생성형ai2 [LLM/NLP] Banishing LLM Hallucinations Requires RethinkingGeneralization TitleBanishing LLM Hallucinations Requires Rethinking GeneralizationLinksarxivhttps://www.lamini.ai/Summary이 연구는 LLM의 일반적인 hallucination problem (사실과 다른 내용을 그럴듯해 보이게 생성해내는 문제) 에 관하여 어떻게 해결할 수 있을지에 대한 방법 및 새로운 모델 구조를 제안한다. 좀 더 구체적으로 답을 구하는 문제들은,What causes hallucinations?Are there other architectures with low generalization error and low hallucinations?Are they computationally feasible?일반적으로 LLM은.. 2024. 7. 8. [LLM/NLP] A Thorough Examination of Decoding Methods in the Era of LLMs TitleA Thorough Examination of Decoding Methods in the Era of LLMsLinksarxiv A Thorough Examination of Decoding Methods in the Era of LLMsDecoding methods play an indispensable role in converting language models from next-token predictors into practical task solvers. Prior research on decoding methods, primarily focusing on task-specific models, may not extend to the current era of general-parxi.. 2024. 7. 7. 이전 1 다음
