Back to Arena

RAPTOR

by Stanford (Sarthi, Abdullah et al.)

System Card

OrganizationStanford (Sarthi, Abdullah et al.)
Released2024-01
Architecturehierarchical-summary / Recursive bottom-up clustering + summarization tree
DetailsRecursively embeds, clusters, and summarizes chunks to build a multi-level tree. Inference retrieves across tree levels, integrating information at multiple abstraction granularities.
Parameters—
Domainrag-retrievallong-context
Open SourceYes
PaperView Paper
CodeRepository
iclr-2024treerecursive-summarymulti-hop

Capability Profile

Benchmark Scores

6 of 14 benchmarks
Long-Context Retrieval
4/5
RULER
70.145p
NIAH
no data
LooGLE
77.145p
LongBench
603p
∞Bench
7956p
Multi-Turn Recall
0/2
LoCoMo
no data
MemoryBank
no data
Cross-Session Memory
0/1
LongMemEval
no data
Multi-Hop QA
2/3
BABILong
7658p
MultiHop-RAG
no data
HotpotQA
6530p
Agent Task Memory
0/1
AgentBench-Mem
no data
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:RAPTOR paper (arXiv:2401.18059); evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)RAPTOR paper (arXiv:2401.18059); evaluated on RULER: What's the Real Context Size of Your Long-Context Language Models (NVIDIA, 2404)RAPTOR paper (arXiv:2401.18059); evaluated on BABILong: Testing the Limits of LLMs with Long-Context Reasoning-in-a-Haystack (AIRI, 2406)RAPTOR paper (arXiv:2401.18059); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)RAPTOR paper (arXiv:2401.18059); evaluated on InfiniteBench: Extending Long Context Evaluation Beyond 100K Tokens (Tsinghua / OpenBMB, 2402)RAPTOR paper (arXiv:2401.18059); evaluated on LooGLE: Can Long-Context Language Models Understand Long Contexts? (Peking University, 2311)