Back to Arena

SID AI

by SID (YC)

System Card

OrganizationSID (YC)
Released2023-06
Architectureagentic-workflow / agentic retrieval model
DetailsSID developed SID-1, an agentic retrieval model that searches, reads results, and iteratively refines queries (multi-hop retrieval) rather than performing a single embedding lookup. SID-1 is 1.8x more accurate than embedding-only search across general, finance, science, and legal domains, and 24x faster than agentic retrieval built on frontier LLMs like Gemini or GPT. Offered as a hosted retrieval API.
Parameters—
Domainrag-retrievalagent-memory
Open SourceNo
WebsiteVisit
agentic-retrievalmulti-hopiterative-searchAPISID-1

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
78.579p
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
77.159p
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
72.454p
HotpotQA
77.184p
Agent Task Memory
1/1
AgentBench-Mem
7226p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:SID AI vendor documentation; evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)SID AI vendor documentation; evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)SID AI vendor documentation; evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)SID AI vendor documentation; evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)SID AI vendor documentation; evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)SID AI vendor documentation; evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)