Back to Arena

Atlas

by Meta AI FAIR (Izacard et al.)

System Card

OrganizationMeta AI FAIR (Izacard et al.)
Released2022-08
Architecturevector-rag / Contriever retriever + fusion-in-decoder few-shot
DetailsRetrieval-augmented few-shot learner combining a dense retriever (Contriever) with a fusion-in-decoder reader. Pretrained end-to-end so the retriever and reader co-evolve for knowledge-intensive tasks.
Parameters—
Domainrag-retrievalknowledge-graph
Open SourceYes
PaperView Paper
CodeRepository
jmlr-2023few-shotfusion-in-decoderknowledge-intensive

Capability Profile

Benchmark Scores

6 of 14 benchmarks
Data Transparency:1 self-reported5 estimated
Long-Context Retrieval
2/5
RULER
69.941pEstimated
NIAH
no data
LooGLE
no data
LongBench
603pEstimated
∞Bench
no data
Multi-Turn Recall
0/2
LoCoMo
no data
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
56.63pEstimated
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
62.913pEstimated
HotpotQA
50.63pSelf-Reported
Agent Task Memory
0/1
AgentBench-Mem
no data
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
1/1
RAGAS
66.331pEstimated
Sources:arXiv:2208.03299 Table 10 — KILT-filtered HotpotQA EM, full-train; 64-shot EM=34.7Arena estimate — derived from capability profile, not independently verified