Sign in to view source links and access this dataset
Description
A benchmark dataset for evaluating memory utilization in conversational LLM agents, created by wenyiwy99 and last updated in May 2026. It is the official dataset for the paper 'Beyond Memorization: Benchmarking Memory Utilization in Conversational LLM Agents.' The dataset's motivation stems from analyzing real human-AI dialogues on ShareChat, where 85.3% of memory usage was found not to be direct inquiry.
Use Cases
Benchmarking memory utilization patterns in LLM agents based on the analysis of real human-AI dialogues.
Evaluating agent performance on long-horizon and personalized tasks based on cross-session memory usage.
Studying non-inquiry memory usage in conversational systems based on the finding that 85.3% of usage is not direct inquiry.
Strengths
Based on analysis of real human-AI dialogues from ShareChat.
Provides a specific quantitative finding: 85.3% of memory usage is not direct inquiry.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Provenance
Source
huggingface user wenyiwy99, associated with the paper 'Beyond Memorization: Benchmarking Memory Utilization in Conversational LLM Agents.'
Collection Method
Analysis of real human-AI dialogues on ShareChat.
Freshness
Last updated 2026-05-18 15:24:49.
License is unknown; terms of use must be verified.