Translation Hallucinations
A discussion unfolds about the intriguing phenomenon of translation hallucinations, where a model opts for grammatical correctness over accurate translation when it encounters uncertainty. This leads to various failure modes, highlighting the complexities of machine learning. Listeners will find humor in the community's playful attempts to decipher the cryptic conversation, showcasing the challenges and quirks of the field.In this clip
From this podcast
Machine Learning Street Talk (MLST)
#039 - Lena Voita - NLP
Related Questions
Is the large language model (LLM) in the episode Reasoning Over Complex Documents with DocLLM with Armineh Nourbakhsh - 672 and the clip Instruction Tuning Insights subject to hallucinations?
Is the large language model (LLM) in the episode "Reasoning Over Complex Documents with DocLLM with Armineh Nourbakhsh - 672" and the clip "Instruction Tuning Insights" subject to hallucinations?
How does the language model work in the episode 42 - Generating Sentences by Editing Prototypes, with Kelvin Guu and the clip Model Perplexity Discussion from the episode 42 - Generating Sentences by Editing Prototypes, with Kelvin Guu and the clip Neural Editing?