Awesome-LLM-Uncertainty-Reliability-Robustness

Why we included this project

Anyone planning to put a large language model into production needs to know when it will be wrong. This reading list collects papers and resources on uncertainty estimation, calibration, hallucination, truthfulness, and robustness under distribution shift or adversarial input, organized by subtopic so you can go straight to the part that matters for your use case. The calibration and confidence sections are particularly useful for teams working on customer-facing answers, medical or financial text, or automated reasoning pipelines, where deciding when to defer to a human is part of the job. Researchers and graduate students get a map of open problems, and practitioners can use the reliability and robustness sections to anticipate failure modes. It is a curated index rather than runnable software, so its value lies in the editorial organization and the breadth of linked sources.