Rahul Madhavan

causal inference (3) causal average treatment effect (2) text generation (2) bias mitigation (2) language model (2) controlled text generation (2) toxicity mitigation (2) large language model (2) reinforcement learning from human feedback (1) reward hacking (1) covering intervention (1) reinforcement learning (1) spurious correlation (1) structural causal model (1) time reversal (1) unsupervised feedback (1) log perplexity (1) safety filter (1) average treatment effect (1) causal bandit (1)