neuron-db claims similar accuracy with far lower token cost
showed 100% accuracy in a for 1-hop, 2-hop, and 3-hop questions. A hop means one step in following connected pieces of information to reach an answer. The result claims stayed fixed at 2, no matter how deep the question was or how large the memory became.
Compared with putting a full markdown dump into the model, matched or beat the accuracy while using about 60 times fewer tokens. The markdown dump approach fell to 83% accuracy at larger size because the model got confused while reasoning over a 67,000-token input.
Key points
- claims 100% accuracy on 1-hop, 2-hop, and 3-hop questions.
- stayed fixed at 2 across different depths and memory sizes.
- The claimed was about 60 times lower than a markdown dump.
- The large markdown dump approach reportedly dropped to 83% accuracy.
- The item is useful for thinking about cheaper in .