11 papers
Better Call Reward: Reward Hacking as Strategic Abstention in Legal Reasoning Models
Subramanyam Sahoo, Justin Shenk
What happens when a legal AI model learns to look like a lawyer instead of reasoning like one? We fine tune Qwen3-8B with Group Relative Policy Optimisation (GRPO) against a proxy…
Creative Collision: Directorial Persona Steering and Competition in Large Language Models
Subramanyam Sahoo, Justin Shenk
Activation steering has emerged as a powerful tool for shaping the behaviour of large language models at inference time, yet most prior work injects a \emph{single} semantic direct…
Evaluation Cards: An Interpretive Layer for AI Evaluation Reporting
Avijit Ghosh, Anka Reuel, Jenny Chim +45
AI evaluation results are produced at scale but reported inconsistently across leaderboards, model cards, benchmark papers, and company blogs. The cost is interpretive: readers can…
Calibration of Structured Ignorance Certificates for Diagnosing Unknown Unknowns in Reasoning Models
Subramanyam Sahoo
Large language models frequently fail in a characteristic way: rather than acknowledging ignorance, they produce fluent but incorrect answers to questions that lie beyond their kno…
Memetic Capture: A Pluralistic Policy Framework for Governing AI-Driven Cultural Disempowerment
Subramanyam Sahoo
Culture is the most insidious vector of gradual human disempowerment by AI: unlike economic or political displacement, cultural displacement attacks the very preferences and values…
From Knowledge to Action: Outcomes of the 2025 Large Language Model (LLM) Hackathon for Applications in Materials Science and Chemistry
Aritra Roy, Kevin Shen, Andrew MacBride +350
Large language models (LLMs) are rapidly changing how researchers in materials science and chemistry discover, organize, and act on scientific knowledge. This paper analyzes a broa…
Calibration Collapse Under Sycophancy Fine-Tuning: How Reward Hacking Breaks Uncertainty Quantification in LLMs
Subramanyam Sahoo
Modern large language models (LLMs) are increasingly fine-tuned via reinforcement learning from human feedback (RLHF) or related reward optimisation schemes. While such procedures…
Catch Me If You Can: How Smaller Reasoning Models Pretend to Reason with Mathematical Fidelity
Subramanyam Sahoo, Vinija Jain, Saanidhya Vats +4
Current evaluation of mathematical reasoning in language models relies primarily on answer accuracy, potentially masking fundamental failures in logical computation. We introduce a…
Position: The Complexity of Perfect AI Alignment -- Formalizing the RLHF Trilemma
Subramanyam Sahoo, Aman Chadha, Vinija Jain +1
Reinforcement Learning from Human Feedback (RLHF) is widely used for aligning large language models, yet practitioners face a persistent puzzle: improving safety often reduces fair…
The Last Vote: A Multi-Stakeholder Framework for Language Model Governance
Subramanyam Sahoo, Aditi Chhawacharia
As artificial intelligence systems become increasingly powerful and pervasive, democratic societies face unprecedented challenges in governing these technologies while preserving c…
The Good, The Bad, and The Hybrid: A Reward Structure Showdown in Reasoning Models Training
Subramanyam Sahoo
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 hy…