Back to Arena

Hebbia

by Hebbia, Inc.

System Card

OrganizationHebbia, Inc.
Released2020-06
Architectureagentic-workflow / ISD (Inference-Search-Decomposition) Multi-Agent
DetailsHebbia's Matrix product decomposes complex queries into structured subtasks, then runs multiple specialized agents in parallel — one for semantic search, one for tabular extraction, one for legal/financial terminology. Every output cell is fully cited back to the source document, eliminating hallucinations. The ISD architecture routes each subtask to the most suitable underlying model.
Parameters—
Domainrag-retrievalagent-memory
Open SourceNo
WebsiteVisit
financelegalenterprise-searchmulti-agentdocument-analysistransparency

Capability Profile

Benchmark Scores

6 of 14 benchmarks
Long-Context Retrieval
1/5
RULER
no data
NIAH
no data
LooGLE
no data
LongBench
603p
∞Bench
no data
Multi-Turn Recall
1/2
LoCoMo
76.160p
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
79.980p
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
74.677p
HotpotQA
75.674p
Agent Task Memory
1/1
AgentBench-Mem
7226p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:Hebbia vendor documentation; evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)Hebbia vendor documentation; evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)Hebbia vendor documentation; evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)Hebbia vendor documentation; evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)Hebbia vendor documentation; evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)Hebbia vendor documentation; evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)