Abstract

Presidential debates provide a recurring, institu-
tionally comparable record of how candidates prioritize public
issues during election campaigns. This paper presents a longi-
tudinal analysis of policy agendas in U.S. presidential and vice-
presidential debates from 1960 to 2024. We analyze 48 debate
transcripts using a hybrid framework that combines interpretable
keyword-based topic identification with large language model
(LLM)-assisted semantic topic extraction and sentence-level topic
clustering. The analysis identifies durable and shifting com-
ponents of debate agendas across six decades. Foreign policy
and security are especially prominent in the early Cold War
era, domestic economic governance remains consistently central
across the entire period, and healthcare, immigration, climate,
and institutional concerns become increasingly visible in later
decades. In addition to topic prominence, we examine agenda
diversity and transitions between dominant policy domains across
election cycles. The results show that presidential debate agendas
evolve gradually rather than abruptly, while becoming more
topically diverse in recent decades. These findings demonstrate
the value of LLM-assisted methods for historically grounded
analysis of political discourse and for tracing long-term changes
in policy emphasis in democratic elections.

Abstract

Automatic evaluation of large language model (LLM) responses requires not only factual correctness but also clarity, particularly in political question-answering. While recent datasets provide human annotations for clarity and evasion, the impact of prompt design on automatic clarity evaluation remains underexplored. In this paper, we study prompt-based clarity evaluation using the CLARITY dataset from the SemEval 2026 shared task. We compare a GPT-3.5 baseline provided with the dataset against GPT-5.2 evaluated under three prompting strategies: simple prompting, chain-of-thought prompting, and chain-of-thought with few-shot examples. Model predictions are evaluated against human annotations using accuracy and class-wise metrics for clarity and evasion, along with hierarchical exact match. Results show that GPT-5.2 consistently outperforms the GPT-3.5 baseline on clarity prediction, with accuracy improving from 56 percent to 63 percent under chain-of-thought with few-shot prompting. Chain-of-thought prompting yields the highest evasion accuracy at 34 percent, though improvements are less stable across fine-grained evasion categories. We further evaluate topic identification and find that reasoning-based prompting improves accuracy from 60 percent to 74 percent relative to human annotations. Overall, our findings indicate that prompt design reliably improves high-level clarity evaluation, while fine-grained evasion and topic detection remain challenging despite structured reasoning prompts.

Abstract

Abstract-- Two-wheelers account for a disproportionately
high share of road fatalities in the Global South. Research
on two-wheeler rider behavior, however, lags far behind fourwheelers, where multimodal datasets have driven major advances in Advanced Driver Assistance Systems (ADAS). To
address this gap, we present the MOtorized TwO-wheeler Rider
(MOTOR) dataset, the first large-scale, multi-view, multimodal
resource dedicated to two-wheelers in dense, unstructured
traffic. MOTOR comprises 1,629 sequences (25+ hours of video
data) collected from 16 riders and integrates synchronized
front, rear, and helmet videos, rider eye-gaze from wearable
trackers, on-road audio, and telemetry (GPS, accelerometer,
gyroscope). Rich annotations capture traffic context, rider state,
12 riding maneuvers spanning conventional and unconventional
behaviors, and legality labels (Legal, Illegal, Unspecified). We
benchmark rider behavior recognition and maneuver legality classification using state-of-the-art video action recognition backbones (CNN and Transformer-based), extended with
multimodal fusion, and find that combining RGB, gaze, and
telemetry consistently yields the best performance. MOTOR
thus provides a unique foundation for advancing safety-critical
understanding of two-wheeler riding. It offers the research
community a benchmark to develop and evaluate models for
behavior analysis, legality-aware prediction, and intelligent
transportation systems. Dataset and code is available at https:
//varuniiith.github.io/MOTOR-Dataset/

Abstract

NA

Abstract

The advent of 6G is expected to enable many use cases which may rely on accurate knowledge of the location and orientation of user equipment (UE). The conventional localization methods suffer from limitations such as synchronization and high power consumption required for multiple active anchors. This can be mitigated by utilizing a large dimensional passive reconfigurable intelligent surface (RIS). This paper presents a novel low-complexity approach for the estimation of 5D pose (i.e. 3D location and 2D orientation) of a UE in near-field RIS-assisted multiple-input multiple-output (MIMO) systems. The proposed approach exploits the symmetric arrangement of uniform planar array of RIS and uniform linear array of UE to decouple the 5D problem into five 1D sub-problems. Further, we solve these sub-problems using a total least squares ESPRIT inspired approach to obtain closed-form solutions.

Abstract

Abstract
Accurately representing the precipitation extremes is a core mandate for both flood risk
assessment and water resource planning in India. This paper evaluates the performance of 36
CMIP6 Global Climate Models in simulating the key characteristics of Indian precipitation by
using a bias-corrected, multi-criteria evaluation framework. The daily precipitation outputs from
the CMIP6 historical simulations (1975-2014) were bias-corrected against the 0.25° gridded
observations provided by the India Meteorological Department through the Daily Quantile
Mapping technique. Two representative precipitation indices have been analyzed: a) Annual Total
Precipitation (PRCPTOTAL), reflecting the mean-state water availability; b) the 99th percentile
precipitation intensity (R99p), regarding extreme rainfall relevant to flood risk. The performance
of the models was evaluated using multiple complementary metrics, such as Pearson correlation,
root mean square error, normalized standard deviation, and the Taylor Skill Score. Then, an
MCDM approach based on the Technique for Order of Preference by Similarity to Ideal Solution
(TOPSIS) was used in order to derive objective rankings of the CMIP6 models for both indices.
Results indicate substantial inter-model variability even after bias correction. The best
performance for PRCPTOTAL was seen for the model EC-Earth3-AerChem, while MIROC6 was
found to be the best among those showing extreme precipitation intensity, R99p. Results highlight
that the best performing GCMs vary with applications, and different models provide the best
performance for mean precipitation and extreme intensity simulations. The developed framework
here offers a transparent and reproducible platform for the selection of suitable CMIP6 models
for hydrological and climate impact studies over India and allows for an easy extension to more
analyses, introducing new indices and bias-correction methods in future works.
Keywords: CMIP6 GCMs; Extreme Precipitation; Bias Correction; MCDM ranking; Climate
model evaluation; Quantile Mapping

Abstract

LLM-as-a-Judge frameworks have emerged as scalable alternatives
to human evaluation, with recent work reporting 70-80% agreement
with human evaluators. However, overall agreement can mask systematic
attribute-level misalignment--a critical concern for educational
validity. We investigate this phenomenon in automated essay evaluation
for English language learners, where the assessment quality depends on
the pedagogically appropriate weighting of criteria such as cohesion, syntax,
and conventions. Through systematic evaluation of six LLMs on
1100 ELL essays, we find that even when most models achieve a reasonable
overall quality agreement (ρ = 0.6-0.74), they exhibit severe
attribution biases. Direct prompting of LLaMA-3.1-8B reveals 18.8%
syntax underweighting and 18.7% grammar overweighting (11% mean
divergence), systematically showing different associations with linguistic
complexity over surface correctness. To address these misalignments, we
propose knowledge distillation based on LLM reasoning traces with explicit
interpretability mechanisms: attribution weighting and consistency
regularization. Our best interpretability-optimized model (AD) outperforms
direct LLM prompting in both prediction accuracy (Spearman ρ =
0.763 vs 0.51, +50% improvement) and attribution alignment (3.2% vs
11.0% mean divergence, -71% reduction), while being more computationally
efficient. The findings establish knowledge distillation as a superior
method to direct prompting for educational assessment, enabling computationally
efficient, accurate, and evaluation-rubric-aligned automated
feedback for English language learners at scale.

Abstract

have emerged as a scalable alternative to expensive
human evaluations. Although, LLM
judges demonstrate 70-80% agreement with
human evaluators, their robustness under semantically
equivalent prompt variations remain
underexplored. Through systematic evaluation
of 8 models across 4 NLG tasks using 10 semantically
equivalent paraphrases per prompt
(≈ 115000 evaluations), we identify a critical
accuracy-robustness gap: attribute verifiability
affects the robustness more than model choice,
with factually verifiable attributes achieving
0.71 accuracy versus 0.19 for subjective attributes.
Our investigations discover three key
insights: 1) Task structure characteristics influence
the robustness and in turn accuracy, 2)
Attribute verifiability as the strongest predictor factually
verifiable attribute achieve 0.71 accuracy
versus 0.19 for subjective attributes,
3) No single winning model-smallest model
(LLaMA-3.1-8B) exhibits second-best performance,
while the strongest model (LLaMA-
4) from the same family significantly lag behind,
thus demonstrating that general capability
improvements do not necessarily result in
evaluation robustness. With these findings, we
propose a diagnostic framework grounded in
attribute verifiability that enables principled decisions
about evaluation automation. Our work
establishes new standards for assessing LLM
judge reliability beyond simple accuracy metrics.

Abstract

Pattern separation enables discrimination between similar experiences and relies on hippocampal function. The Mnemonic Similarity Task (MST) indexes this process using the Lure Discrimination (LDI). Recent studies, on the other hand, suggest MST performance is influenced by perceptual processing. We investigated the extent of mnemonic processes in MST performance across increasing delays between encoding and retention phases (0 minutes, 30 minutes, 8 hours, 24 hours). LDI declined in the 8-hour and 24-hour groups, with significantly worse lure discrimination especially for low similarity lures. Response patterns revealed a bias, shorter delays showed an "Old" response bias, whereas longer delays were characterized by a "New" response bias. Additionally, decreasing lure similarity was associated with increased "New" responses. Together, these findings suggest that the MST is sensitive to mnemonic processing over short delays, however performance at longer delays is constrained and may reflect non-mnemonic factors.

Abstract

Urban cities have been intensely prone to floods during extreme rainfall events and water
scarcity issues during dry periods in recent years. In this context, identifying rainwater har
vesting potential (RWHP) regions in urban environments provides a sustainable approach
to mitigate both urban flooding and water security, thereby improving urban stormwater
management. Geospatial mapping of RWHP has tried to consider various hydrometeoro
logical, topographical and other geospatial datasets, but integrating socio-economic factors
over urban environments has not beenexplored much. Thepresentstudyintegrated remote
sensing and hydrological-based information, such as slope, soil type, drainage density,
geomorphology, topographic wetness index (TWI), land use land cover (LULC), rainfall,
runoff coefficient, proximity to roads, and proximity to settlements for geospatial mapping
of RWHpotential zones for Hyderabad city using multi-criteria decision analysis (MCDA)
and weighted overlay analysis (WOA). The resulting RWH potential map indicates that
80.20% of the area falls within the "low" potential category, 17.53% as "moderate", 2.0% as
"very low", and only 0.25% as "high" potential, mainly in the southeastern portion near the
Hussain Sagar outlet. These categories are spatially verified using Sentinel-2 LULC and
Google Earth imagery to assess the qualitative plausibility of the mapped RWH potential
zones. Northwestern areas, with loamy soils and mild slopes, demonstrate suitability for
rooftop collection and percolation structures, highlighting the effectiveness of the proposed
modelling framework for sustainable stormwater management for urban environments.