SalwaMK·May 28What If the Problem in AI Trading Isn’t the Model, But the Reward Function?When people talk about AI in trading, the conversation usually focuses on models:
ML Point·Apr 18LongAct: Training Models to Stay Stable When Context Gets LongThere is a point where language models stop failing apparently but failing due to long context.
Anvitha H·Mar 24Rethinking PPO: My Empirical Study of A*-PO for LLM Self-CorrectionThe Consistency Problem in LLM Reasoning Large Language Models (LLMs) have become surprisingly good at reasoning tasks, but their…
Dixon·Oct 20, 2025Information Gain-Based Policy Optimization: A Simple and Effective Approach for Multi-Turn LLM…How dense intrinsic rewards could revolutionize agentic reinforcement learning for large language models.
Nikhil Verma·Sep 2, 2025Modern Alignment and Post-training of LLMsLLMs, though powerful, often generate unhelpful or unpredictable outputs if left solely to their pretraining data. Post-training alignment…
