arXiv subjects

Subramanyam Sahoo

Publications and source records attributed to Subramanyam Sahoo.

25 records · Page 2Linked to original sources

arXiv2025

The Deepfake Detective: Interpreting Neural Forensics Through Sparse Features and Manifolds

Deepfake detection models have achieved high accuracy in identifying synthetic media, but their decision processes remain largely opaque. In this paper we present a mechanistic interpretability framework for deepfake detection applied to a vision-language model. Our approach combines a sparse autoencoder (SAE) analysis of internal network representations with a novel forensic manifold analysis that probes how the model's features respond to controlled forensic artifact manipulations. We demonstrate that only a small fraction of latent features are actively used in each layer, and that the geometric properties of the model's feature manifold, including intrinsic dimensionality, curvature, and feature selectivity, vary systematically with different types of deepfake artifacts. These insights provide a first step toward opening the "black box" of deepfake detectors, allowing us to identify which learned features correspond to specific forensic artifacts and to guide the development of more interpretable and robust models.

cs.CV↗

arXiv2025

Catch Me If You Can: How Smaller Reasoning Models Pretend to Reason with Mathematical Fidelity

Current evaluation of mathematical reasoning in language models relies primarily on answer accuracy, potentially masking fundamental failures in logical computation. We introduce a diagnostic framework that distinguishes genuine mathematical reasoning from superficial pattern matching through four complementary axes: forward-backward consistency, transitivity coverage, counterfactual sensitivity, and perturbation robustness. Through a case study applying this framework to Qwen3-0.6B on the MenatQA dataset, we reveal a striking disconnect between surface performance and reasoning fidelity. While the model achieves reasonable answer accuracy (70%+), it demonstrates poor backward consistency (15%), limited transitivity coverage (32.2%), and brittle sensitivity to perturbations. Our diagnostics expose reasoning failures invisible to traditional accuracy metrics, suggesting that this small model relies heavily on pattern matching rather than genuine logical computation. While our empirical findings are based on a single 600M-parameter model, the diagnostic framework itself is model-agnostic and generalizable. We release our evaluation protocols to enable the research community to assess reasoning fidelity across different model scales and architectures, moving beyond surface-level accuracy toward verifiable mathematical reasoning.

cs.CL↗

arXiv2025

Position: The Complexity of Perfect AI Alignment -- Formalizing the RLHF Trilemma

Reinforcement Learning from Human Feedback (RLHF) is widely used for aligning large language models, yet practitioners face a persistent puzzle: improving safety often reduces fairness, scaling to diverse populations becomes computationally intractable, and making systems robust often amplifies majority biases. We formalize this tension as the Alignment Trilemma: no RLHF system can simultaneously achieve (i) epsilon-representativeness across diverse human values, (ii) polynomial tractability in sample and compute complexity, and (iii) delta-robustness against adversarial perturbations and distribution shift. Through a complexity-theoretic analysis integrating statistical learning theory and robust optimization, we prove that achieving both representativeness (epsilon <= 0.01) and robustness (delta <= 0.001) for global-scale populations requires Omega(2^{d_context}) operations, which is super-polynomial in the context dimensionality. We show that current RLHF implementations resolve this trilemma by sacrificing representativeness: they collect only 10^3--10^4 samples from homogeneous annotator pools while 10^7--10^8 samples are needed for true global representation. Our framework provides a unified explanation for documented RLHF pathologies including preference collapse, sycophancy, and systematic bias amplification. We conclude with concrete directions for navigating these fundamental trade-offs through strategic relaxations of alignment requirements.

cs.LG↗

arXiv2025

The Horcrux: Mechanistically Interpretable Task Decomposition for Detecting and Mitigating Reward Hacking in Embodied AI Systems

Embodied AI agents exploit reward signal flaws through reward hacking, achieving high proxy scores while failing true objectives. We introduce Mechanistically Interpretable Task Decomposition (MITD), a hierarchical transformer architecture with Planner, Coordinator, and Executor modules that detects and mitigates reward hacking. MITD decomposes tasks into interpretable subtasks while generating diagnostic visualizations including Attention Waterfall Diagrams and Neural Pathway Flow Charts. Experiments on 1,000 HH-RLHF samples reveal that decomposition depths of 12 to 25 steps reduce reward hacking frequency by 34 percent across four failure modes. We present new paradigms showing that mechanistically grounded decomposition offers a more effective way to detect reward hacking than post-hoc behavioral monitoring.

cs.LG↗

arXiv2025

The Good, The Bad, and The Hybrid: A Reward Structure Showdown in Reasoning Models Training

Reward design is central to reinforcement learning from human feedback (RLHF) and alignment research. In this work, we propose a unified framework to study hard, continuous, and hybrid reward structures for fine-tuning large language models (LLMs) on mathematical reasoning tasks. Using Qwen3-4B with LoRA fine-tuning on the GSM8K dataset, we formalize and empirically evaluate reward formulations that incorporate correctness, perplexity, reasoning quality, and consistency. We introduce an adaptive hybrid reward scheduler that transitions between discrete and continuous signals, balancing exploration and stability. Our results show that hybrid reward structures improve convergence speed and training stability over purely hard or continuous approaches, offering insights for alignment via adaptive reward modeling.

cs.LG↗

arXiv2025

The Last Vote: A Multi-Stakeholder Framework for Language Model Governance

As artificial intelligence systems become increasingly powerful and pervasive, democratic societies face unprecedented challenges in governing these technologies while preserving core democratic values and institutions. This paper presents a comprehensive framework to address the full spectrum of risks that AI poses to democratic societies. Our approach integrates multi-stakeholder participation, civil society engagement, and existing international governance frameworks while introducing novel mechanisms for risk assessment and institutional adaptation. We propose: (1) a seven-category democratic risk taxonomy extending beyond individual-level harms to capture systemic threats, (2) a stakeholder-adaptive Incident Severity Score (ISS) that incorporates diverse perspectives and context-dependent risk factors, and (3) a phased implementation strategy that acknowledges the complex institutional changes required for effective AI governance.

cs.CY↗

arXiv2024

Boardwalk Empire: How Generative AI is Revolutionizing Economic Paradigms

The relentless pursuit of technological advancements has ushered in a new era where artificial intelligence (AI) is not only a powerful tool but also a critical economic driver. At the forefront of this transformation is Generative AI, which is catalyzing a paradigm shift across industries. Deep generative models, an integration of generative and deep learning techniques, excel in creating new data beyond analyzing existing ones, revolutionizing sectors from production and manufacturing to finance. By automating design, optimization, and innovation cycles, Generative AI is reshaping core industrial processes. In the financial sector, it is transforming risk assessment, trading strategies, and forecasting, demonstrating its profound impact. This paper explores the sweeping changes driven by deep learning models like Large Language Models (LLMs), highlighting their potential to foster innovative business models, disruptive technologies, and novel economic landscapes. As we stand at the threshold of an AI-driven economic era, Generative AI is emerging as a pivotal force, driving innovation, disruption, and economic evolution on a global scale.

cs.CE↗