Does a fixed relation vocabulary help agent memory?
Tests whether a small fixed set of relations (is_a, part_of, if_then, before and a few more) gives an LLM agent better recall per stored token than letting it write any relation it likes.
"Where did Maya work before she moved to Toronto?"
- data analyst at Northbridge Labs"I was a data analyst at Northbridge Labs when I lived in Waterloo."
- Northbridge Labs at Waterloo"I was a data analyst at Northbridge Labs when I lived in Waterloo."
- Waterloo before Toronto"That was before the move to Toronto."
- Maya does data analyst"I was a data analyst at Northbridge Labs when I lived in Waterloo."
Northbridge Labs, in Waterloo.