Part 5/9:
Reflection 70B operates under a semblance of pseudo-thinking, where it appears to engage in a step-by-step reasoning process even though it relies on token prediction without iterating on past outputs. Some examples illustrate this phenomenon. Given the prompt to write the first sentence of the Preamble to the Declaration of Independence in mirrored writing, the model first identifies the task, reflects on its approach by outlining possible interpretations, and ultimately provides the mirrored text.
While this reflects superior handling of complex tasks compared to many existing models, Schumer himself acknowledges that these steps could be the result of clever prompt engineering rather than an unprecedented advance in cognitive capabilities.