HOT RADAR
agentHEAT 26.3°

AI CURATED HOT STORY · 11/21/2026

Tencent Yuanbao and KCL Launch Group Chat Agent Memory Benchmark

2 reports archived2 independent sourcesupdated 11/22/2026, 01:07:07
Synthesis & Latest Updates
2 Sources Cross-Validated

Tencent Yuanbao and the NLP team at King's College London have jointly released GroupAssistBench, the first benchmark for evaluating agent memory in group chat scenarios. Built from 50 real group chats with 80,299 historical messages, the benchmark comprises 2,395 evaluation questions covering single-memory recall, multi-memory integration, and update conflicts. It is the first to split memory capability into two levels: L1 factual recall and L2 open-ended service.

Experiments across 14 configurations reveal that stronger memory recall does not equal stronger service capability. For instance, EverMemOS scores 0.621 on L1, higher than full context's 0.570, but only 0.247 on L2, far below the latter's 0.431. This indicates that retrieval-augmented and external memory approaches generally lag behind full context on open-ended service tasks, offering a critical warning for agent memory system design.

The paper, code, and dataset are all publicly available, which is expected to promote standardization and reproducibility in group chat agent memory research. As AI assistants increasingly integrate into group collaboration, balancing memory recall and service capability will become a core issue for developers, and this work points the way forward for future optimization.

LATEST/GroupAssistBench has drawn attention since its release, with the finding that stronger memory recall does not equal stronger service capability becoming a focal point. The team has open-sourced the paper, code, and dataset, which may spur more optimization solutions for group chat agent memory.

Hourly heat curve

10 fully observed hours
Now
26.3
Peak
27.102:00
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

Coverage timeline

Total 2 reports · Latest first
  1. 量子位 (微信公众号)T2·78 pts
    • 2,395 tasks built from 50 real group chats and 80,299 messages, covering single-memory, multi-memory integration and update-conflict structures
    • Memory evaluation split into L1 factual recall and L2 open-ended service, where L2 has no fixed wording but must respect valid decisions and member constraints
    • Across 14 configurations, higher recall scores did not translate into better service, with retrieval and external-memory setups generally trailing full context on L2