“Reasoning models ... enumerate ... then later going back and choosing from that”
Eric Bigelow · "The Cognitive Revolution" · How Agents Decide: Goodfire's Eric Bigelow on Critical Tokens, Phase Shifts, & In-Context Learning
Play from 43:40
43:40
Audio streams from the publisher’s own feed. Verbatim does not host or copy it.
What was described
Describes reasoning models enumerating many possible solution paths then back-tracking to choose the best, increasing diversity within a chain of thought
Said nearby on the same subject
“…we do that through this chain of thought and reasoning and traces… term… serial depth opacity… make sure we can actually still see what the model is thinking”
“…chain of thought reasoning…are there as a band-a-aid for latency…we struggled with how much to narrate…progressive disclosure…does it really need to be there for everybody”