self-preference bia

large language model (13587) llm evaluation (141) self-recognition capability (1) evaluation alignment (2) controlled experiment (6) human preference (124) bias reduction (48) evaluation bia (15) attention analysis (25) human alignment (58)