Abstract

In this paper, we develop the resource theory of quantum secret key. Operating under the assumption that entangled states with zero distillable key do not exist, we define the key cost of a quantum state, and device. We study its properties through the lens of a quantity that we call the key of formation. The main result of our paper is that the regularized key of formation is an upper bound on the key cost of a quantum state. The core protocol underlying this result is privacy dilution, which converts states containing ideal privacy into ones with diluted privacy. Next, we show that the key cost is bounded from below by the regularized relative entropy of entanglement, which implies the irreversibility of the privacy creation-distillation process for a specific class of states. We further focus on mixed-state analogues of pure quantum states in the domain of privacy, and we prove that a number of entanglement measures are equal to each other for these states, similar to the case of pure entangled states. The privacy cost and distillable key in the single-shot regime exhibit a yield-cost relation, and basic consequences for quantum devices are also provided. Importantly, our results presented here will remain valid even if entangled states with zero distillable key were shown to exist.

Abstract

This article examines how digital trade diplomacy can play a strategic role in advancing India's economic interests by shaping rules, standards, and cooperation frameworks for digital trade, data flows and emerging technologies while supporting inclusive growth and innovation.

Abstract

The Supreme Court of India decriminalized same sex relations in 2018 through the partial abrogation of Article 377. However, the marginalization of queer communities in India persists, via stigma, inequalities, and unsafe social conditions. This reality makes digital intimacy crucial for understanding queer experiences, as online platforms often serve as the only accessible safe spaces for connection and visibility.
Research in the growing field of "Grindr studies" has extensively documented these digital dynamics. Early work--largely focused on the Global North--conceptualizing the platform as a "hook-up" space (Wortham, 2013) and "meat market" (Bonner-Thompson, 2017, p.1), often emphasizing public health and self-presentation. More recent studies from the Global South have shifted attention to the dangers of using the app in restrictive environments. Research from the global south shows how users face serious risks--from physical violence to "predators, scammers, fakers" and develop strategies of "doing trust" to handle the intense need for connection that comes from isolation (Afzal, 2025). However, despite the detailed documentation of social risks, there is a lack of work connecting these vulnerabilities directly to the design of the application themselves.
Situated in this conundrum, we investigate the experiences of queer individuals on Grindr, and what shapes these experiences. We employ a mixed-methods approach applied to a text corpus of over 350 posts from a prominent Indian LGBTQ+ subreddit, where users anonymously share personal experiences. Reddit is particularly suited for this analysis as its anonymity enables more candid and community-driven accounts of experiences. Methodologically, we first utilized a fine-tuned DistilRoBERTa-base (Sanh et al., 2019) classifier to categorize (Ekman et al., 1992) these narratives into six primary emotional states: anger, sadness, fear, joy, surprise, and disgust. We subsequently conducted a manual discourse analysis on the posts within each emotional cluster. This qualitative examination allowed us to deconstruct recurring narratives and group them into seven overarching themes. For instance, the theme of 'Societal and Cultural Context' encompasses sub-themes such as caste-based hierarchies, the rural-urban divide, and familial pressure, illustrating how the platform reproduces offline social stratifications rather than functioning as a neutral space.
This analysis reveals that negative emotions are primarily shaped by toxic masculinity, self- image insecurities, and difficulties in forming meaningful connections. By categorizing these user narratives into themes, we were able to map these emotional outcomes directly to the platform's design features. A critical example of this is the intersection between Relationship Dynamics and the app's Marketplace Interface. Our findings suggest that the infinite grid layout

commodifies users, presenting them as products to be 'shopped for' rather than individuals to be discovered and inevitably compared against a wall of alternatives. This visual-centric design encourages side-by-side evaluation, necessitating snap judgments based on a singular image.
We argue that the very platforms where users seek freedom and visibility become sites of vulnerability due to their universal design affordances. Thus, we contend that there is an urgent need to create platforms with design ideations that are embedded in local cultural specificities. To help define these nuances, our future work will expand the scope of inquiry by conducting qualitative interviews with a cohort of queer users in India, thereby contextualizing our design critique within the realities of the global south.

Abstract

How should we evaluate generation systems that combine autoregressive (AR) and diffusion decoding? We study this question through Speculative Refinement (SpecRef), a training-free hybrid method that warm-starts a masked diffusion language model from an AR draft using entropy-guided selective masking. Evaluating SpecRef across six benchmarks (HumanEval, MBPP, GSM8K, BBH, ARC-Challenge, HellaSwag) with three distinct evaluation protocols (execution-based pass@1, exact-match, log-likelihood scoring), we surface several findings relevant beyond our specific system: (1) code benchmarks conflate structural discovery} with logical correctness: providing a syntactic scaffold lifts accuracy from near zero to over 20% without changing the model, indicating that much of the baseline failure is structural; (2) a refinement tension phenomenon where multi-stage correction degrades already-correct tokens, exposing benchmark saturation ceilings invisible to single-model evaluation; (3) log-likelihood and generative evaluation produce different model rankings for the same model pair, suggesting they measure different capabilities; (4) standard Python post-processing silently breaks code evaluation for non-AR generators. These observations apply to any multi-stage or non-autoregressive generation pipeline and point toward more diagnostic evaluation practices.

Abstract

Raz (2009) proved that multilinear formulas computing the determinant of a generic $n times n$ matrix require size $n^{Omega(log n)}$. A fundamental question in understanding this lower bound is identifying which structural properties of the determinant drive this hardness. Is it the quadratic number of variables? The dense connectivity? Specific algebraic symmetries?

We establish that density is not essential. We prove the existence of $n times n$ symbolic matrices with only $Theta(n log^6 n)$ nonzero entries - reducing the variable count by a factor of $n/{log^6 n}$ - such that any multilinear formula computing their determinants still requires size $n^{Omega(log n)}$. Our construction uses rectangle sampling from the complete bipartite graph to generate sparse matrices that simultaneously maintain perfect matchings (ensuring nonzero determinant) while exhibiting diagonal imbalance under random vertex permutations - a geometric property we identify as the key driver of factor imbalance in Raz's framework.

This demonstrates that Raz's partial derivatives method is remarkably robust to sparsification, and suggests that the fundamental source of multilinear hardness for determinant lies in expansion-like combinatorial structure rather than density. Our techniques combine concentration inequalities for dependent random variables with insights from random graph theory.

Abstract

Large Language Models often achieve strong performance by generating long intermediate chain-of-thought reasoning. However, it remains unclear when a model's final answer is actually determined during generation. If the answer is already fixed at an intermediate stage, subsequent reasoning tokens may constitute post-decision explanation, increasing inference cost and latency without improving correctness. We study the evolution of predicted answers over reasoning steps using forced answer completion, which elicits the model's intermediate predictions at partial reasoning prefixes. Focusing on Qwen3-4B and averaging results across all datasets considered, we find that predicted answers change in only 32% of queries. Moreover, once the final answer switch occurs, the model generates an average of 760 additional reasoning tokens per query, accounting for a substantial fraction of the total reasoning budget. Motivated by these findings, we investigate early stopping strategies that halt generation once the answer has stabilized. We show that simple heuristics, including probe-based stopping, can reduce reasoning token usage by 500 tokens per query while incurring only a 2% drop in accuracy. Together, our results indicate that a large portion of chain-of-thought generation is redundant and can be reduced with minimal impact on performance.

Abstract

Large language models (LLMs) exhibit failures on elementary symbolic tasks such as character counting in a word, despite excelling on complex benchmarks. Although this limitation has been noted, the internal reasons remain unclear. We use character counting (e.g., "How many p's are in apple?") as a minimal, controlled probe that isolates token-level reasoning from higher-level confounds. Using this setting, we uncover a consistent phenomenon across modern architectures, including LLaMA, Qwen, and Gemma: models often compute the correct answer internally yet fail to express it at the output layer.
Through mechanistic analysis combining probing classifiers, activation patching, logit lens analysis, and attention head tracing, we show that character-level information is encoded in early and mid-layer representations. However, this information is attenuated by a small set of components in later layers, especially the penultimate and final layer MLP. We identify these components as negative circuits: subnetworks that downweight correct signals in favor of higher-probability but incorrect outputs.
Our results lead to two contributions. First, we show that symbolic reasoning failures in LLMs are not due to missing representations or insufficient scale, but arise from structured interference within the model's computation graph. This explains why such errors persist and can worsen under scaling and instruction tuning. Second, we provide evidence that LLM forward passes implement a form of competitive decoding, in which correct and incorrect hypotheses coexist and are dynamically reweighted, with final outputs determined by suppression as much as by amplification.
These findings carry implications for interpretability and robustness: simple symbolic reasoning exposes weaknesses in modern LLMs, underscoring need for design strategies that ensure information is encoded and reliably used.

Abstract

We provide a comprehensive review of the FRUIT (Faithfully Reflecting Updated Information in Text) task, which formalizes the challenge of accurately updating textual information with large language models (LLMs). Our work begins with an in-depth analysis of the FRUIT dataset, revealing key structural insights. We also investigate the unsupervised capabilities of LLMs--such as zero-shot learning, chain-of-thought reasoning, self-reflection, and evidence ordering. Experimental results demonstrate that unsupervised approaches perform competitively with supervised methods in faithful text updating. Qualitative analysis shows that updates utilizing table-structured evidence outperform those based on unstructured text. We also discuss important limitations, including the need for new datasets and the risks of information leakage in this domain. These findings have significant implications for applications requiring precise document updates, such as software engineering, technical documentation, and legal document maintenanc

Abstract

Reliable recognition of motorcycle driving events is important for rider safety and intelligent mobility applications. This work addresses the limited generalisability of earlier studies by introducing a more diverse real-world time-series dataset collected using 3-axis accelerometer and gyroscope sensors from both male and female riders, recorded during day and night riding. The dataset captures wider variations in riding behaviour and sensor dynamics, making event recognition more realistic and challenging. To model this increased variability, we focus on a Time-Series Transformer (TST) architecture that can capture long-range temporal dependencies while remaining efficient for embedded deployment. The proposed approach is evaluated on edge hardware, where inference latency, memory usage, and model size are analysed. Experimental results show that the TST achieves strong and consistent class-wise performance while maintaining low computational overhead, making it well suited for real-time on-device two-wheeler event recognition.

Abstract

Automatic syllable stress detection is a key component of computer-assisted language learning (CALL) systems, as it enables precise feedback to second-language (L2) learners. Most of the existing detection methods depend on explicit phoneme or syllable-level boundaries derived from manual annotation or forced alignment. However, manual annotation is costly, and the forced alignment methods are prone to errors. To overcome these limitations, we introduce a fully boundary-independent framework for syllable stress detection based on hierarchical time-compression. Our approach employs an encoder-decoder architecture with Hierarchical compact self-Attention based Time-compression (HAT) to progressively compress the frame-level embeddings into syllable-level representations. This enables the model to capture stress cues directly from speech without any boundary supervision. We evaluate the proposed framework on the ISLE corpus of second language English learners (German and Italian), leveraging self-supervised wav2vec 2.0 embeddings. Experimental results show that our approach consistently outperforms both boundary-dependent and prior partially boundary-independent baselines. Our approach achieves a maximum accuracy of 95.07% on GER and 96.18% on ITA, with highest absolute improvements of up to 23% and 19%.