Back to Arena

GAM

by VectorSpaceLab (BAAI-related)

System Card

OrganizationVectorSpaceLab (BAAI-related)
Released2025-10
Architecturehierarchical-summary / Deep-research memory filesystem
DetailsLLM-based intelligent chunking with semantic boundary detection; builds structured summary + TLDR per chunk; auto-builds hierarchical taxonomies. Multi-modal (text and video) with incremental append and agent trajectory compression.
Parameters—
Domainagent-memorylong-contextlifelong-learning
Open SourceYes
CodeRepository
deep-researchmultimodaltrajectory-compressionhierarchical

Capability Profile

Benchmark Scores

6 of 14 benchmarks
Long-Context Retrieval
1/5
RULER
no data
NIAH
no data
LooGLE
no data
LongBench
no data
∞Bench
7956p
Multi-Turn Recall
1/2
LoCoMo
77.873p
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
76.554p
Multi-Hop QA
2/3
BABILong
8088p
MultiHop-RAG
no data
HotpotQA
70.145p
Agent Task Memory
1/1
AgentBench-Mem
7226p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:GAM (VectorSpaceLab/general-agentic-memory); evaluated on BABILong: Testing the Limits of LLMs with Long-Context Reasoning-in-a-Haystack (AIRI, 2406)GAM (VectorSpaceLab/general-agentic-memory); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)GAM (VectorSpaceLab/general-agentic-memory); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)GAM (VectorSpaceLab/general-agentic-memory); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)GAM (VectorSpaceLab/general-agentic-memory); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)GAM (VectorSpaceLab/general-agentic-memory); evaluated on InfiniteBench: Extending Long Context Evaluation Beyond 100K Tokens (Tsinghua / OpenBMB, 2402)