Abstract

This thesis investigates the dynamics of answer formation in Large Language Models (LLMs) along
two dimensions: vertically across network layers, and horizontally across sequential reasoning tokens.
To evaluate vertical evolution, we apply layer-wise linear probes to open-weight models on a controlled character-counting task. The probing results show that early and middle layers reliably encode
sub-token character counts. Late-stage network components, specifically the penultimate and final multilayer perceptrons, actively suppress this information and default to degenerate prediction strategies. The
models thus fail at utilization, despite having the required representational capacity.
For horizontal evolution, we introduce Forced Answer Completion (FAC) to extract a model's intermediate answer commitment at each step of its chain-of-thought trace. Across multiple-choice, numeric,
and search tasks, FAC shows that models decide their final answer long before the reasoning trace concludes. After filtering out Transient Answer Flips (TAFs) with a causal smoothing operator, we find
that nearly half of all reasoning tokens are generated after the model's decision has stabilized. These
late-stage tokens function primarily as post-hoc justification.
These findings enable direct interventions in LLM inference. Vertical tracking isolates suppressed
knowledge, providing targets for structural adjustments. Horizontal tracking supports lightweight earlystopping gates by training linear probes on intermediate hidden states, we predict when an answer has
stabilized and halt generation. Testing on the Qwen3 architecture demonstrates that this early-stopping
method saves hundreds of tokens per query with a marginal drop in accuracy

Abstract

The Internet of Things (IoT) has revolutionized various fields by enabling the collection of data
in real-time and interacting remotely with physical systems. In the education sector, one of the most
impactful uses of IoT is remote labs, which enable students to perform experiments virtually. This
has been particularly impactful during the COVID-19 pandemic or for those who are located remotely.
Though remote labs are becoming extremely popular for physics and electronics due to their relatively
simpler requirements, chemistry has not yet found a suitable place due to the complexity involved. In
addition, existing remote labs are not focused on creating a deep level of reflection during learning.
This thesis presents a two-pronged approach to bridging the gap between traditional and remote
laboratory learning by focusing on chemistry education and pedagogical enhancement. First, it details
the design and implementation of a fully retrofitted, IoT-enabled titration system. The proposed system
has been implemented on the RLabs platform at IIIT Hyderabad and provides a facility for users to
conduct acid-base titration experiments virtually using a web-based platform. It consists of hardware
components that help in precise measurement of fluids, visual feedback mechanisms, and automatic
cleaning facilities. This helps users conduct repeated experiments without any physical interference.
The experimental setup has been tested for accuracy, usability, and cost-effectiveness, demonstrating its
potential as a scalable solution for remote chemistry education.
The second part of this work addresses the pedagogical shortcomings of existing remote labs, which
are mostly limited in their instruction and lack active engagement with learners during the experiment.
To fill this gap, a large language model (LLM)-powered assistant is integrated into the RLabs framework. Grounded in a dual-theoretical foundation of Kolb's Experiential Learning Cycle and Bloom's
Taxonomy, the assistant facilitates active experimentation, conceptual reflection, and assessment across
multiple cognitive levels. It includes context-based tutoring, structured viva examination, and experiment completion verification, mimicking the role of a human lab instructor. The system was evaluated
through user studies that benchmarked candidate language models and quantified the impact of staged
LLM-assisted guidance, showing clear improvements in learner performance across all six levels of
Bloom's Taxonomy.
These contributions show a marked advancement in the development of remote laboratory systems
by bringing together robust experimentation via IoT and intelligent tutoring. The contributions can be
seen as a step towards the development of more holistic and capable remote laboratory systems that can
effectively mimic in-person learning while being accessible to a wide audience.

Abstract

Speech prominence refers to the relative emphasis placed on certain units in speech to express the
communicative intent. This prominence is realized at two different levels. One, at the syllable level,
where one syllable within a word is made more prominent than the rest, referred to as lexical or syllablelevel prominence. Lexical prominence is generally produced correctly by native speakers and is commonly incorporated into speech synthesis through pronunciation lexicons, but non-native speakers frequently misplace it due to the influence of their native language, motivating the need for automatic
lexical stress detection. Two, at the word level, where certain words within an utterance are emphasized more than others, commonly referred to as sentence or word-level prominence. Unlike lexical
prominence, automatic detection and natural generation of sentence prominence remain challenging
problems for both native and non-native speech. Despite their importance, existing prominence modeling approaches continue to face several limitations. First, most methods rely on handcrafted prosodic
features, such as pitch, energy, and duration, which may not fully capture prominence-related information. Second, prominence modeling at both syllable and word levels typically depends on accurate
boundary annotations, obtained either through costly manual labeling or error-prone forced alignment.
Third, training these systems requires expensive prominence labels, limiting scalability to low resource
language backgrounds.
In this thesis, these limitations are addressed progressively, moving from supervised to unsupervised learning and from boundary-dependent to boundary-independent modeling across both linguistic
levels. At the syllable level, it is first demonstrated that self-supervised speech representations provide a substantially stronger foundation for prominence modeling than handcrafted features, with sequential modeling further improving performance by capturing inter-syllable dependencies. The dependence on boundary annotations is then removed through hierarchical temporal compression, introducing
boundary-independent frameworks that match and often exceed boundary-dependent counterparts. The
reliance on prominence labels is further addressed through Post-Net and Post-Net2.0, which leverage
pseudo-label generation and linguistic constraints to enable effective unsupervised prominence detection. A fully unsupervised and boundary-independent framework, SRAMA, is proposed, combining
Adaptive Local Monotonic Attention (AMA) with iterative deep clustering, approaching supervised
performance without requiring boundaries, transcriptions, or prominence labels. Finally, the scope is
extended to the word level through ProTTS, which integrates prominence discovery directly into SOTA
TTS, namely, FastSpeech 2, using word-level prominence inferred from prosodic predictions to guide
synthesis -- producing more natural and expressive speech and bridging the two central applications of
this thesis: automatic speech prominence detection and expressive speech synthesis.

Abstract

Despite negation being one of the core elements in any language, it remains a challenging
phenomenon for modern Large Language Models (LLMs). Recently, there have been growing efforts to evaluate how models handle negation. However, the existing probing datasets
are mostly English-centric, leaving morphologically rich languages largely unevaluated. To facilitate evaluation for Indian languages, especially Telugu, which expresses negation not only
through separate words but also through morphemes attached directly to words, we present
the NEGTEL benchmark. As part of this work, we first compile and verify an inventory
of Telugu negative morphemes and introduce NEGSEED, a manually curated set of example
sentences covering the full range of morphological negation patterns in Telugu. Building on
this resource, the NEGTEL benchmark is a test suite that contains five tasks: Negation Detection, Translation, Paraphrase Detection, Sentiment Classification, and Polarity Flipping. These
tasks are designed to evaluate models at increasing levels of linguistic and logical complexity,
starting from whether models can detect the presence of negation, to whether they can preserve
it during translation, distinguish negated sentence pairs, correctly classify sentiment in the presence of negation, and actively manipulate negation by flipping the polarity of a sentence. The
test suite is designed grounded in detailed linguistic analysis and includes annotations of different negation types. We evaluate a diverse set of multilingual LLMs on this benchmark, and
our evaluation reveals that most models struggle significantly with Telugu negation across all
tasks. A recurring finding is that standard automated metrics remain high even when models
fail semantically, highlighting the need for human evaluation when assessing negation-related
capabilities.

Abstract

In multi-armed bandit settings with Markovian rewards, the Gittins index policy is well known to
maximize the expected discounted return and is therefore the optimal arm-selection strategy. In practical
applications, however, the transition dynamics of the underlying Markov chains are seldom available,
making it impossible to compute the Gittins indices analytically. This motivates the use of reinforcement
learning (RL) techniques that balance exploration and exploitation to infer these indices directly from
data.
In this work, we introduce two learning-based approaches for estimating the Gittins index: a tabular method, QGI, and a deep reinforcement learning counterpart, DGN. Both algorithms are built upon
the retirement-formulation perspective of the multi-armed bandit problem. Compared to existing RL
approaches for Gittins index estimation, our methods offer substantially lower computational overhead,
require less memory due to a smaller Q-table in QGI and a reduced replay buffer in DGN, and exhibit faster and more stable empirical convergence to the true index values. These properties make the
proposed algorithms particularly effective for large state spaces and attractive alternatives to current
techniques. As an illustrative application, we show how these methods can be used to learn the optimal
scheduling policy that minimizes mean flow time in a batch-arrival job scheduling system with unknown
service-time distributions.

Abstract

Music's ability to communicate emotions is universally acknowledged, yet the relative contribution
of its constituent components, namely instrumentals, vocals, and lyrics, to perceived emotion remains
insufficiently understood. This thesis systematically investigates this question through a series of computational and perceptual experiments within the continuous valence-arousal (VA) emotion space.
We first establish that a substantial discrepancy exists between lyrics-based emotion annotations and
acoustically derived emotion estimates. Across three datasets spanning English, Hindi, and Telugu, we
find that lyrics-based labels frequently disagree with the emotion conveyed by the music itself, and that
this mismatch is not confined to a single language or cultural context.
To determine where the emotional signal resides, we construct a unified corpus of over 1,500 Englishlanguage music excerpts with continuous human-annotated VA values. Using source separation, we
isolate instrumental and vocal audio stems; lyrics are transcribed and human-verified. Deep learning
models are employed to predict VA values for each component independently. The results reveal a
consistent hierarchy: instrumental features exhibit the strongest alignment with human emotional annotations, followed by vocals, with lyrics trailing substantially. A perceptual validation study with human
participants confirms that these predictions are perceptually grounded.
Quadrant-wise analysis reveals a notable exception: in the low-valence, low-arousal region of the
VA space, all three modalities converge, with lyrics achieving their highest alignment with perceived
emotion. This pattern is unique to this region, suggesting that lyrics play a particularly important role
in conveying sadness-related emotions. Cross-corpus validation against independently derived emotion
sources, including Spotify audio features and crowd-sourced social tags, confirms that both the modality
hierarchy and this convergence pattern generalize beyond the primary corpus.
These findings carry implications for both dataset design and system architecture in music emotion recognition. Datasets annotated from lyrics alone measure textual sentiment rather than perceived
musical emotion, and current multimodal approaches that uniformly weight lyrics are suboptimal. Our
results motivate adaptive, context-sensitive fusion frameworks in which lyrical information is selectively
leveraged where it meaningfully enhances affective interpretation.

Abstract

Breast cancer is among the most prevalent and clinically heterogeneous malignancies
worldwide, and its characterization at multiple biological scales such as radiological
phenotype and genomic architecture represents one of the defining computational challenges
at the interface of information science and biomedicine. Despite significant algorithmic
progress in both medical image analysis and genomic data processing, reliable computational
characterization of breast cancer is constrained by a common and underappreciated
bottleneck, which is the absence of rigorous, standardized benchmarking infrastructure that
enables reproducible evaluation and fair comparison of methods. This thesis addresses two
distinct but complementary computational problems in breast cancer research: (i) the
development and evaluation of benchmark datasets and deep learning frameworks for
mammographic image analysis, and (ii) the systematic benchmarking of long-read sequencing
tools for structural variant detection in germline and somatic cancer contexts. Both problems
are unified by the same underlying bottleneck, which is the absence of rigorous evaluation
infrastructure. The contributions presented here address this bottleneck at both scales. The
first contribution is Mammo-Bench, a large-scale unified mammography benchmark dataset
comprising 19,731 images from 6,500 patients across six geographically diverse publicly
available resources, standardized through a comprehensive preprocessing pipeline and
annotated for five clinical tasks including cancer status, breast density, BI-RADS score,
abnormality type, and molecular subtype. Mammo-Bench is, to the best of our knowledge, the
largest open-access multi-task mammography benchmark currently available and establishes
a reproducible evaluation standard for mammographic computer-aided detection and
diagnosis systems. The second contribution is MammoInsight, a unified multi-task deep
learning framework for mammographic analysis, trained on an extended benchmark of 65,602
images. Built on the MedImageInsight foundation model backbone, MammoInsight
simultaneously performs five clinical classification tasks and lesion segmentation in a single
forward pass, incorporating SAM-guided region-of-interest pooling, cross-view attention
fusion, and ordinal regression losses for clinically ordered tasks. Evaluated against four large
foundation model baselines, MammoInsight achieves state-of-the-art performance on cancer
status prediction, breast density classification, and BI-RADS scoring, and attains a
segmentation Dice score of 0.804, substantially outperforming a fine-tuned SAM-Med2D
baseline (0.734) despite using only 367 million parameters compared to baselines up to 74
times larger. The third contribution is a systematic multi-context evaluation of long-read
structural variant (SV) callers across three experimental paradigms, which include simulated
chromosome 17 data at four coverage levels, germline benchmarking against the GIAB
HG002 Tier 1 truth set, and paired tumor-normal somatic analysis of HER2+ (HCC1954) and
triple-negative breast cancer (HCC1937) cell lines on both ONT PromethION R10 and PacBio
Revio platforms. Six germline callers and four dedicated somatic callers are evaluated using
a novel three-tier benchmarking framework that accounts for systematic differences in VCF
output representation across tools. Key findings include caller-specific coverage thresholds
for reliable SV detection (F1 > 0.85 achievable at 5x for Sniffles2, Kled, and SVDF), nearceiling germline performance at high coverage, and a consistent, substantial performance gap
between dedicated somatic callers and germline callers adapted through normal subtraction in
the cancer context. The thesis establishes benchmarking infrastructure and validates
computational frameworks across the imaging and genomic modalities of breast cancer
characterization and provides guidance for tool selection and evaluation design in both
research and clinical genomics applications.

Abstract

Large Language Models (LLMs) have undergone remarkable growth in capability over recent years,
evolving from text-generation systems into autonomous agents capable of reasoning, planning, execution and performing complex tasks across diverse domains through the use of external tools, memory,
and multi-step interactions with their environment. However, this progress and widespread adoption
has also led to the expansion of associated risks and misuse. As LLMs become increasingly integrated
into real-world systems, ensuring their safety, robustness, and trustworthiness has emerged as a critical
challenge.
This thesis investigates the trustworthiness of LLM systems, with a particular focus on their safety,
reliability, and robustness across settings. First, we examine the fragility of AI-generated text detection
methods in low-resource non-English settings, specifically Hindi language. While existing detection
methods have largely focused on English, the rapid adoption of LLMs across multilingual settings raise
important concerns regarding the reliability of current detection frameworks beyond their original training distributions. To address this gap, we present a comprehensive study on Hindi AI-generated news
detection, introducing AGhi, a large-scale benchmark dataset consisting of human-written and LLMgenerated Hindi news articles and propose AI-Detectability Index for Hindi (ADIhi). Our findings
show that current methods fail to generalize effectively to non-English settings, revealing significant
limitations in the multilingual robustness of AI-generated text detection systems.
Second, we evaluate the ability of Small Language Models (SLMs) for function selection and generation, a critical capability for agentic systems. While large frontier models achieve strong function
calling performance, their deployment is often impractical in resource-constrained or privacy-sensitive
environments. This work presents a systematic evaluation of multiple SLMs on large-scale function generation tasks under zero-shot, few-shot, and fine-tuned settings, including adversarial prompt injection
scenarios. The findings show that many smaller models continue to struggle with consistent function
generation despite improvements from fine-tuning.
Next we investigate the safety and robustness of multi-agent LLM systems, where multiple agents
collaborate to solve complex tasks through communication and coordination. Existing safety evaluations primarily assume single-agent settings and therefore fail to capture the emergent risks introduced
by inter-agent interactions. To address this limitation, we introduce TAMAS (Threats and Attacks in
Multi-Agent Systems), a benchmark specifically designed to evaluate adversarial threats in multi-agent
LLM systems across diverse domains and attack scenarios. The benchmark consists of six attack types spanning five high-stakes domains, and reveals the failure modes of multi-agent systems under adversarial settings. The findings show that multi-agent systems inherit many vulnerabilities from single-agent
setups while also introducing new risks that emerge specifically from inter-agent communication and
coordination.
Collectively, this thesis highlights that the trustworthiness of modern LLM systems cannot be measured solely through improvements in capability or benchmark performance. As these systems become
increasingly integrated into real-world applications, they must be evaluated not only for what they can
achieve under ideal conditions, but also for how reliably and safely they behave under challenging,
adversarial, and diverse deployment settings.
By examining multiple dimensions of trustworthiness across different levels of the LLM ecosystem,
this thesis emphasizes the importance of rigorous evaluation methodologies for identifying hidden failure modes and advancing the development of safer, more robust, and dependable AI systems. In support
of this goal, the thesis contributes novel datasets, benchmarks, metrics, and evaluation frameworks for
systematically studying reliability and safety across diverse real-world deployment scenarios.

Abstract

Contactless fingerprint recognition provides a number of benefits in terms of hygiene and minimizing
sensor-induced artifacts, over traditional contact-based systems, but it is also associated with a number
of challenges that are currently hindering its adoption. The 2D representation of 3D features, in combination with unconstrained illumination, affects the quality of ridges, resulting in inconsistent features.
Also, the lack of interaction information in contactless recognition makes it more prone to presentation
attacks. Current techniques for acquisition, enhancement, and detection of spoofing attacks are mostly
independent and focused on single images, without exploiting any complementary information for improved recognition. As the trend towards contactless biometric recognition is becoming increasingly
necessary for usability and cost-effectiveness in terms of deployment in mobile and security-critical
applications, lower matching accuracy and increased susceptibility to spoofing attacks are significant
challenges. This highlights the need for a method that jointly addresses enhancement, recognition, and
spoof detection while improving cross-modal interoperability.
In our work, we present a unified framework for contactless fingerprint recognition by exploiting
the flash-non-flash acquisition pair as a lightweight active sensing mechanism that extracts complementary illumination information. We thus present the Flash-Non-Flash Fingerphoto (FNF) Database for
a thorough analysis of the illumination impact, and Fusion2Print (F2P), an end-to-end deep learning
framework that learns the optimal enhancement in the spatial, frequency, and color domains to enhance
the quality of the ridge-valley representations and cross-domain recognition accuracy. Furthermore, we
investigate a method for the detection of presentation attacks by exploiting the illumination-dependent
photometric differences between genuine and spoofed fingerprints, and a low-cost 3D pipeline for the
reconstruction of the fingerprint surface geometry from just two images, facilitating the simulation of
digital spoof attacks.
Experimental results show that all components benefit from improvements. F2P obtains AUC of
0.999 and EER of 1.12%. The illumination-aware technique separates real and spoofing samples effectively, and the 3D reconstruction pipeline rebuilds finger geometry and ridge-valley information with
improved contrast. These components constitute a refined unified framework for contactless fingerprint
recognition.

Abstract

Two-wheelers are among the most widely used modes of transportation in many regions of
the world, particularly in developing countries. However, the lack of structural protection for
riders makes them highly vulnerable to road accidents. Understanding and recognising driving
events from vehicle motion data is therefore an important step toward developing intelligent
safety systems, rider assistance technologies, and accident detection mechanisms.
This thesis investigates time-series based deep learning approaches for two-wheeler driving
event recognition using inertial sensor data. A multivariate time-series dataset was collected
using an embedded sensing platform consisting of a tri-axial accelerometer, gyroscope, and GPS
module mounted on a motorcycle. Unlike many existing datasets, the collected data incorpo-
rates diverse riding conditions, including recordings from multiple riders, variations in riding
behaviour, and both daytime and nighttime sessions, providing a more realistic representation
of real-world riding dynamics.
Several temporal deep learning architectures are evaluated for driving event classification, in-
cluding Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), attention-enhanced
recurrent networks, Temporal Convolutional Networks (TCN), and transformer-based models.
To capture both short-term motion patterns and long-range temporal dependencies, a Time-
Series Transformer (TST) architecture is explored and compared against these baseline models.
All models are trained and evaluated using a consistent preprocessing pipeline and hyperpa-
rameter optimisation framework.
Experimental results show that transformer-based temporal modelling achieves competitive
or superior performance compared to recurrent and convolutional approaches, particularly for
challenging manoeuvre-related events such as turns, stops, and bumps. To evaluate practical
feasibility, the trained models are deployed on a Raspberry Pi edge platform, where inference
latency, memory usage, and real-time processing capability are analysed.
The results demonstrate that accurate driving event recognition can be achieved using
lightweight temporal models operating directly on raw sensor streams, enabling the development
of practical on-device safety monitoring systems for two-wheelers.