List your company
ResearchFormerly WizWand

Text-based Question Answering on HotpotQA

21Exact Match (EM)Full Context
4.0488.44912.8517.251Jun 9, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Full ContextGeneration LLM=Mistral-7B-Instruct2026.062144.9—1,701
Qwen3-Emb-0.6B-ftGeneration LLM=Mistral-7B-Instruct, k=5, Fine-tuned=true2026.0618.642.580.5267
Latent MemoryGeneration LLM=Mistral-7B-Instruct, k=5, Backbone=LLaMA-3.2-1B-Instruct2026.0617.537.886.676
Qwen3-Emb-0.6B-ftGeneration LLM=Mistral-7B-Instruct, k=2, Fine-tuned=true2026.0616.338.459.7127
BM25 RetrievalGeneration LLM=Mistral-7B-Instruct, k=52026.0615.937.865.5255
Qwen3-Emb-0.6BGeneration LLM=Mistral-7B-Instruct, k=52026.0615.136.670.1256
Latent MemoryGeneration LLM=Mistral-7B-Instruct, k=2, Backbone=LLaMA-3.2-1B-Instruct2026.0614.834.862.246
Dense RetrievalGeneration LLM=Mistral-7B-Instruct, k=52026.0613.93462.4244
BM25 RetrievalGeneration LLM=Mistral-7B-Instruct, k=22026.0613.833.845.5120
Qwen3-Emb-0.6BGeneration LLM=Mistral-7B-Instruct, k=22026.0612.632.750.1124
Qwen3-Emb-0.6B-ftGeneration LLM=Mistral-7B-Instruct, k=1, Fine-tuned=true2026.0612.332.237.980
Latent MemoryGeneration LLM=Mistral-7B-Instruct, k=1, Backbone=LLaMA-3.2-1B-Instruct2026.0612.330.434.836
xRAGGeneration LLM=Mistral-7B-Instruct2026.0612.125.3—42
Dense RetrievalGeneration LLM=Mistral-7B-Instruct, k=22026.0611.330.242.2117
BM25 RetrievalGeneration LLM=Mistral-7B-Instruct, k=12026.061129.630.276
Qwen3-Emb-0.6BGeneration LLM=Mistral-7B-Instruct, k=12026.0610.328.433.279
Dense RetrievalGeneration LLM=Mistral-7B-Instruct, k=12026.06926.427.975
LLMLinguaGeneration LLM=Mistral-7B-Instruct, Compression Rate=20%2026.068.927.6—406
LLMLinguaGeneration LLM=Mistral-7B-Instruct, Compression Rate=10%2026.066.822.6—217
CLaRa-16xGeneration LLM=Mistral-7B-Instruct, k=52026.065.215.355.8119
LLMLinguaGeneration LLM=Mistral-7B-Instruct, Compression Rate=5%2026.064.717.9—117