Comment by narmiouh
17 hours ago
The most interesting part to me is the "Group accuracy by Model" section, because it underscores that a single agent having all the relevant information consistently scores significantly higher than a group of agents with parts of the information.
Is it fair to then infer that when decisions are to be made, single agent environments are going to make them better than multi-agent if the relevant information can fit into a single agents context window?
> Is it fair to then infer that when decisions are to be made, single agent environments are going to make them better than multi-agent if the relevant information can fit into a single agents context window?
Context window for most frontier models is 1 million tokens. They all start to lose their minds around 300K, if not sooner.
they all operate on the same assumption, that a single token has a single meaning and that meaning doesn't change as more information is added. So regardless of size, context poisoning is a near certainty approaching 1 as the context grows. Few tasks are so clinical that they include zero ambiguity in the context chain.
[flagged]