Papers by Supriyo Ghosh
CARMO: Dynamic Criteria Generation for Context Aware Reward Modelling (2025.findings-acl)
Copied to clipboard
Taneesh Gupta, Shivam Shandilya, Xuchao Zhang, Rahul Madhavan, Supriyo Ghosh, Chetan Bansal, Huaxiu Yao, Saravan Rajmohan
| Challenge: | Reward modeling in large language models is susceptible to reward hacking . flawed reward signals often lead to outputs that optimize for spurious correlates . |
| Approach: | They propose a new approach that generates dynamic, context-relevant criteria to ground the reward model prior to producing reward scores. |
| Outcome: | The proposed approach generates dynamic, context-relevant criteria to ground the model prior to producing reward scores. |
TACO-RL: Task Aware Prompt Compression Optimization with Reinforcement Learning (2025.findings-acl)
Copied to clipboard
Shivam Shandilya, Menglin Xia, Supriyo Ghosh, Huiqiang Jiang, Jue Zhang, Qianhui Wu, Victor Rühle, Saravan Rajmohan
| Challenge: | Existing prompt compression techniques rely on sub-optimal metrics such as information entropy or model it as a task-agnostic token classification problem that fails to capture task-specific information. |
| Approach: | They propose a task-aware prompt compression method that leverages existing Transformer encoders and a lightweight REINFORCE algorithm to ensure low latency requirements. |
| Outcome: | The proposed method improves task performance by 8% - 189% on three diverse and challenging tasks over state-of-the-art techniques while satisfying the same compression rate and latency requirements. |
Learning Optimal Message Representations for Agentic Communication (2026.findings-acl)
Copied to clipboard
Shashwat Gupta, Anson Bastos, Mayukh Das, Supriyo Ghosh, Nagarajan Natarajan, Chetan Bansal, Saravan Rajmohan
| Challenge: | Existing approaches lack the intelligence necessary to understand, learn or apply optimal communication representations adaptively. |
| Approach: | They propose to dynamically learn the optimal message representations to enhance agentic performance by using an Expanding Markov Decision Process. |
| Outcome: | The proposed framework improves agentic performance while maintaining efficiency. |