List your company
ResearchFormerly WizWand

Open-domain Question Answering on HotpotQA (test)

50.18Accuracy (Exact Match)Clean
1.008813.774426.5439.3056Aug 24, 2025Oct 11, 2025Nov 28, 2025Jan 15, 2026Mar 4, 2026Apr 21, 2026Jun 9, 2026
Updated 4mo ago

Evaluation results

MethodLinks
CleanLLM Backbone=GPT-4o, Pipeline=Chunking, Retrieval, and Reranking2026.0650.18———
PoisonedRAG-BBLLM Backbone=GPT-4o, Pipeline=Chunking, Retrieval, and Reranking2026.0649.33——16.71
The RAG ParadoxLLM Backbone=GPT-4o, Pipeline=Chunking, Retrieval, and Reranking2026.0642.15——15.25
Joint-GCGLLM Backbone=GPT-4o, Pipeline=Chunking, Retrieval, and Reranking2026.0639.76——34.75
Deepseek-V3 (671B)Downstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=Compression of top 10 docs, Compressor Model=Deepseek-V3 (671B)2025.0834.6245.6950—
CORE (1B)Downstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=Compression of top 10 docs, Compressor Model=CORE (1B), Compressor Training Backbone=llama3.2-1B-Instruct2025.0834.124548—
Top10 DocumentsDownstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=RAG without compression2025.0833.9544.881,471—
Deepseek-V3 (671B)Downstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=Compression of top 5 docs, Compressor Model=Deepseek-V3 (671B)2025.0833.5944.8348—
CORE (1B)Downstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=Compression of top 5 docs, Compressor Model=CORE (1B), Compressor Training Backbone=llama3.2-1B-Instruct2025.0833.4444.5445—
Top5 DocumentsDownstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=RAG without compression2025.0832.9943.69737—
Top3 DocumentsDownstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=RAG without compression2025.0831.6441.87442—
RECOMP (1B)Downstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=Compression of top 5 docs, Compressor Model=RECOMP (1B), Compressor Training Backbone=llama3.2-1B-Instruct2025.0829.8741.2149—
RECOMP (1B)Downstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=Compression of top 10 docs, Compressor Model=RECOMP (1B), Compressor Training Backbone=llama3.2-1B-Instruct2025.0829.8241.0555—
Top1 DocumentDownstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=RAG without compression2025.0829.238.93147—
llama3.2-1BDownstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=Compression of top 10 docs, Compressor Model=llama3.2-1B2025.0826.6336.6156—
llama3.2-1BDownstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=Compression of top 5 docs, Compressor Model=llama3.2-1B2025.0826.4836.3958—
No RetrievalDownstream LLM (M)=Qwen2.5-14B-Instruct, Retrieval/Compression Setting=No Retrieval2025.0821.0529.480—
EVE-Agent2026.0520.9———
CRCPLLM Backbone=DeepSeek-R1, Pipeline=Chunking, Retrieval, and Reranking2026.0619.78——95.01
CRCPLLM Backbone=LLaMA-2-7B, Pipeline=Chunking, Retrieval, and Reranking2026.0618.93——90.12
CRCPLLM Backbone=Vicuna-7B, Pipeline=Chunking, Retrieval, and Reranking2026.0618.69——90.87
CRCPLLM Backbone=GPT-4o, Pipeline=Chunking, Retrieval, and Reranking2026.0617.39——91.76
CRCPLLM Backbone=Qwen3-Max, Pipeline=Chunking, Retrieval, and Reranking2026.0616.17——92.75
CRCPLLM Backbone=Qwen2.5-7B, Pipeline=Chunking, Retrieval, and Reranking2026.0614.92——91.36
Dr. Zero2026.0511———
Initialsearch=disabled2026.055.7———
Initialsearch=enabled2026.052.9———