Kunvar Thaman, a 26-year-old independent researcher from Chandigarh, India, has gained recognition in the field of machine learning after his solo-authored paper was accepted at the prestigious ICML 2026 conference in Seoul. His paper, “Reward Hacking Benchmark: Measuring Exploits in LLM Agents with Tool Use,” introduces a framework to assess how AI systems exploit shortcuts in multi-step tasks, a growing concern in AI safety research. The study evaluates 13 cutting-edge AI models from leading organizations such as OpenAI and Google, revealing exploit rates between 0% and 13.9%, with safety measures mitigating these behaviors effectively. Thaman’s achievement is particularly remarkable given the dominance of major AI companies and elite universities in the field, making his solo acceptance a rare feat. His work highlights the increasing importance of understanding reward hacking in AI, as models gain more autonomy and tool access. Thaman’s success underscores the potential for independent researchers to make significant contributions to the competitive landscape of machine learning.
Who is Kunvar Thaman, the solo Indian researcher making waves at ICML 2026?
