not equal, but probabisticly better. Best example: give the agent a tight spec and it will perform better compared with a spec that leaves room for interpretati

not equal, but probabisticly better. Best example: give the agent a tight spec and it will perform better compared with a spec that leaves room for interpretation. This is true for all models, more or less. (purely anecdotal of course)