List your company

Document Question Answering on Qasper

40.8AccuracyFP8
-1.38249.568820.5231.4712Nov 3, 2025Nov 8, 2025Nov 13, 2025Nov 18, 2025Nov 23, 2025Nov 28, 2025Dec 3, 2025

Evaluation results

MethodLinks
FP8Backbone=Llama 3.1 8B, Compression Ratio (CR)=2, Evaluation Protocol=2-shot2025.1140.8—
kvtc16×Backbone=Llama 3.1 8B, Compression Ratio (CR)=18-22, Evaluation Protocol=2-shot2025.1140.7—
VanillaBackbone=Llama 3.1 8B, Compression Ratio (CR)=1, Evaluation Protocol=2-shot2025.1140.4—
GEAR2-bitBackbone=Llama 3.1 8B, Compression Ratio (CR)=5, Evaluation Protocol=2-shot2025.1140.4—
kvtc8×Backbone=Llama 3.1 8B, Compression Ratio (CR)=9-10, Evaluation Protocol=2-shot2025.1140.1—
kvtc32×Backbone=Llama 3.1 8B, Compression Ratio (CR)=34-44, Evaluation Protocol=2-shot2025.1139.4—
KIVI2-bitBackbone=Llama 3.1 8B, Compression Ratio (CR)=5, Evaluation Protocol=2-shot2025.1139.1—
kvtc8×Backbone=MN-Minitron 8B, Compression Ratio (CR)=10-11, Evaluation Protocol=2-shot2025.1139.1—
TOVABackbone=Llama 3.1 8B, Compression Ratio (CR)=8, Evaluation Protocol=2-shot2025.1138.6—
kvtc16×Backbone=MN-Minitron 8B, Compression Ratio (CR)=17-21, Evaluation Protocol=2-shot2025.1138.6—
GEAR2-bitBackbone=Mistral NeMo 12B, Compression Ratio (CR)=5, Evaluation Protocol=2-shot2025.1138.6—
VanillaBackbone=Mistral NeMo 12B, Compression Ratio (CR)=1, Evaluation Protocol=2-shot2025.1138.4—
FP8Backbone=MN-Minitron 8B, Compression Ratio (CR)=2, Evaluation Protocol=2-shot2025.1138.3—
VanillaBackbone=MN-Minitron 8B, Compression Ratio (CR)=1, Evaluation Protocol=2-shot2025.1138.2—
GEAR2-bitBackbone=MN-Minitron 8B, Compression Ratio (CR)=5, Evaluation Protocol=2-shot2025.1138.2—
KIVI2-bitBackbone=MN-Minitron 8B, Compression Ratio (CR)=5, Evaluation Protocol=2-shot2025.1138.2—
KIVI2-bitBackbone=Mistral NeMo 12B, Compression Ratio (CR)=5, Evaluation Protocol=2-shot2025.1138.2—
kvtc64×Backbone=MN-Minitron 8B, Compression Ratio (CR)=53-95, Evaluation Protocol=2-shot2025.1138.1—
kvtc64×Backbone=Mistral NeMo 12B, Compression Ratio (CR)=51-87, Evaluation Protocol=2-shot2025.1138—
FP8Backbone=Mistral NeMo 12B, Compression Ratio (CR)=2, Evaluation Protocol=2-shot2025.1137.9—
kvtc64×Backbone=Llama 3.1 8B, Compression Ratio (CR)=60-88, Evaluation Protocol=2-shot2025.1137.8—
kvtc32×Backbone=MN-Minitron 8B, Compression Ratio (CR)=32-46, Evaluation Protocol=2-shot2025.1137.7—
kvtc8×Backbone=Mistral NeMo 12B, Compression Ratio (CR)=10-11, Evaluation Protocol=2-shot2025.1137.6—
kvtc16×Backbone=Mistral NeMo 12B, Compression Ratio (CR)=17-20, Evaluation Protocol=2-shot2025.1137.6—
kvtc32×Backbone=Mistral NeMo 12B, Compression Ratio (CR)=31-43, Evaluation Protocol=2-shot2025.1137.5—
TOVABackbone=Mistral NeMo 12B, Compression Ratio (CR)=8, Evaluation Protocol=2-shot2025.1136—
xKVBackbone=Llama 3.1 8B, Compression Ratio (CR)=1-5, Evaluation Protocol=2-shot2025.1135.6—
xKVBackbone=MN-Minitron 8B, Compression Ratio (CR)=1-5, Evaluation Protocol=2-shot2025.1134.5—
H2OBackbone=Llama 3.1 8B, Compression Ratio (CR)=8, Evaluation Protocol=2-shot2025.1134.3—
TOVABackbone=MN-Minitron 8B, Compression Ratio (CR)=8, Evaluation Protocol=2-shot2025.1133.9—
xKVBackbone=Mistral NeMo 12B, Compression Ratio (CR)=1-5, Evaluation Protocol=2-shot2025.1133.5—
H2OBackbone=MN-Minitron 8B, Compression Ratio (CR)=8, Evaluation Protocol=2-shot2025.1130—
H2OBackbone=Mistral NeMo 12B, Compression Ratio (CR)=8, Evaluation Protocol=2-shot2025.1129.5—
BookRAGBaseline Type=Our proposed2025.120.5520.611
DocETLBaseline Type=Layout segmented RAG2025.120.4230.504
Layout + VanillaBaseline Type=Conventional RAG2025.120.4070.446
Vanilla RAGBaseline Type=Conventional RAG2025.120.4060.444
RAPTORBaseline Type=Graph-based RAG2025.120.3940.441
BM25Baseline Type=Conventional RAG2025.120.3810.425
GraphRAG-LocalBaseline Type=Graph-based RAG2025.120.3590.392
GraphRankerBaseline Type=Layout segmented RAG2025.120.3290.376
MM-VanillaBaseline Type=Layout segmented RAG2025.120.2790.293
Tree-TraverseBaseline Type=Layout segmented RAG2025.120.2730.321
GraphRAG-GlobalBaseline Type=Graph-based RAG2025.120.240.241