[Paper Discrepancy] Data Inconsistency in Table 1: A-Mem Baseline Metrics vs. Original Paper

[onford]

Hi Mem0 team,

I enjoyed reading your paper "Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory".

While reviewing the baseline comparisons, I noticed a discrepancy in Table 1 regarding the reported performance of the A-Mem system.

In Table 1 of the Mem0 paper, the F1 and BLEU-1 scores for A-Mem are listed as follows:

Question Type	A-Mem's F1 score	A-Mem's BLEU-1 score
Single Hop	27.02	20.09
Multi-Hop	12.14	12.00
Open Domain	44.65	37.06
Temporal	45.85	36.67

However, these values appear to contradict the data reported in the original A-Mem paper (A-MEM: Agentic Memory for LLM Agents, Table 1).

Comparing the two tables, it seems that the results for Single Hop, Multi-Hop, and Open Domain may have been mixed up or mislabeled in the Mem0 paper. For example, the values assigned to Single Hop in Mem0 seem to correspond to a different category (or vice versa) in the original source.

Reference from A-Mem Paper (Table 1):

Could you please verify if this is a transcription error during the compilation of the results?

Thank you!