arXiv subjects

Shreyansh Padarha

Publications and source records attributed to Shreyansh Padarha.

7 recordsLinked to original sources

arXiv2026

Running the Gauntlet: Challenging Agentic Tasks

As agentic systems continue to evolve and are widely deployed in real-world scenarios, there is a growing demand to faithfully evaluate their capabilities. However, current benchmarks are typically built on popular applications with relatively simple tasks and focus on a narrow set of capabilities while overlooking broader dimensions, resulting in saturated performance on modern agents and failing to probe their limitations. To this end, we introduce GauntletBench, a web-based benchmark for evaluating agent generalisation in challenging scenarios, focusing on three underexplored capabilities (temporal perception, graphical understanding, and 3D reasoning), across five less-covered professional applications (Video Editor, Workflow Builder, 3D Modeller, Flight Analyser, and Circuit Designer), each with 27 vision-intensive tasks (135 in total). Our benchmark provides a modular pipeline that comprises an environment compatible with both open- and closed-source agent frameworks, a controlled web-based application, a well-structured task suite, and an automated evaluation engine with diverse metrics. Contrary to widespread expectations, our empirical results reveal that frontier agentic systems remain far from achieving human-level performance. Even the state-of-the-art agent achieves only a 28.2% success rate on our GauntletBench, highlighting the limitations in these overlooked capabilities and generalisation. By comparison, non-expert human annotators achieve over 80% success on our challenging yet feasible tasks, revealing the substantial gap between current agent capabilities and those required for complex real-world scenarios.

cs.LG↗

arXiv2026

Agentic Reinforcement Learning for Search Misaligns Instruction-Tuning

Agentic reinforcement learning (RL) trains large language models to use tools, but its impact on alignment is poorly understood. We study how agentic RL for search affects the alignment of instruction-tuned (IT) models. We find that RL-trained models inherit refusal reasoning by deflecting harmful requests into benign search queries, but this breaks down under a simple diagnostic trigger that elicits a search call before refusal can occur. Under this condition, RL models produce multi-step unsafe search actions and reasoning, reducing search query safety by up to 68.6% in Qwen and Llama models relative to their IT counterparts. The effect generalises across model families, scales, and RL algorithms. To understand why, we identify linear directions in the residual stream that control search query safety, and show that RL training progressively shifts search behaviour toward the harmful end of this direction. We thus propose representation-guided RL training, which adds a reward penalty based on projection toward the harmful search direction. Training on benign data alone, it restores IT-level alignment without reducing task accuracy and requires no additional training data. Together, our work provides the first framework for diagnosing, mechanistically analysing, and mitigating alignment degradation in agentic RL for search.

cs.CL↗

arXiv2026

Evaluating AI-based Scientific Knowledge Synthesis with Epidemiological Systematic Reviews

Systematic literature reviews (SLRs) are a demanding and high-stakes form of scientific knowledge synthesis that remains underspecified as an evaluation setting for large language models (LLMs). We introduce AgentSLR, a large-scale evaluation harness comprising an SLR automation workflow and an expert annotated dataset covering 16,248 articles, designed to test LLM capabilities across the stages of SLRs in epidemiology. Reference annotations were derived from peer-reviewed studies on WHO priority pathogens and produced by domain experts. The harness evaluates each review stage as a separate unit with dedicated metrics enabling targeted failure analysis. We evaluated five frontier reasoning models and found that no single model dominated across all tasks, showing sub-task specialisation often hidden by aggregate benchmarks. Structured data extraction is a major bottleneck, with no model exceeding an average field-level F1 of 0.67. Estimated costs vary substantially, by up to 96 times across evaluated models. Documented failure modes suggest that the evaluated models are not yet reliable enough for unsupervised deployment in epidemiology, where findings can inform public policy.

cs.IR↗

arXiv2026

Strategic Navigation or Stochastic Search? How Agents and Humans Reason Over Document Collections

Multimodal agents offer a promising path to automating complex document-intensive workflows. Yet, a critical question remains: do these agents demonstrate genuine strategic reasoning, or merely stochastic trial-and-error search? To address this, we introduce MADQA, a benchmark of 2,250 human-authored questions grounded in 800 heterogeneous PDF documents. Guided by Classical Test Theory, we design it to maximize discriminative power across varying levels of agentic abilities. To evaluate agentic behaviour, we introduce a novel evaluation protocol measuring the accuracy-effort trade-off. Using this framework, we show that while the best agents can match human searchers in raw accuracy, they succeed on largely different questions and rely on brute-force search to compensate for weak strategic planning. They fail to close the nearly 20% gap to oracle performance, persisting in unproductive loops. We release the dataset and evaluation harness to help facilitate the transition from brute-force retrieval to calibrated, efficient reasoning.

cs.CL↗

arXiv2025

Enhancing Reasoning Capabilities in SLMs with Reward Guided Dataset Distillation

The push to compress and impart the proficiency of Large Language Models (LLMs) into more deployable and efficient Small Language Models (SLMs) has benefited from improvements in knowledge distillation (KD) techniques. These techniques allow a smaller student model to learn from a more capable and larger teacher model's responses. However, distillation often revolves around the student model merely copying the teacher's in-distribution responses, limiting its generalisability. This limitation is amplified on reasoning tasks and can be computationally expensive. In this study, we propose AdvDistill, a reward-guided dataset distillation framework. We utilise multiple generations (responses) from a teacher for each prompt and assign rewards based on rule-based verifiers. These varying and normally distributed rewards serve as weights when training student models. Our methods and their subsequent behavioural analysis demonstrate a significant improvement in student model performance for mathematical and complex reasoning tasks, showcasing the efficacy and benefits of incorporating a rewarding mechanism in dataset distillation processes.

cs.AI↗

arXiv2025

Party Ideologies and Political Polarization-Driven Conflicts: A Study of the Global South

Post-World War II armed conflicts have often been viewed with higher scrutiny in order to avoid a full-scale global war. This scrutiny has led to the establishment of determinants of war such as poverty, inequalities, literacy, and many more. There is a gap that exists in probing countries in the Global South for political party fragmentation and examining ideology-driven polarization's effect on armed conflicts. This paper fills this gap by asking the question: How does political identity-induced polarization affect conflicts in the Global South region? Polarization indices are created based on socially relevant issues and party stances from the V-Party Dataset. Along with control variables, they are tested against the response variables conflict frequency and conflict severity created from the UCDP (Uppsala Conflict Data Program). Through Chow's test, Regional Structural Breaks are found between regions when accounting for polarization-conflict dynamics. A multilevel mixed effects modelling approach is used to create region-specific models to find what types of polarization affect conflict in different geographies and their adherence to normative current developments. The paper highlights that vulnerable regions of the world are prone to higher polarization-induced violence. Modelling estimates indicate polarization of party credo on Minority Rights, Rejection of Political Violence, Religious Principles, and Political Pluralism are strong proponents of cultivated violence. The Global South's inhibitions and slow progress towards development are caused by hindrances from armed conflicts; this paper's results show self-inflicted political instability and fragmentation's influence on these events, making the case for urgency in addressing and building inter-group homogeneity and tolerance.

physics.soc-ph↗

arXiv2023

Data-Driven Dystopia: an uninterrupted breach of ethics

This article discusses the risks and complexities associated with the exponential rise in data and the misuse of data by large corporations. The article presents instances of data breaches and data harvesting practices that violate user privacy. It also explores the concept of "Weapons Of Math Destruction" (WMDs), which refers to big data models that perpetuate inequality and discrimination. The article highlights the need for companies to take responsibility for safeguarding user information and the ethical use of data models, AI, and ML. The article also emphasises the significance of data privacy for individuals in their daily lives and the need for a more conscious and responsible approach towards data management.

cs.CY↗