive results across all five subtasks of SemEval- 2026 Task 1 (MWAHAHA), placing 2nd in two subtasks. We argue that these results demonstrate the viability of structured, theory-driven decomposition for solving complex tasks and that how a model reasons about humor is just as important as how large the model is.

Abstract

Introduction According to the industry' latest definitive benchmark, the "India Influencer Marketing Report 2025," released by WPP's The Goat Agency and Kantar in June 2025, the sector was valued at |3,600 crore in 2024 and is projected to grow by 25% in 2025. For content creators, this financial influx signifies a transition from casual social presence to a professionalized economy where earnings are derived from online content, ranging from fixed-fee brand endorsements and affiliate marketing to long-term ambassadorships. While influencers have been widely researched globally, there are limited studies on the Indian influencer economy. Recent studies that do cover them primarily focus on topics such as their cultural impact, consumer culture, trust between brands and customers, para-social relationships, and the ethical risks of being an influencer. The global scholarship on influencers has also examined how inequalities across race, gender, social and socioeconomic status, and sexual orientation shape the influencer economy. Another strand of scholarship investigates monetary and non-monetary transactions, such as gifts, free products, discount coupons, or gift cards, within the influencer economy (Duffy, 2017; Christin and Lu, 2023). The payment amount is highly context-dependent, making it vague and open to interpretation. This ambiguity, Carbone (2019) argues, opens the door to different forms of discrimination. However, there is limited scholarly attention to how monetary and non-monetary compensation is negotiated in practice. This gap also indicates the presence of diverse spheres of circulation within the influencer economy. We use the anthropology of barter exchange as a framework to investigate the economic circuits in which early-career influencers engage. Treating the influencer economy as having distinct transactional cultures, we envisage interpreting value creation and circulation within local contexts. Recent studies on barter demonstrate that it is not a remnant of the old order but is present across various demographics (Stapleton, 2022). Borrowing from Gudeman (2001), we interpret the initial value circulation in the influencers' lifecycle through a lens of 'self-currency', i.e., the non-monetary exchanges a participant in this economy requires to be engaged. Thus, rather than a monetary valuation, the commodity itself is exchanged for its inherent value. Further, Gudeman (2001) argues that in the contemporary economy, there exists a trader's dilemma: should one try to increase their profits or maintain their relations? Thus, this investigation centers the influencer economy in India to inquire into the subterranean circulation of values and their logics. Objective/Research Question To investigate the above scholarly gap, we ask: How does barter as an exchange logic appear in the influencer economy? Methods In the influencer industry, as Trevisan & Farinosi (2024) argue, fashion influencers are among the most successful. Influencer marketing has the largest reach in the fashion, beauty, and lifestyle segments. Thus, we focus on fashion micro-influencers in our study, as research suggests that marketers are shifting away from mega and macro influencers toward micro-influencers, who appear more relatable and thus able to attract more commercial prospects (Britt et al., 2020). For this investigation, the third author conducted an interpretivist ethnographic enquiry. The methods used in the field were content analysis and semi-structured interviews. 10 female Instagram micro-influencers were selected and interviewed from Kolkata, India. In addition, we interviewed 2 brand owners (both females). All of them engaged in the lifestyle and clothing sectors. The influencers were selected through a snowball sampling method. In the definition of micro-influencers, influencers have a following between 10,000 to 100,000. All the microinfluencers that we interviewed and whose Instagrams we studied fell within this category. The interviews and their social media actions demonstrated the high frequency of collaborations between them and various other stakeholders (for instance, friends and brands). Following this thread, we investigate the spheres of circulation as a social phenomenon during various stages of influencers' growth. The collected data were analyzed through thematic interpretation. Some of the major themes that emerged are: social ties in content production, commodity exchange, relatability, and network size. For ethical purposes, all identity markers have been anonymized. Results Our interview findings indicate three distinct phases of growth for influencers. First, the altruistic phase, where micro-influencers create content with acquaintances. The second stage is the commodity barter phase, where the influencers create content for brands and boutiques in exchange for a commodity. The third phase is the monetisation phase, which involves producing content for brands in exchange for money. Both the altruistic phase and the commodity barter phase are embedded in the logic of barter. In the altruistic phase, aspiring influencers form friendships to collaborate with aspiring photographers, editors, and beauty professionals to create content. The idea being that this free labor, donated by everyone, will lead to the growth of all involved. In terms of Instagram growth, this would translate into more followers and, in turn, an expansive fan base. Apart from their labour, micro-influencers create relatable content. This 'relatability' means that the community of followers resonates and approves of the content created; in addition, the community feels that the micro-influencer is similar to them in terms of social position and life choices. In platform terminology, the altruistic phase is key to achieving network effects and building a community. This phase is followed by the commodity barter phase, in which either the micro-influencer or the brands can reach out to the other party to collaborate. Our respondents stated that they created content while wearing the brands' products (mostly clothes) and, in return, kept the products themselves. If there are multiple commodities, they are still paid a single remuneration. In such barter collaborations, brands bear the production costs (for instance, hiring professional photographers). Money is not considered an exchange tool in such barter collaborations, and it is not discussed. Until the micro-influencer has reached a minimum number of followers (close to 50k) and is working on multiple barter collaborations, they are not offered collaborations with monetary compensation. The key ingredient is the exchange of one's followership (network size) for the commodity (for instance, a dress) in the commodity barter phase. Whereas in the altruistic phase, there is an exchange of labour to increase the network size for all participants. The findings consist of the anthropologies of exchange, where diverse spheres of circulation exist, and in the digital world, it is not exclusively impregnated by monetary logics. In a similar line to the extant literature on anthropology of barter, there is the exchange of self-currency the network size becomes a non-monetary commodity being exchanged. Just as in the trader's dilemma, our respondents want to achieve monetary gains through content production, yet they are aware of the social ties (here, the community of followers). Thus, micro-influencers carefully choose only those collaborations that will be appreciated by their followers. Future Work Building on these findings, we envisage investigating the monetization phase of micro-influencers. In addition to the above research interest, we are looking at the aesthetic performativity of fashion micro-influencers. Thus, we ask, How do micro-influencers negotiate authenticity in a competitive social media landscape in India.