7 papers
GRAIL: Gradient-Reweighted Advantages for Reinforcement Learning with Verifiable Rewards
Tej Deep Pala, Vernon Toh, Soujanya Poria
Reinforcement learning with verifiable rewards (e.g. GRPO) is now a common way to improve mathematical reasoning in Large Language Models (LLMs). However, current methods usually b…
LLMs Can't Handle Peer Pressure: Crumbling under Multi-Agent Social Interactions
Maojia Song, Tej Deep Pala, Ruiwen Zhou +5
Large language models (LLMs) are increasingly integrated into multi-agent systems (MAS), where peer interactions shape individual decisions. While prior work has mainly examined co…
Training Vision-Language Process Reward Models for Test-Time Scaling in Multimodal Reasoning: Key Insights and Lessons Learned
Brandon Ong, Tej Deep Pala, Vernon Toh +2
Process Reward Models (PRMs) provide step-level supervision that improves the reliability of reasoning in large language models. While PRMs have been extensively studied in text-ba…
Lessons from Training Grounded LLMs with Verifiable Rewards
Shang Hong Sim, Tej Deep Pala, Vernon Toh +5
Generating grounded and trustworthy responses remains a key challenge for large language models (LLMs). While retrieval-augmented generation (RAG) with citation-based grounding hol…
Error Typing for Smarter Rewards: Improving Process Reward Models with Error-Aware Hierarchical Supervision
Tej Deep Pala, Panshul Sharma, Amir Zadeh +2
Large Language Models (LLMs) are prone to hallucination, especially during multi-hop and reasoning-intensive tasks such as mathematical problem solving. While Outcome Reward Models…
PromptDistill: Query-based Selective Token Retention in Intermediate Layers for Efficient Large Language Model Inference
Weisheng Jin, Maojia Song, Tej Deep Pala +4
As large language models (LLMs) tackle increasingly complex tasks and longer documents, their computational and memory costs during inference become a major bottleneck. To address…