Abstract

Developments in Graph-Language Models (GLMs) aim to integrate the structural reasoning capabilities of Graph Neural Networks (GNNs) with the semantic understanding of Large Language Models (LLMs). However, we demonstrate that current evaluation benchmarks for GLMs, which are primarily repurposed node-level classification datasets, are insufficient to assess multimodal reasoning. Our analysis reveals that strong performance on these benchmarks is achievable using unimodal information alone, suggesting that they do not necessitate graph-language integration. To address this evaluation gap, we introduce the CLEGR(Compositional Language-Graph Reasoning) benchmark, designed to evaluate multimodal reasoning at various complexity levels. Our benchmark employs a synthetic graph generation pipeline paired with questions that require joint reasoning over structure and textual semantics. We perform a thorough evaluation of representative GLM architectures and find that soft-prompted LLM baselines perform on par with GLMs that incorporate a full GNN backbone. This result calls into question the architectural necessity of incorporating graph structure into LLMs. We further show that GLMs exhibit significant performance degradation in tasks that require structural reasoning. These findings highlight limitations in the graph reasoning capabilities of current GLMs and provide a foundation for advancing the community toward explicit multimodal reasoning involving graph structure and language.

Abstract

Visualising all dimensions of multi-dimensional real data while providing semantics is a challenging problem in data analysis. In this paper, we present the visualisation of multi-dimensional real data using our tool, 360◦ RAYS. RAYS plots all features and points of the dataset and provides information such as the geometric positioning of points or classes in the orthants of space, thereby aiding understanding of the closeness of points, classification of points, and ordering of subspaces. This visualisation technique is built after exploring viable options to plot multi-dimensional data, which included parallel coordinates, concentric coordinates, and circle segments, among others. The key idea of our approach is a hierarchical concentric ring structure, where each successive ring incrementally adds a dimension to the existing subspace, ordered by the increasing or decreasing coefficient of variation of data in the added dimension. Normalised data points are mapped to ring sectors based on a bitwise encoding of their coordinate values (non-negative or negative). The advantage of RAYS is that it presents the complete data with all dimensions and provides a clear visual grouping of points by subspace concentration. The visualisation also shows pure and overlapping orthants within the data in a flattened 2D radial layout.

Abstract

Speech enhancement models optimised for percep-
tual quality metrics such as PESQ do not necessarily improve
automatic speech recognition (ASR) performance, as aggres-
sive noise suppression can distort the phonemic cues that
acoustic models rely on. We propose a time-domain speech
frontend based on the Mamba selective state space model,
trained exclusively with ASR-oriented loss functions to pre-
serve phonetically discriminative structure rather than max-
imise perceptual quality. The model follows a U-Net encoder-
decoder architecture with 4× temporal downsampling and six
stacked bidirectional Mamba blocks operating on the raw wave-
form, trained with a multi-component loss combining multi-
resolution spectral supervision, log-mel feature matching, speech
correlation preservation, and ideal-ratio-mask spectral objec-
tives. On VoiceBank-DEMAND, the proposed model achieves
10.26% WER with wav2vec2-base-960h and 7.68% with
whisper-base, achieving statistically equivalent WER to SE-
Mamba (10.30% / 7.73%) while producing a lower PESQ
of 3.10 versus 3.69. Phoneme-class analysis confirms that the
proposed model matches or outperforms SEMamba on five
of seven articulatory classes, demonstrating that ASR-oriented
training objectives achieve equivalent robustness to perceptual
enhancement models without modifying the downstream ASR
system.

Abstract

Subtitles are essential for accessibility and global dissemination of video content, yet conventional formats often omit critical cues such as speaker identities. Character-Aware Subtitling (CAS) aims to address this gap, however, existing approaches are limited by the absence of public checkpoints, weak pipeline coordination leading to detrimental performance, and insufficient failure analysis. This work presents a systematic study of subtitling pipelines, demonstrating that improved transcript-speech alignment substantially enhances downstream performance, achieving state-of-the-art results on the LLR dataset by 2.77% DER-point zero-shot, and by 4.78% DER-point after VAD finetuning. In addition, we propose design refinements, adopt open, out-of-the-box model alternatives, and introduce a scalable method for curating diverse character priors, thereby improving the robustness and replicability of CAS.

Abstract

We study the connection between audio observations and the underlying physics of a mundane yet intriguing everyday activity: pouring liquids. Given only the sound of liquid pouring into a container, our objective is to automatically infer physical properties such as the liquid level, the shape and size of the container, the pouring rate and the time to fill. To this end, we: (i) show in theory that these properties can be determined from the fundamental frequency (pitch); (ii) train a pitch detection model with supervision from simulated data and visual data with a physics-inspired objective; (iii) introduce a new large dataset of real pouring videos for a systematic study; (iv) show that the trained model can indeed infer these physical properties for real data; and finally, (v) we demonstrate strong generalization to various container shapes, other datasets, and in-the-wild YouTube videos. Our work presents a keen understanding of a narrow yet rich problem at the intersection of acoustics, physics, and learning. It opens up applications to enhance multisensory perception in robotic pouring.

Abstract

Adapting CLIP for videos has gained popularity due to its semantic and rich representation. While CLIP is a good starting point, it typically undergoes post-pretraining (contrastive finetuning) on large video narration or caption datasets (e.g. HowTo100M, WebVid2.5M). However, such narrations or captions often lack comprehensive information needed to represent a video holistically. As the learning signal from text is sparse, the visual learning is inefficient and adaptation requires millions of samples to post-pretrain. In this work, we ask: is it possible to efficiently adapt CLIP for general and holistic video understanding? We use videos labeled with structured and dense Semantic Role Labels (SRLs) that capture actions, people or objects, their attributes, adverbs (manner), and location in a structured format representing the entire video in a holistic way. We generate rule-based captions from SRLs and demonstrate that simple contrastive finetuning on a mere 23k video-caption pairs is adequate to learn powerful, transferable representations applicable across a diverse range of video understanding tasks that require varying levels of perceptual granularity. Our adapted CLIP model, SRL-CLIP, exhibits comparable or superior performance on zero-shot text-to-video retrieval compared to state-of-the-art models that possess 4-8x more parameters and are post-pretrained on up to 6000x more data. SRL-CLIP surpasses CLIP on multiple video benchmarks, underscoring the efficient learning and improved representations.

Abstract

The maximum entropy principle, as applied to quantum systems, is a fundamental prescript positing that for a quantum system for which we only have partial knowledge, the maximum entropy state consistent with the partial knowledge is a valuable choice as the system's state. An intriguing result is that in case the only prior knowledge is of a fixed energy, the maximum entropy state turns out to be the thermal state, a ubiquitous state in several arenas, especially in statistical mechanics. We extend the consequences of this principle from quantum states to quantum processes. We establish that a quantum channel attains maximal output entropy under a fixed energy constraint if and only if it is an absolutely thermalizing channel, where the fixed output is the thermal state corresponding to that energy. Our results have potential implications for understanding the informational and thermodynamic utility of quantum channels under physical constraints. As an application, we examine the consequences for private randomness distillation from fixed energy constrained quantum processes.

Abstract

Developing Automatic Speech Recognition (ASR)
systems for specialized medical domains is challenging in low-
resource settings due to the limited availability of annotated
speech data. Synthetic speech generated using text to speech
(TTS) systems is often used to augment training data, but directly
mixing synthetic and real speech can introduce distribution
mismatch that degrades encoder representations. This work
introduces Real-Governed Representation-Calibrated Training
(RG-RCT), a training strategy that regulates the influence of syn-
thetic speech at the representation level. The proposed framework
combines reliability-based confidence estimation, representation
alignment, and layer-wise representation governance to stabilize
the encoder feature space while preserving acoustic diversity from
multi-speaker synthetic data. Experiments on medical-domain
speech datasets using Wav2Vec2 and Whisper demonstrate that
RG-RCT consistently outperforms conventional strategies such
as direct mixing and confidence-aware training. The proposed
method achieves the lowest Word Error Rate (WER), reaching
17.1% and 16.6% for Telugu and 23.3% and 25.6% for Kannada.
These results indicate that regulating encoder representation
geometry enables more effective utilization of synthetic speech
for low-resource medical ASR

Abstract

Natural and intuitive human-computer interaction
has created a growing demand for digital writing and gesture based input systems that can operate beyond the constraints of touchscreens and specialized tablets. Inertial sensing offers a low cost and flexible solution for enabling pen-based interaction on arbitrary surfaces; however, accurate motion tracking remains challenging due to the influence of gravitational acceleration during dynamic pen movements.
This work addresses the problem of dynamic tilt estimation in inertial sensing for pen-based trajectory localization. Inertial Measurement Units (IMUs), consisting of tri-axial accelerometers and gyroscopes, are commonly used to estimate motion by integrating acceleration signals. However, during handwriting, the pen undergoes continuous tilt, causing the gravitational acceleration vector to project onto all sensor axes. This contaminates the measured acceleration and makes direct integration unreliable, a challenge commonly referred to as the dynamic tilt problem.
In this paper, we focus on estimating the orientation of the pen in order to separate the gravitational component from the measured acceleration using a single tri-axial accelerometer. Under practical assumptions on the writing surface and motion continuity, we formulate tilt estimation as a constrained optimization problem. The proposed framework provides a robust and computationally efficient approach for gravity compensation, serving as a fundamental step toward accurate trajectory reconstruction. Experimental results demonstrate that the proposed approach provides stable and accurate tilt estimates, thereby improving the quality of motion signals for handwriting and gesture tracking applications.

Abstract

Ambient air pollution and environmental exposures
impose a measurable but often invisible burden on community health, particularly in rapidly urbanising regions where populations face simultaneous exposures from traffic emissions, construction activities, agrochemical residues, and degraded water quality. Despite this burden, low-cost integrated monitoring
frameworks that combine real-time IoT-based air quality sensing with individual-level health data in demographically matched cohorts remain scarce in Indian settings. This pilot study deploys a calibrated low-cost IoT-based air quality monitoring
framework alongside portable water quality instruments and structured self-reported health surveys to assess rural-urban environmental health disparities in Telangana, India. Urban PM2.5 (mean 78.7 µg/m3
, peak 684 µg/m3 ) substantially exceeded
rural levels (mean 50.2 µg/m3, peak 117 µg/m3
), with 51.3% of urban monitoring hours surpassing the 60 µg/m3 acute-concern threshold; both sites exceeded the CPCB National Ambient Air Quality Standards (NAAQS) PM2.5 limit of 40 µg/m3 . Water quality parameters were broadly similar across sites, isolating
air and noise pollution as the dominant differential exposures. Urban residents, particularly women and children, reported higher burdens of recurrent respiratory infections, antibiotic use, hair loss, headaches, and premenstrual symptoms. Together,
these findings demonstrate that a low-cost IoT-driven monitoring framework can identify patterns consistent with pollution-linked health outcomes in matched communities, offering a scalable model for environmental health surveillance across rural-urban
India.